qs_backtest/

report.rs

//! Backtest reporting — trade log, aggregate statistics, and enhanced analytics.
//!
//! This module provides per-trade results, per-position summaries, and rich
//! aggregate statistics including risk-adjusted metrics (Sharpe, Sortino, Calmar),
//! streak analysis, duration stats, monthly returns, and breakdowns by symbol,
//! group, side, and close reason.

use std::collections::{BTreeMap, BTreeSet, HashMap};

use chrono::NaiveDateTime;
use serde::{Deserialize, Serialize};

use qs_core::CostKind;
use qs_core::types::{CloseReason, GroupId, PositionId, Side};

use crate::artifacts::{
    CloseEvent, CompletedPosition, CostEvent, ExecutionMetadata, FutureBacktestArtifacts,
    NetPnlOutcome, OpenPositionSnapshot, PendingOrderLifecycleEvent, PendingOrderLifecycleState,
    PendingOrderSnapshot, RecordedFill,
};
use crate::evaluation::{
    EvaluationOptions, EvaluationReport, EvaluationRequest, ExcursionInput,
    ExecutionDiagnosticsInput, LifecycleCounts, OutcomeClassification, PositionCostsInput,
    PositionDimensions, PositionOutcome, PositionSide, evaluate,
};
use crate::ledger::{ActionDisposition, ActionDispositionStatus};
use crate::mtm::MtmOutputSummary;
use crate::portfolio::EquityPoint;

// ─── Serde helper for f64 fields that may be INFINITY or NaN ────────────────

/// Serializes non-finite f64 (INFINITY, NEG_INFINITY, NaN) as JSON null.
mod finite_f64 {
    use serde::{self, Deserialize, Deserializer, Serializer};

    pub fn serialize<S>(value: &f64, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        if value.is_finite() {
            serializer.serialize_f64(*value)
        } else {
            serializer.serialize_none()
        }
    }

    pub fn deserialize<'de, D>(deserializer: D) -> Result<f64, D::Error>
    where
        D: Deserializer<'de>,
    {
        // Accept either a number or null (null → 0.0).
        let opt = Option::<f64>::deserialize(deserializer)?;
        Ok(opt.unwrap_or(0.0))
    }
}

// ─── TradeResult ────────────────────────────────────────────────────────────

/// Result of a single closed trade (or partial close).
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct TradeResult {
    pub position_id: PositionId,
    pub symbol: String,
    pub side: Side,
    pub entry_price: f64,
    pub exit_price: f64,
    pub size: f64,
    /// Realized profit and loss for this close, already net of `commission` and `swap`.
    pub pnl: f64,
    /// Account-currency commission settled on this row and already subtracted from `pnl`.
    ///
    /// This close's own exit commission always settles here. The row that fully closes a position additionally settles that position's entry commission, because entry commission belongs to the position rather than to any one close. Summing this column over a position therefore reproduces [`CompletedPosition::commission_total`], while an individual row of a partially closed position carries only its own exit commission.
    #[serde(default)]
    pub commission: f64,
    /// Account-currency swap settled on this row and already subtracted from `pnl`.
    ///
    /// Swap accrues to a position rather than to a close, so it settles entirely on the row that fully closes the position and is zero on every other row.
    #[serde(default)]
    pub swap: f64,
    /// Profit and loss before `commission` and `swap`, present only when either was settled.
    #[serde(default)]
    pub gross_pnl: Option<f64>,
    pub open_ts: NaiveDateTime,
    pub close_ts: NaiveDateTime,
    pub close_reason: CloseReason,
    /// Group this position belonged to (for per-group reporting).
    #[serde(default)]
    pub group: Option<GroupId>,
}

// ─── SubsetStats ────────────────────────────────────────────────────────────

/// Reusable statistics block computed from any subset of trades.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct SubsetStats {
    /// Number of trades in this subset.
    pub total_trades: usize,
    /// Trades with positive P&L.
    pub winning_trades: usize,
    /// Trades with negative P&L.
    pub losing_trades: usize,
    /// Trades with exactly zero P&L.
    pub breakeven_trades: usize,
    /// Sum of all P&L.
    pub total_pnl: f64,
    /// Sum of positive P&L.
    pub gross_profit: f64,
    /// Sum of absolute negative P&L.
    pub gross_loss: f64,
    /// winning / total (0.0 if no trades).
    pub win_rate: f64,
    /// gross_profit / gross_loss (INFINITY if no losers, 0.0 if no trades).
    #[serde(with = "finite_f64")]
    pub profit_factor: f64,
    /// gross_profit / winning_trades (0.0 if no winners).
    pub avg_win: f64,
    /// gross_loss / losing_trades (0.0 if no losers).
    pub avg_loss: f64,
    /// avg_win / avg_loss (INFINITY if no losers, 0.0 if no winners).
    #[serde(with = "finite_f64")]
    pub win_loss_ratio: f64,
    /// (win_rate * avg_win) - (loss_rate * avg_loss). Expected P&L per trade.
    pub expectancy: f64,
    /// Largest single winning trade P&L.
    pub largest_win: f64,
    /// Largest single losing trade (as positive number).
    pub largest_loss: f64,
    /// Commission settled on this subset's rows.
    ///
    /// A subset that contains every row of a position reports that position's complete commission, because the row that fully closes a position settles its entry commission. A subset that splits a position reports only the commission settled on the rows it kept.
    #[serde(default)]
    pub commission: f64,
    /// Swap settled on this subset's rows, under the same subset rule as `commission`.
    #[serde(default)]
    pub swap: f64,
    /// Subset profit and loss before `commission` and `swap`, present only when either was settled.
    #[serde(default)]
    pub gross_pnl: Option<f64>,
}

impl SubsetStats {
    /// Compute statistics from a slice of trade references.
    pub fn from_trades(trades: &[&TradeResult]) -> Self {
        let total_trades = trades.len();
        let winning_trades = trades.iter().filter(|t| t.pnl > 0.0).count();
        let losing_trades = trades.iter().filter(|t| t.pnl < 0.0).count();
        let breakeven_trades = trades.iter().filter(|t| t.pnl == 0.0).count();

        let total_pnl: f64 = trades.iter().map(|t| t.pnl).sum();
        let gross_profit: f64 = trades.iter().filter(|t| t.pnl > 0.0).map(|t| t.pnl).sum();
        let gross_loss: f64 = trades
            .iter()
            .filter(|t| t.pnl < 0.0)
            .map(|t| t.pnl.abs())
            .sum();

        let win_rate = if total_trades > 0 {
            winning_trades as f64 / total_trades as f64
        } else {
            0.0
        };

        let profit_factor = if gross_loss > 0.0 {
            gross_profit / gross_loss
        } else if gross_profit > 0.0 {
            f64::INFINITY
        } else {
            0.0
        };

        let avg_win = if winning_trades > 0 {
            gross_profit / winning_trades as f64
        } else {
            0.0
        };

        let avg_loss = if losing_trades > 0 {
            gross_loss / losing_trades as f64
        } else {
            0.0
        };

        let win_loss_ratio = if avg_loss > 0.0 {
            avg_win / avg_loss
        } else if avg_win > 0.0 {
            f64::INFINITY
        } else {
            0.0
        };

        let loss_rate = if total_trades > 0 {
            losing_trades as f64 / total_trades as f64
        } else {
            0.0
        };
        let expectancy = (win_rate * avg_win) - (loss_rate * avg_loss);

        let largest_win = trades
            .iter()
            .filter(|t| t.pnl > 0.0)
            .map(|t| t.pnl)
            .fold(0.0_f64, f64::max);

        let largest_loss = trades
            .iter()
            .filter(|t| t.pnl < 0.0)
            .map(|t| t.pnl.abs())
            .fold(0.0_f64, f64::max);

        let commission: f64 = trades.iter().map(|t| t.commission).sum();
        let swap: f64 = trades.iter().map(|t| t.swap).sum();
        let gross_pnl = (commission != 0.0 || swap != 0.0).then_some(total_pnl + commission + swap);

        Self {
            total_trades,
            winning_trades,
            losing_trades,
            breakeven_trades,
            total_pnl,
            gross_profit,
            gross_loss,
            win_rate,
            profit_factor,
            avg_win,
            avg_loss,
            win_loss_ratio,
            expectancy,
            largest_win,
            largest_loss,
            commission,
            swap,
            gross_pnl,
        }
    }

    /// Compute statistics from an owned slice (convenience wrapper).
    pub fn from_trade_slice(trades: &[TradeResult]) -> Self {
        let refs: Vec<&TradeResult> = trades.iter().collect();
        Self::from_trades(&refs)
    }
}

// ─── StreakStats ─────────────────────────────────────────────────────────────

/// Consecutive win/loss streak analysis.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct StreakStats {
    /// Maximum consecutive winning trades.
    pub max_consecutive_wins: u32,
    /// Maximum consecutive losing trades.
    pub max_consecutive_losses: u32,
    /// Current streak (positive = wins, negative = losses, 0 = no trades or breakeven).
    pub current_streak: i32,
}

impl StreakStats {
    /// Compute streak statistics from a chronologically-ordered trade log.
    pub fn from_trades(trades: &[&TradeResult]) -> Self {
        let mut current_streak: i32 = 0;
        let mut max_wins: u32 = 0;
        let mut max_losses: u32 = 0;

        for trade in trades {
            if trade.pnl > 0.0 {
                if current_streak > 0 {
                    current_streak += 1;
                } else {
                    current_streak = 1;
                }
                max_wins = max_wins.max(current_streak as u32);
            } else if trade.pnl < 0.0 {
                if current_streak < 0 {
                    current_streak -= 1;
                } else {
                    current_streak = -1;
                }
                max_losses = max_losses.max(current_streak.unsigned_abs());
            } else {
                // Breakeven resets streak.
                current_streak = 0;
            }
        }

        Self {
            max_consecutive_wins: max_wins,
            max_consecutive_losses: max_losses,
            current_streak,
        }
    }

    /// Compute FutureQuote streaks from completed campaigns in deterministic close order.
    pub fn from_completed_positions(positions: &[CompletedPosition]) -> Self {
        let mut ordered: Vec<&CompletedPosition> = positions.iter().collect();
        ordered.sort_by(|left, right| {
            left.close_ts
                .cmp(&right.close_ts)
                .then_with(|| left.position_id.cmp(&right.position_id))
                .then_with(|| left.open_ts.cmp(&right.open_ts))
        });

        let mut current_streak: i32 = 0;
        let mut max_wins: u32 = 0;
        let mut max_losses: u32 = 0;
        for position in ordered {
            match position.outcome {
                NetPnlOutcome::Win => {
                    current_streak = if current_streak > 0 {
                        current_streak + 1
                    } else {
                        1
                    };
                    max_wins = max_wins.max(current_streak as u32);
                }
                NetPnlOutcome::Loss => {
                    current_streak = if current_streak < 0 {
                        current_streak - 1
                    } else {
                        -1
                    };
                    max_losses = max_losses.max(current_streak.unsigned_abs());
                }
                NetPnlOutcome::Breakeven => current_streak = 0,
            }
        }

        Self {
            max_consecutive_wins: max_wins,
            max_consecutive_losses: max_losses,
            current_streak,
        }
    }
}

// ─── RiskMetrics ────────────────────────────────────────────────────────────

/// Risk-adjusted return metrics.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct RiskMetrics {
    /// Annualized Sharpe ratio. `None` if fewer than 2 trades.
    pub sharpe_ratio: Option<f64>,
    /// Annualized Sortino ratio (penalizes only downside). `None` if fewer than 2 trades or no downside.
    pub sortino_ratio: Option<f64>,
    /// Calmar ratio: annualized_return / max_drawdown_pct. `None` if max_drawdown is zero or duration < 1 day.
    pub calmar_ratio: Option<f64>,
    /// total_pnl / max_drawdown. `None` if max_drawdown is zero.
    pub return_on_max_drawdown: Option<f64>,
    /// Largest peak-to-trough drawdown in absolute terms.
    pub max_drawdown: f64,
    /// Largest peak-to-trough drawdown as percentage of peak.
    pub max_drawdown_pct: f64,
    /// Duration of the longest drawdown period (seconds).
    pub max_drawdown_duration_secs: Option<i64>,
}

impl RiskMetrics {
    /// Compute risk metrics from the trade log, equity curve, and drawdown values.
    fn compute(
        trade_log: &[TradeResult],
        initial_balance: f64,
        max_drawdown: f64,
        max_drawdown_pct: f64,
        equity_curve: &[(NaiveDateTime, f64)],
        total_pnl: f64,
    ) -> Self {
        let return_on_max_drawdown = if max_drawdown > 0.0 {
            Some(total_pnl / max_drawdown)
        } else {
            None
        };

        // Compute per-trade returns (relative to balance before the trade).
        let mut balance = initial_balance;
        let mut returns = Vec::with_capacity(trade_log.len());
        for trade in trade_log {
            let ret = if balance.abs() > f64::EPSILON {
                trade.pnl / balance
            } else {
                0.0
            };
            returns.push(ret);
            balance += trade.pnl;
        }

        let sharpe_ratio = compute_sharpe(&returns, trade_log);
        let sortino_ratio = compute_sortino(&returns, trade_log);
        let calmar_ratio = compute_calmar(trade_log, initial_balance, total_pnl, max_drawdown_pct);

        // Max drawdown duration.
        let max_drawdown_duration_secs = compute_max_dd_duration(equity_curve, initial_balance);

        Self {
            sharpe_ratio,
            sortino_ratio,
            calmar_ratio,
            return_on_max_drawdown,
            max_drawdown,
            max_drawdown_pct,
            max_drawdown_duration_secs,
        }
    }
}

/// Annualized Sharpe: mean(returns) / std(returns) * sqrt(trades_per_year).
fn compute_sharpe(returns: &[f64], trade_log: &[TradeResult]) -> Option<f64> {
    if returns.len() < 2 {
        return None;
    }
    let n = returns.len() as f64;
    let mean = returns.iter().sum::<f64>() / n;
    let variance = returns.iter().map(|r| (r - mean).powi(2)).sum::<f64>() / (n - 1.0);
    let std_dev = variance.sqrt();
    if std_dev < f64::EPSILON {
        return None;
    }
    let trades_per_year = annualization_factor(trade_log)?;
    Some((mean / std_dev) * trades_per_year.sqrt())
}

/// Annualized Sortino: mean(returns) / downside_dev * sqrt(trades_per_year).
fn compute_sortino(returns: &[f64], trade_log: &[TradeResult]) -> Option<f64> {
    if returns.len() < 2 {
        return None;
    }
    let n = returns.len() as f64;
    let mean = returns.iter().sum::<f64>() / n;
    let downside_sq_sum: f64 = returns
        .iter()
        .filter(|&&r| r < 0.0)
        .map(|r| r.powi(2))
        .sum();
    let downside_count = returns.iter().filter(|&&r| r < 0.0).count();
    if downside_count == 0 {
        return None; // No downside deviation — Sortino undefined.
    }
    let downside_dev = (downside_sq_sum / n).sqrt();
    if downside_dev < f64::EPSILON {
        return None;
    }
    let trades_per_year = annualization_factor(trade_log)?;
    Some((mean / downside_dev) * trades_per_year.sqrt())
}

/// Calmar: annualized_return / max_drawdown_pct.
fn compute_calmar(
    trade_log: &[TradeResult],
    initial_balance: f64,
    total_pnl: f64,
    max_drawdown_pct: f64,
) -> Option<f64> {
    if trade_log.len() < 2 || max_drawdown_pct < f64::EPSILON {
        return None;
    }
    let first_ts = trade_log.first()?.open_ts;
    let last_ts = trade_log.last()?.close_ts;
    let duration = last_ts - first_ts;
    let days = duration.num_seconds() as f64 / 86400.0;
    if days < 1.0 {
        return None;
    }
    let years = days / 365.25;
    let annualized_return = (total_pnl / initial_balance) / years;
    Some(annualized_return / max_drawdown_pct)
}

/// Estimate trades per year from the backtest span and trade count.
fn annualization_factor(trade_log: &[TradeResult]) -> Option<f64> {
    if trade_log.len() < 2 {
        return None;
    }
    let first_ts = trade_log.first()?.open_ts;
    let last_ts = trade_log.last()?.close_ts;
    let duration = last_ts - first_ts;
    let days = duration.num_seconds() as f64 / 86400.0;
    if days < f64::EPSILON {
        return None;
    }
    Some(trade_log.len() as f64 / (days / 365.25))
}

/// Compute max drawdown duration from the equity curve.
fn compute_max_dd_duration(
    equity_curve: &[(NaiveDateTime, f64)],
    initial_balance: f64,
) -> Option<i64> {
    if equity_curve.is_empty() {
        return None;
    }

    let mut peak = initial_balance;
    let mut peak_ts = equity_curve[0].0;
    let mut max_dd_dur_secs: i64 = 0;

    for &(ts, bal) in equity_curve {
        if bal >= peak {
            // Recovered or new peak — measure duration of the drawdown that just ended.
            let dur = (ts - peak_ts).num_seconds();
            if dur > max_dd_dur_secs {
                max_dd_dur_secs = dur;
            }
            peak = bal;
            peak_ts = ts;
        }
    }

    // Check unrecovered drawdown at end.
    if let Some(&(last_ts, last_bal)) = equity_curve.last()
        && last_bal < peak
    {
        let dur = (last_ts - peak_ts).num_seconds();
        if dur > max_dd_dur_secs {
            max_dd_dur_secs = dur;
        }
    }

    if max_dd_dur_secs > 0 {
        Some(max_dd_dur_secs)
    } else {
        None
    }
}

// ─── DurationStats ──────────────────────────────────────────────────────────

/// Trade or completed-campaign holding time statistics (all values in seconds).
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DurationStats {
    /// Average holding time across all trades (seconds).
    pub avg_duration_secs: i64,
    /// Shortest trade duration (seconds).
    pub min_duration_secs: i64,
    /// Longest trade duration (seconds).
    pub max_duration_secs: i64,
    /// Average holding time for winning trades (seconds).
    pub avg_winner_duration_secs: i64,
    /// Average holding time for losing trades (seconds).
    pub avg_loser_duration_secs: i64,
}

impl DurationStats {
    /// Compute duration statistics from trades. Returns `None` if no trades.
    pub fn from_trades(trades: &[&TradeResult]) -> Option<Self> {
        if trades.is_empty() {
            return None;
        }

        let durations: Vec<i64> = trades
            .iter()
            .map(|t| (t.close_ts - t.open_ts).num_seconds())
            .collect();

        let total: i64 = durations.iter().sum();
        let avg_duration_secs = total / durations.len() as i64;
        let min_duration_secs = *durations.iter().min().unwrap();
        let max_duration_secs = *durations.iter().max().unwrap();

        let winner_durations: Vec<i64> = trades
            .iter()
            .filter(|t| t.pnl > 0.0)
            .map(|t| (t.close_ts - t.open_ts).num_seconds())
            .collect();
        let avg_winner_duration_secs = if winner_durations.is_empty() {
            0
        } else {
            winner_durations.iter().sum::<i64>() / winner_durations.len() as i64
        };

        let loser_durations: Vec<i64> = trades
            .iter()
            .filter(|t| t.pnl < 0.0)
            .map(|t| (t.close_ts - t.open_ts).num_seconds())
            .collect();
        let avg_loser_duration_secs = if loser_durations.is_empty() {
            0
        } else {
            loser_durations.iter().sum::<i64>() / loser_durations.len() as i64
        };

        Some(Self {
            avg_duration_secs,
            min_duration_secs,
            max_duration_secs,
            avg_winner_duration_secs,
            avg_loser_duration_secs,
        })
    }

    /// Compute FutureQuote holding times once per completed campaign.
    pub fn from_completed_positions(positions: &[CompletedPosition]) -> Option<Self> {
        if positions.is_empty() {
            return None;
        }

        let duration =
            |position: &CompletedPosition| (position.close_ts - position.open_ts).num_seconds();
        let durations: Vec<i64> = positions.iter().map(duration).collect();
        let winner_durations: Vec<i64> = positions
            .iter()
            .filter(|position| position.outcome == NetPnlOutcome::Win)
            .map(duration)
            .collect();
        let loser_durations: Vec<i64> = positions
            .iter()
            .filter(|position| position.outcome == NetPnlOutcome::Loss)
            .map(duration)
            .collect();
        let average = |values: &[i64]| {
            if values.is_empty() {
                0
            } else {
                values.iter().sum::<i64>() / values.len() as i64
            }
        };

        Some(Self {
            avg_duration_secs: average(&durations),
            min_duration_secs: *durations
                .iter()
                .min()
                .expect("completed positions are non-empty"),
            max_duration_secs: *durations
                .iter()
                .max()
                .expect("completed positions are non-empty"),
            avg_winner_duration_secs: average(&winner_durations),
            avg_loser_duration_secs: average(&loser_durations),
        })
    }
}

// ─── MonthlyReturn ──────────────────────────────────────────────────────────

/// P&L summary for one calendar month.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct MonthlyReturn {
    /// Year (e.g. 2026).
    pub year: i32,
    /// Month (1–12).
    pub month: u32,
    /// Sum of P&L for rows closed in this month.
    pub pnl: f64,
    /// Number of close events (Legacy) or completed campaigns (FutureQuote).
    pub trade_count: usize,
    /// Balance at end of month.
    pub ending_balance: f64,
}

/// Compute monthly returns from a chronologically-ordered trade log.
fn compute_monthly_returns(trade_log: &[TradeResult], initial_balance: f64) -> Vec<MonthlyReturn> {
    if trade_log.is_empty() {
        return Vec::new();
    }

    // Group by (year, month).
    let mut groups: Vec<((i32, u32), Vec<&TradeResult>)> = Vec::new();
    for trade in trade_log {
        let key = (trade.close_ts.date().year(), trade.close_ts.date().month());
        if let Some(last) = groups.last_mut()
            && last.0 == key
        {
            last.1.push(trade);
            continue;
        }
        groups.push((key, vec![trade]));
    }

    let mut balance = initial_balance;
    groups
        .into_iter()
        .map(|((year, month), trades)| {
            let pnl: f64 = trades.iter().map(|t| t.pnl).sum();
            let trade_count = trades.len();
            balance += pnl;
            MonthlyReturn {
                year,
                month,
                pnl,
                trade_count,
                ending_balance: balance,
            }
        })
        .collect()
}

fn compute_monthly_returns_from_completed(
    positions: &[CompletedPosition],
    initial_balance: f64,
) -> Vec<MonthlyReturn> {
    let mut ordered: Vec<&CompletedPosition> = positions.iter().collect();
    ordered.sort_by(|left, right| {
        left.close_ts
            .cmp(&right.close_ts)
            .then_with(|| left.position_id.cmp(&right.position_id))
            .then_with(|| left.net_pnl.total_cmp(&right.net_pnl))
    });

    let mut groups: BTreeMap<(i32, u32), (f64, usize)> = BTreeMap::new();
    for position in ordered {
        let key = (
            position.close_ts.date().year(),
            position.close_ts.date().month(),
        );
        let (pnl, count) = groups.entry(key).or_default();
        *pnl += position.net_pnl;
        *count += 1;
    }

    let mut balance = initial_balance;
    groups
        .into_iter()
        .map(|((year, month), (pnl, trade_count))| {
            balance += pnl;
            MonthlyReturn {
                year,
                month,
                pnl,
                trade_count,
                ending_balance: balance,
            }
        })
        .collect()
}

// We need chrono's Datelike for year()/month().
use chrono::Datelike;

// ─── PositionSummary ────────────────────────────────────────────────────────

/// Aggregated result for one position across all its close events.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct PositionSummary {
    pub position_id: PositionId,
    pub symbol: String,
    pub side: Side,
    pub group: Option<GroupId>,
    /// Campaign entry basis, size-weighted across the basis recorded at each close event.
    /// This preserves a shared basis while reflecting average-cost changes after scale-ins.
    pub entry_price: f64,
    /// Size-weighted average exit price across all closes.
    pub avg_exit_price: f64,
    /// Total size across all close events.
    pub original_size: f64,
    /// Number of close events (partial + final).
    pub close_count: usize,
    /// Net P&L across all close events.
    pub net_pnl: f64,
    /// Ordered list of close reasons.
    pub close_reasons: Vec<CloseReason>,
    /// When the position was opened.
    pub open_ts: NaiveDateTime,
    /// When the last close event occurred.
    pub final_close_ts: NaiveDateTime,
    /// Total holding duration in seconds.
    pub duration_seconds: i64,
}

impl PositionSummary {
    /// Build a summary from all trade results for one position.
    pub fn from_trades(trades: &[&TradeResult]) -> Self {
        assert!(
            !trades.is_empty(),
            "PositionSummary requires at least one trade"
        );

        let first = trades[0];
        let net_pnl: f64 = trades.iter().map(|t| t.pnl).sum();
        let original_size: f64 = trades.iter().map(|t| t.size).sum();

        let entry_price = if original_size > 0.0 {
            trades.iter().map(|t| t.entry_price * t.size).sum::<f64>() / original_size
        } else {
            first.entry_price
        };
        let avg_exit_price = if original_size > 0.0 {
            trades.iter().map(|t| t.exit_price * t.size).sum::<f64>() / original_size
        } else {
            0.0
        };

        let final_close_ts = trades.iter().map(|t| t.close_ts).max().unwrap();
        let close_reasons: Vec<CloseReason> = trades.iter().map(|t| t.close_reason).collect();

        Self {
            position_id: first.position_id.clone(),
            symbol: first.symbol.clone(),
            side: first.side,
            group: first.group.clone(),
            entry_price,
            avg_exit_price,
            original_size,
            close_count: trades.len(),
            net_pnl,
            close_reasons,
            open_ts: first.open_ts,
            final_close_ts,
            duration_seconds: (final_close_ts - first.open_ts).num_seconds(),
        }
    }

    /// Net position result: positive P&L.
    pub fn is_winner(&self) -> bool {
        self.net_pnl > 0.0
    }

    /// Net position result: negative P&L.
    pub fn is_loser(&self) -> bool {
        self.net_pnl < 0.0
    }
}

// ─── CloseReasonStats ───────────────────────────────────────────────────────

/// Statistics for one close reason.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct CloseReasonStats {
    /// The close reason.
    pub reason: CloseReason,
    /// How many trades closed for this reason.
    pub count: usize,
    /// Sum of P&L for trades with this reason.
    pub total_pnl: f64,
    /// Average P&L per trade for this reason.
    pub avg_pnl: f64,
    /// Fraction of all trades that closed for this reason.
    pub percentage: f64,
}

/// Compute per-close-reason statistics.
fn compute_close_reason_stats(trade_log: &[TradeResult]) -> Vec<CloseReasonStats> {
    if trade_log.is_empty() {
        return Vec::new();
    }

    let total_count = trade_log.len();
    let mut by_reason: HashMap<CloseReason, Vec<f64>> = HashMap::new();
    for trade in trade_log {
        by_reason
            .entry(trade.close_reason)
            .or_default()
            .push(trade.pnl);
    }

    let mut stats: Vec<CloseReasonStats> = by_reason
        .into_iter()
        .map(|(reason, pnls)| {
            let count = pnls.len();
            let total_pnl: f64 = pnls.iter().sum();
            CloseReasonStats {
                reason,
                count,
                total_pnl,
                avg_pnl: total_pnl / count as f64,
                percentage: count as f64 / total_count as f64,
            }
        })
        .collect();

    // Sort by count descending, then reason name for deterministic ties.
    stats.sort_by(|left, right| {
        right
            .count
            .cmp(&left.count)
            .then_with(|| left.reason.to_string().cmp(&right.reason.to_string()))
    });
    stats
}

// ─── BacktestResult ─────────────────────────────────────────────────────────

/// Aggregate backtest statistics produced by [`BacktestRunner`](crate::runner::BacktestRunner).
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct BacktestResult {
    // ── Existing fields (preserved for backward compatibility) ───────
    /// Starting account balance.
    pub initial_balance: f64,
    /// Final account balance (initial + total realized P&L).
    pub final_balance: f64,
    /// Sum of all realized P&L.
    pub total_pnl: f64,
    /// Number of completed trades (full + partial closes).
    pub total_trades: usize,
    /// Number of trades with positive P&L.
    pub winning_trades: usize,
    /// Number of trades with negative P&L.
    pub losing_trades: usize,
    /// `winning_trades / total_trades` (0.0 if no trades).
    pub win_rate: f64,
    /// Largest peak-to-trough drawdown in absolute terms.
    pub max_drawdown: f64,
    /// Largest peak-to-trough drawdown as a percentage of the peak.
    pub max_drawdown_pct: f64,
    /// Sum of winning P&L / abs(sum of losing P&L). `f64::INFINITY` if no losers.
    #[serde(with = "finite_f64")]
    pub profit_factor: f64,
    /// Equity value at each trade close: `(timestamp, balance)`.
    pub equity_curve: Vec<(NaiveDateTime, f64)>,
    /// Full trade log (one entry per close event).
    pub trade_log: Vec<TradeResult>,

    /// Full aggregate stats in SubsetStats form.
    pub summary: SubsetStats,

    /// Stats broken down by symbol.
    pub per_symbol: BTreeMap<String, SubsetStats>,

    /// Stats broken down by group (empty if no positions were grouped).
    pub per_group: BTreeMap<GroupId, SubsetStats>,

    /// Stats for long (Buy) trades.
    pub long_stats: SubsetStats,
    /// Stats for short (Sell) trades.
    pub short_stats: SubsetStats,

    /// Breakdown by close reason, sorted by count descending.
    pub per_close_reason: Vec<CloseReasonStats>,

    /// Consecutive win/loss streak analysis.
    pub streaks: StreakStats,

    /// Risk-adjusted return metrics (Sharpe, Sortino, Calmar, drawdown duration).
    pub risk_metrics: RiskMetrics,

    /// Trade holding time statistics. `None` if no trades.
    pub duration_stats: Option<DurationStats>,

    /// Monthly P&L breakdown.
    pub monthly_returns: Vec<MonthlyReturn>,

    // ── Per-position aggregation ────────────────────────────────────
    /// Per-position summaries. Legacy reports aggregate close-event rows; FutureQuote
    /// reports include only campaigns present in `completed_positions`.
    pub positions: Vec<PositionSummary>,

    /// Number of unique legacy positions or completed FutureQuote campaigns.
    pub total_positions: usize,
    /// Winning legacy positions or epsilon-classified FutureQuote campaigns.
    pub winning_positions: usize,
    /// Losing legacy positions or epsilon-classified FutureQuote campaigns.
    pub losing_positions: usize,
    /// Position-level win rate: winning_positions / total_positions.
    pub position_win_rate: f64,

    // ── FutureQuoteV1 additive artifacts ─────────────────────────────
    #[serde(default)]
    pub future_format_version: Option<u32>,
    #[serde(default)]
    pub execution_metadata: Option<ExecutionMetadata>,
    #[serde(default)]
    pub recorded_fills: Vec<RecordedFill>,
    #[serde(default)]
    pub action_dispositions: Vec<ActionDisposition>,
    #[serde(default)]
    pub close_events: Vec<CloseEvent>,
    #[serde(default)]
    pub completed_positions: Vec<CompletedPosition>,
    #[serde(default)]
    pub open_position_snapshots: Vec<OpenPositionSnapshot>,
    #[serde(default)]
    pub pending_order_snapshots: Vec<PendingOrderSnapshot>,
    #[serde(default)]
    pub pending_order_lifecycle: Vec<PendingOrderLifecycleEvent>,
    #[serde(default)]
    pub mtm_equity_curve: Vec<EquityPoint>,
    #[serde(default)]
    pub mtm_output_summary: MtmOutputSummary,
    #[serde(default)]
    pub mtm_max_drawdown: Option<f64>,
    #[serde(default)]
    pub mtm_max_drawdown_pct: Option<f64>,
    #[serde(default)]
    pub provider_evaluation: Option<EvaluationReport>,
    /// Unfiltered normalized outcomes used to build provider evaluation.
    #[serde(default)]
    pub provider_positions: Vec<PositionOutcome>,
    /// Total account-currency commission charged across every entry and exit fill.
    #[serde(default)]
    pub total_commission: f64,
    /// Total account-currency swap charged across every rollover.
    #[serde(default)]
    pub total_swap: f64,
    /// Realized profit and loss before commission and swap, present only when a cost was charged.
    #[serde(default)]
    pub gross_pnl: Option<f64>,
    /// Commission and swap charges in application order.
    #[serde(default)]
    pub cost_events: Vec<CostEvent>,
}

impl BacktestResult {
    /// Build aggregate statistics from a trade log.
    pub fn from_trade_log(initial_balance: f64, trade_log: Vec<TradeResult>) -> Self {
        let total_pnl: f64 = trade_log.iter().map(|t| t.pnl).sum();
        let final_balance = initial_balance + total_pnl;
        let total_trades = trade_log.len();

        let winning_trades = trade_log.iter().filter(|t| t.pnl > 0.0).count();
        let losing_trades = trade_log.iter().filter(|t| t.pnl < 0.0).count();

        let win_rate = if total_trades > 0 {
            winning_trades as f64 / total_trades as f64
        } else {
            0.0
        };

        let gross_profit: f64 = trade_log
            .iter()
            .filter(|t| t.pnl > 0.0)
            .map(|t| t.pnl)
            .sum();
        let gross_loss: f64 = trade_log
            .iter()
            .filter(|t| t.pnl < 0.0)
            .map(|t| t.pnl.abs())
            .sum();
        let profit_factor = if gross_loss > 0.0 {
            gross_profit / gross_loss
        } else if gross_profit > 0.0 {
            f64::INFINITY
        } else {
            0.0
        };

        // Build equity curve and compute max drawdown.
        let mut balance = initial_balance;
        let mut equity_curve = Vec::with_capacity(trade_log.len());
        let mut peak = initial_balance;
        let mut max_drawdown = 0.0_f64;
        let mut max_drawdown_pct = 0.0_f64;

        for trade in &trade_log {
            balance += trade.pnl;
            equity_curve.push((trade.close_ts, balance));

            if balance > peak {
                peak = balance;
            }
            let dd = peak - balance;
            if dd > max_drawdown {
                max_drawdown = dd;
            }
            let dd_pct = if peak > 0.0 { dd / peak } else { 0.0 };
            if dd_pct > max_drawdown_pct {
                max_drawdown_pct = dd_pct;
            }
        }

        // ── SubsetStats (overall summary) ───────────────────────────
        let all_refs: Vec<&TradeResult> = trade_log.iter().collect();
        let summary = SubsetStats::from_trades(&all_refs);

        // ── Per-symbol breakdown ────────────────────────────────────
        let mut by_symbol: HashMap<String, Vec<&TradeResult>> = HashMap::new();
        for trade in &trade_log {
            by_symbol
                .entry(trade.symbol.clone())
                .or_default()
                .push(trade);
        }
        let per_symbol: BTreeMap<String, SubsetStats> = by_symbol
            .iter()
            .map(|(sym, trades)| (sym.clone(), SubsetStats::from_trades(trades)))
            .collect();

        // ── Per-group breakdown ─────────────────────────────────────
        let mut by_group: HashMap<GroupId, Vec<&TradeResult>> = HashMap::new();
        for trade in &trade_log {
            if let Some(ref g) = trade.group {
                by_group.entry(g.clone()).or_default().push(trade);
            }
        }
        let per_group: BTreeMap<GroupId, SubsetStats> = by_group
            .iter()
            .map(|(g, trades)| (g.clone(), SubsetStats::from_trades(trades)))
            .collect();

        // ── Per-side breakdown ──────────────────────────────────────
        let longs: Vec<&TradeResult> = trade_log.iter().filter(|t| t.side == Side::Buy).collect();
        let shorts: Vec<&TradeResult> = trade_log.iter().filter(|t| t.side == Side::Sell).collect();
        let long_stats = SubsetStats::from_trades(&longs);
        let short_stats = SubsetStats::from_trades(&shorts);

        // ── Per-close-reason breakdown ──────────────────────────────
        let per_close_reason = compute_close_reason_stats(&trade_log);

        // ── Streak analysis ─────────────────────────────────────────
        let streaks = StreakStats::from_trades(&all_refs);

        // ── Risk metrics ────────────────────────────────────────────
        let risk_metrics = RiskMetrics::compute(
            &trade_log,
            initial_balance,
            max_drawdown,
            max_drawdown_pct,
            &equity_curve,
            total_pnl,
        );

        // ── Duration stats ──────────────────────────────────────────
        let duration_stats = DurationStats::from_trades(&all_refs);

        // ── Monthly returns ─────────────────────────────────────────
        let monthly_returns = compute_monthly_returns(&trade_log, initial_balance);

        // ── Position summaries ──────────────────────────────────────
        let mut by_position: HashMap<PositionId, Vec<&TradeResult>> = HashMap::new();
        for trade in &trade_log {
            by_position
                .entry(trade.position_id.clone())
                .or_default()
                .push(trade);
        }
        let mut positions: Vec<PositionSummary> = by_position
            .values()
            .map(|trades| PositionSummary::from_trades(trades))
            .collect();
        // Stable lifecycle tie-breaks keep full-result serialization deterministic.
        positions.sort_by(|left, right| {
            left.open_ts
                .cmp(&right.open_ts)
                .then_with(|| left.final_close_ts.cmp(&right.final_close_ts))
                .then_with(|| left.position_id.cmp(&right.position_id))
        });

        let total_positions = positions.len();
        let winning_positions = positions.iter().filter(|p| p.is_winner()).count();
        let losing_positions = positions.iter().filter(|p| p.is_loser()).count();
        let position_win_rate = if total_positions > 0 {
            winning_positions as f64 / total_positions as f64
        } else {
            0.0
        };

        Self {
            initial_balance,
            final_balance,
            total_pnl,
            total_trades,
            winning_trades,
            losing_trades,
            win_rate,
            max_drawdown,
            max_drawdown_pct,
            profit_factor,
            equity_curve,
            trade_log,
            summary,
            per_symbol,
            per_group,
            long_stats,
            short_stats,
            per_close_reason,
            streaks,
            risk_metrics,
            duration_stats,
            monthly_returns,
            positions,
            total_positions,
            winning_positions,
            losing_positions,
            position_win_rate,
            future_format_version: None,
            execution_metadata: None,
            recorded_fills: Vec::new(),
            action_dispositions: Vec::new(),
            close_events: Vec::new(),
            completed_positions: Vec::new(),
            open_position_snapshots: Vec::new(),
            pending_order_snapshots: Vec::new(),
            pending_order_lifecycle: Vec::new(),
            mtm_equity_curve: Vec::new(),
            mtm_output_summary: MtmOutputSummary::default(),
            mtm_max_drawdown: None,
            mtm_max_drawdown_pct: None,
            provider_evaluation: None,
            provider_positions: Vec::new(),
            total_commission: 0.0,
            total_swap: 0.0,
            gross_pnl: None,
            cost_events: Vec::new(),
        }
    }

    /// Build legacy close-event rows plus completed-position FutureQuoteV1 statistics
    /// and additive artifacts using the backward-compatible all-sections report.
    pub fn from_future_artifacts(artifacts: FutureBacktestArtifacts) -> Self {
        Self::from_future_artifacts_with_options(artifacts, EvaluationOptions::default())
    }

    /// Build a FutureQuoteV1 result and apply typed provider-evaluation selection.
    pub fn from_future_artifacts_with_options(
        artifacts: FutureBacktestArtifacts,
        evaluation_options: EvaluationOptions,
    ) -> Self {
        let trade_log = future_trade_log(&artifacts);
        let (provider_evaluation, provider_positions) =
            evaluate_future_positions(&artifacts, evaluation_options);
        let settled: BTreeSet<&str> = artifacts
            .completed_positions
            .iter()
            .filter(|position| !position.close_events.is_empty())
            .map(|position| position.position_id.as_str())
            .collect();
        let mut result = Self::from_trade_log(artifacts.execution.initial_balance, trade_log);
        result.apply_cost_events(&artifacts.cost_events, &settled);
        result.replace_position_statistics(&artifacts.completed_positions);
        result.future_format_version = Some(artifacts.format_version);
        result.execution_metadata = Some(artifacts.execution);
        result.recorded_fills = artifacts.fills;
        result.action_dispositions = artifacts.lifecycle.as_slice().to_vec();
        result.close_events = artifacts.close_events;
        result.completed_positions = artifacts.completed_positions;
        result.open_position_snapshots = artifacts.open_positions;
        result.pending_order_snapshots = artifacts.pending_orders;
        result.pending_order_lifecycle = artifacts.pending_order_lifecycle;
        result.mtm_equity_curve = artifacts.equity_curve;
        result.mtm_output_summary = artifacts.mtm_output_summary;
        result.mtm_max_drawdown = artifacts.max_drawdown;
        result.mtm_max_drawdown_pct = artifacts.max_drawdown_pct;
        result.provider_evaluation = Some(provider_evaluation);
        result.provider_positions = provider_positions;
        result.cost_events = artifacts.cost_events;
        result
    }

    /// Fold commission and swap totals into the realized result.
    ///
    /// Exit commission is already inside each trade's profit and loss, so only entry commission and swap still have to be applied to the run totals. The call is a no-op when nothing was charged, which keeps cost-free runs identical to runs produced before costs existed.
    fn apply_cost_events(&mut self, cost_events: &[CostEvent], settled: &BTreeSet<&str>) {
        if cost_events.is_empty() {
            return;
        }
        let mut commission = 0.0;
        let mut swap = 0.0;
        let mut outside_trade_log = 0.0;
        for event in cost_events {
            // A position that closed fully settles its entry commission and swap on its final trade row, so only a position still open when the run ended still has charges outside the trade log.
            let unsettled = !settled.contains(event.position_id.as_str());
            match event.kind {
                CostKind::EntryCommission => {
                    commission += event.amount;
                    if unsettled {
                        outside_trade_log += event.amount;
                    }
                }
                CostKind::ExitCommission => commission += event.amount,
                CostKind::Swap => {
                    swap += event.amount;
                    if unsettled {
                        outside_trade_log += event.amount;
                    }
                }
            }
        }
        self.total_commission = commission;
        self.total_swap = swap;
        self.total_pnl -= outside_trade_log;
        self.gross_pnl = Some(self.total_pnl + commission + swap);
        self.final_balance = self.initial_balance + self.total_pnl;
    }

    fn replace_position_statistics(&mut self, completed_positions: &[CompletedPosition]) {
        self.positions = completed_positions
            .iter()
            .map(position_summary_from_completed)
            .collect();
        self.positions.sort_by(|left, right| {
            left.open_ts
                .cmp(&right.open_ts)
                .then_with(|| left.final_close_ts.cmp(&right.final_close_ts))
                .then_with(|| left.position_id.cmp(&right.position_id))
        });
        self.streaks = StreakStats::from_completed_positions(completed_positions);
        self.duration_stats = DurationStats::from_completed_positions(completed_positions);
        self.monthly_returns =
            compute_monthly_returns_from_completed(completed_positions, self.initial_balance);
        self.total_positions = completed_positions.len();
        self.winning_positions = completed_positions
            .iter()
            .filter(|position| position.outcome == NetPnlOutcome::Win)
            .count();
        self.losing_positions = completed_positions
            .iter()
            .filter(|position| position.outcome == NetPnlOutcome::Loss)
            .count();
        self.position_win_rate = if self.total_positions > 0 {
            self.winning_positions as f64 / self.total_positions as f64
        } else {
            0.0
        };
    }
}

fn position_summary_from_completed(position: &CompletedPosition) -> PositionSummary {
    let closed_size = position
        .close_events
        .iter()
        .map(|event| event.size)
        .sum::<f64>();
    let avg_exit_price = if closed_size > 0.0 {
        position
            .close_events
            .iter()
            .map(|event| event.price * event.size)
            .sum::<f64>()
            / closed_size
    } else {
        0.0
    };
    let close_reasons = if position.close_events.is_empty() {
        position.close_reasons.clone()
    } else {
        position
            .close_events
            .iter()
            .map(|event| event.reason)
            .collect()
    };

    PositionSummary {
        position_id: position.position_id.clone(),
        symbol: position.symbol.clone(),
        side: position.side,
        group: position.group.clone(),
        entry_price: position.average_entry_price,
        avg_exit_price,
        original_size: position.entry_size,
        close_count: position.close_events.len(),
        net_pnl: position.net_pnl,
        close_reasons,
        open_ts: position.open_ts,
        final_close_ts: position.close_ts,
        duration_seconds: (position.close_ts - position.open_ts).num_seconds(),
    }
}

/// Entry commission and swap charged to one position, which belong to the position rather than to any single close.
#[derive(Debug, Clone, Copy, Default)]
struct PositionCostSettlement {
    commission: f64,
    swap: f64,
}

/// Costs that a completed position settles on its final close row, keyed by the identifier of that row.
///
/// Only fully closed positions appear. A position still open when the run ended has no final row to settle against, so its charges stay on the run totals alone.
fn final_row_settlements(
    artifacts: &FutureBacktestArtifacts,
) -> BTreeMap<&str, PositionCostSettlement> {
    let mut by_position: BTreeMap<&str, PositionCostSettlement> = BTreeMap::new();
    for event in &artifacts.cost_events {
        let entry = by_position.entry(event.position_id.as_str()).or_default();
        match event.kind {
            CostKind::EntryCommission => entry.commission += event.amount,
            CostKind::Swap => entry.swap += event.amount,
            // An exit commission is already inside its own close event's profit and loss.
            CostKind::ExitCommission => {}
        }
    }
    let mut by_row = BTreeMap::new();
    for position in &artifacts.completed_positions {
        let Some(settlement) = by_position.get(position.position_id.as_str()) else {
            continue;
        };
        let Some(final_close) = position.close_events.last() else {
            continue;
        };
        by_row.insert(final_close.id.as_str(), *settlement);
    }
    by_row
}

fn future_trade_log(artifacts: &FutureBacktestArtifacts) -> Vec<TradeResult> {
    let settlements = final_row_settlements(artifacts);
    let mut rows = Vec::with_capacity(artifacts.close_events.len());
    for event in &artifacts.close_events {
        let completed = artifacts
            .completed_positions
            .iter()
            .find(|position| position.position_id == event.position_id);
        let open = artifacts
            .open_positions
            .iter()
            .find(|position| position.position_id == event.position_id);
        let entry_price = event
            .entry_price
            .or_else(|| completed.map(|position| position.average_entry_price))
            .or_else(|| open.map(|position| position.average_entry_price))
            .unwrap_or(event.price);
        let open_ts = completed
            .map(|position| position.open_ts)
            .or_else(|| open.and_then(|position| position.open_ts))
            .unwrap_or(event.ts);
        let group = completed
            .and_then(|position| position.group.clone())
            .or_else(|| open.and_then(|position| position.group.clone()));
        let settled = settlements
            .get(event.id.as_str())
            .copied()
            .unwrap_or_default();
        let commission = event.commission + settled.commission;
        let swap = settled.swap;
        let pnl = event.pnl - settled.commission - settled.swap;
        rows.push(TradeResult {
            position_id: event.position_id.clone(),
            symbol: event.symbol.clone(),
            side: event.side,
            entry_price,
            exit_price: event.price,
            size: event.size,
            pnl,
            commission,
            swap,
            gross_pnl: (commission != 0.0 || swap != 0.0).then_some(pnl + commission + swap),
            open_ts,
            close_ts: event.ts,
            close_reason: event.reason,
            group,
        });
    }
    rows.sort_by(|left, right| {
        left.close_ts
            .cmp(&right.close_ts)
            .then_with(|| left.position_id.cmp(&right.position_id))
            .then_with(|| {
                left.close_reason
                    .to_string()
                    .cmp(&right.close_reason.to_string())
            })
            .then_with(|| left.size.total_cmp(&right.size))
            .then_with(|| left.pnl.total_cmp(&right.pnl))
    });
    rows
}

fn evaluate_future_positions(
    artifacts: &FutureBacktestArtifacts,
    options: EvaluationOptions,
) -> (EvaluationReport, Vec<PositionOutcome>) {
    let positions: Vec<PositionOutcome> = artifacts
        .completed_positions
        .iter()
        .map(|position| {
            let initial_risk = position.initial_risk();
            let excursions = initial_risk.and_then(|risk| {
                (risk > 0.0).then_some(ExcursionInput {
                    favorable_r: position.mfe.map(|value| value / risk),
                    adverse_r: position.mae.map(|value| value / risk),
                })
            });
            let fills: Vec<_> = artifacts
                .fills
                .iter()
                .filter(|fill| fill.position_id == position.position_id)
                .collect();
            let execution = (!fills.is_empty()).then(|| {
                let latency_ms = fills
                    .iter()
                    .map(|fill| {
                        (fill.execution_ts.unwrap_or(fill.quote_ts) - fill.effective_ts)
                            .num_milliseconds() as f64
                    })
                    .sum::<f64>()
                    / fills.len() as f64;
                let slippage_bps = fills
                    .iter()
                    .filter(|fill| fill.fill.quote_price.is_finite() && fill.fill.quote_price > 0.0)
                    .map(|fill| {
                        let raw = (fill.fill.price - fill.fill.quote_price) / fill.fill.quote_price
                            * 10_000.0;
                        let adverse_sign = match (fill.fill.purpose.is_entry(), fill.fill.side) {
                            (true, Side::Buy) | (false, Side::Sell) => 1.0,
                            (true, Side::Sell) | (false, Side::Buy) => -1.0,
                        };
                        raw * adverse_sign
                    })
                    .sum::<f64>()
                    / fills.len() as f64;
                ExecutionDiagnosticsInput {
                    slippage_bps: Some(slippage_bps),
                    latency_ms: Some(latency_ms),
                    fill_ratio: position_fill_ratio(position, artifacts),
                }
            });
            PositionOutcome {
                id: position.position_id.clone(),
                trade_id: position.trade_id.clone(),
                ordinal: position.close_ts.and_utc().timestamp_millis(),
                dimensions: PositionDimensions {
                    symbol: position.symbol.clone(),
                    side: match position.side {
                        Side::Buy => PositionSide::Long,
                        Side::Sell => PositionSide::Short,
                    },
                    group: position.group.clone(),
                    close_reasons: position
                        .close_reasons
                        .iter()
                        .map(ToString::to_string)
                        .collect(),
                    // Every position of a run shares the run's labels, which is what lets a breakdown group positions across runs by parameter or window; a label of the position itself, such as its instance, joins them.
                    tags: {
                        let mut tags = artifacts.execution.run_tags.clone();
                        if let Some(own) =
                            artifacts.execution.position_tags.get(&position.position_id)
                        {
                            tags.extend(own.clone());
                        }
                        tags
                    },
                },
                outcome: position.net_pnl,
                outcome_classification: Some(match position.outcome {
                    NetPnlOutcome::Win => OutcomeClassification::Win,
                    NetPnlOutcome::Loss => OutcomeClassification::Loss,
                    NetPnlOutcome::Breakeven => OutcomeClassification::Breakeven,
                }),
                r_multiple: position.realized_r,
                excursions,
                execution,
                costs: (position.commission_total != 0.0 || position.swap_total != 0.0).then_some(
                    PositionCostsInput {
                        commission: position.commission_total,
                        swap: position.swap_total,
                    },
                ),
            }
        })
        .collect();
    let entry_dispositions: Vec<_> = artifacts
        .lifecycle
        .iter()
        .filter(|disposition| disposition.action_kind.as_deref() == Some("entry"))
        .collect();
    let accepted = entry_dispositions
        .iter()
        .filter(|disposition| disposition.status == ActionDispositionStatus::Applied)
        .count() as u64;
    let rejected = entry_dispositions.len() as u64 - accepted;
    let lifecycle = LifecycleCounts {
        candidates: entry_dispositions.len() as u64,
        accepted,
        opened: artifacts
            .fills
            .iter()
            .filter(|fill| fill.fill.purpose.is_entry())
            .map(|fill| fill.position_id.as_str())
            .collect::<std::collections::HashSet<_>>()
            .len() as u64,
        completed: artifacts.completed_positions.len() as u64,
        rejected,
        filled: artifacts
            .pending_order_lifecycle
            .iter()
            .filter(|event| event.state == PendingOrderLifecycleState::Filled)
            .count() as u64,
        cancelled: artifacts
            .pending_order_lifecycle
            .iter()
            .filter(|event| event.state == PendingOrderLifecycleState::Cancelled)
            .count() as u64,
        unfilled_at_end: artifacts
            .pending_order_lifecycle
            .iter()
            .filter(|event| event.state == PendingOrderLifecycleState::UnfilledAtEnd)
            .count() as u64,
        open_at_end: artifacts.open_positions.len() as u64,
    };
    let report = evaluate(&EvaluationRequest {
        positions: positions.clone(),
        lifecycle: Some(lifecycle),
        options,
    });
    (report, positions)
}

fn position_fill_ratio(
    position: &CompletedPosition,
    artifacts: &FutureBacktestArtifacts,
) -> Option<f64> {
    let entry_fills: Vec<_> = artifacts
        .fills
        .iter()
        .filter(|fill| fill.position_id == position.position_id && fill.fill.purpose.is_entry())
        .collect();
    if entry_fills.is_empty() {
        return None;
    }

    let total_filled = entry_fills
        .iter()
        .map(|fill| fill.size)
        .filter(|size| size.is_finite() && *size > 0.0)
        .sum::<f64>();
    if total_filled <= 0.0 {
        return None;
    }

    let pending_fill = artifacts.pending_order_lifecycle.iter().find(|event| {
        event.position_id == position.position_id
            && event.state == PendingOrderLifecycleState::Filled
    });
    let Some(pending_fill) = pending_fill else {
        // Market entry and scale-in fills are fill-or-reject in FutureQuoteV1.
        return Some(1.0);
    };
    let pending_filled = pending_fill.filled_size.filter(|size| size.is_finite())?;
    if !pending_fill.requested_size.is_finite() || pending_fill.requested_size <= 0.0 {
        return None;
    }
    let other_filled = (total_filled - pending_filled).max(0.0);
    let requested = pending_fill.requested_size + other_filled;
    (requested > 0.0).then_some(total_filled / requested)
}

// ─── Display ────────────────────────────────────────────────────────────────

/// Format duration seconds into a human-readable string (e.g. "3d 14h 5m").
fn fmt_duration(secs: i64) -> String {
    if secs < 0 {
        return format!("-{}", fmt_duration(-secs));
    }
    let days = secs / 86400;
    let hours = (secs % 86400) / 3600;
    let minutes = (secs % 3600) / 60;
    if days > 0 {
        format!("{}d {}h {}m", days, hours, minutes)
    } else if hours > 0 {
        format!("{}h {}m", hours, minutes)
    } else {
        format!("{}m", minutes)
    }
}

/// Format a SubsetStats one-line summary for breakdown sections.
fn fmt_subset_line(label: &str, stats: &SubsetStats) -> String {
    format!(
        "{:<14}: {} trades, P&L: {:+.2}, WR: {:.1}%, PF: {:.2}",
        label,
        stats.total_trades,
        stats.total_pnl,
        stats.win_rate * 100.0,
        stats.profit_factor,
    )
}

impl std::fmt::Display for BacktestResult {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        let be = self.summary.breakeven_trades;

        writeln!(f, "═══ Backtest Result ═══")?;
        writeln!(
            f,
            "Balance      : {:.2} -> {:.2}",
            self.initial_balance, self.final_balance
        )?;
        writeln!(f, "Total P&L    : {:+.2}", self.total_pnl)?;
        writeln!(f, "Trades       : {}", self.total_trades)?;

        if be > 0 {
            writeln!(
                f,
                "Win / Lose   : {} / {} / {} (BE)",
                self.winning_trades, self.losing_trades, be
            )?;
        } else {
            writeln!(
                f,
                "Win / Lose   : {} / {}",
                self.winning_trades, self.losing_trades
            )?;
        }

        writeln!(f, "Win Rate     : {:.1}%", self.win_rate * 100.0)?;
        writeln!(f, "Profit Factor: {:.2}", self.profit_factor)?;
        writeln!(f, "Expectancy   : {:.2} per trade", self.summary.expectancy)?;

        // ── Risk Metrics ────────────────────────────────────────────
        writeln!(f)?;
        writeln!(f, "-- Risk Metrics --")?;

        match self.risk_metrics.sharpe_ratio {
            Some(v) => writeln!(f, "Sharpe Ratio       : {:.2}", v)?,
            None => writeln!(f, "Sharpe Ratio       : N/A")?,
        }
        match self.risk_metrics.sortino_ratio {
            Some(v) => writeln!(f, "Sortino Ratio      : {:.2}", v)?,
            None => writeln!(f, "Sortino Ratio      : N/A")?,
        }
        match self.risk_metrics.calmar_ratio {
            Some(v) => writeln!(f, "Calmar Ratio       : {:.2}", v)?,
            None => writeln!(f, "Calmar Ratio       : N/A")?,
        }
        writeln!(
            f,
            "Max Drawdown       : {:.2} ({:.1}%)",
            self.max_drawdown,
            self.max_drawdown_pct * 100.0
        )?;
        match self.risk_metrics.max_drawdown_duration_secs {
            Some(s) => writeln!(f, "Max DD Duration    : {}", fmt_duration(s))?,
            None => writeln!(f, "Max DD Duration    : N/A")?,
        }
        match self.risk_metrics.return_on_max_drawdown {
            Some(v) => writeln!(f, "Return / Max DD    : {:.2}", v)?,
            None => writeln!(f, "Return / Max DD    : N/A")?,
        }

        // ── Win / Loss Analysis ─────────────────────────────────────
        writeln!(f)?;
        writeln!(f, "-- Win / Loss Analysis --")?;
        writeln!(
            f,
            "Avg Win    : {:.2}    Largest Win  : {:.2}",
            self.summary.avg_win, self.summary.largest_win
        )?;
        writeln!(
            f,
            "Avg Loss   : {:.2}    Largest Loss : {:.2}",
            self.summary.avg_loss, self.summary.largest_loss
        )?;
        writeln!(
            f,
            "Win/Loss   : {:.2}     Expectancy   : {:.2}",
            self.summary.win_loss_ratio, self.summary.expectancy
        )?;
        writeln!(
            f,
            "Max Consec Wins  : {}",
            self.streaks.max_consecutive_wins
        )?;
        writeln!(
            f,
            "Max Consec Losses: {}",
            self.streaks.max_consecutive_losses
        )?;

        // ── Side Breakdown ──────────────────────────────────────────
        writeln!(f)?;
        writeln!(f, "-- Side Breakdown --")?;
        writeln!(f, "{}", fmt_subset_line("Long", &self.long_stats))?;
        writeln!(f, "{}", fmt_subset_line("Short", &self.short_stats))?;

        // ── Symbol Breakdown ────────────────────────────────────────
        if !self.per_symbol.is_empty() {
            writeln!(f)?;
            writeln!(f, "-- Symbol Breakdown --")?;
            let mut symbols: Vec<_> = self.per_symbol.iter().collect();
            symbols.sort_by(|a, b| {
                b.1.total_trades
                    .cmp(&a.1.total_trades)
                    .then_with(|| a.0.cmp(b.0))
            });
            for (sym, stats) in &symbols {
                writeln!(f, "{}", fmt_subset_line(sym, stats))?;
            }
        }

        // ── Group Breakdown ─────────────────────────────────────────
        if !self.per_group.is_empty() {
            writeln!(f)?;
            writeln!(f, "-- Group Breakdown --")?;
            let mut groups: Vec<_> = self.per_group.iter().collect();
            groups.sort_by(|a, b| {
                b.1.total_trades
                    .cmp(&a.1.total_trades)
                    .then_with(|| a.0.cmp(b.0))
            });
            for (grp, stats) in &groups {
                writeln!(f, "{}", fmt_subset_line(grp, stats))?;
            }
        }

        // ── Close Reasons ───────────────────────────────────────────
        if !self.per_close_reason.is_empty() {
            writeln!(f)?;
            writeln!(f, "-- Close Reasons --")?;
            for cr in &self.per_close_reason {
                writeln!(
                    f,
                    "{:<14}: {:>3} ({:>4.1}%), P&L: {:+.2}",
                    cr.reason.to_string(),
                    cr.count,
                    cr.percentage * 100.0,
                    cr.total_pnl,
                )?;
            }
        }

        // ── Duration ────────────────────────────────────────────────
        if let Some(ref ds) = self.duration_stats {
            writeln!(f)?;
            writeln!(f, "-- Duration --")?;
            writeln!(
                f,
                "Avg Duration     : {}",
                fmt_duration(ds.avg_duration_secs)
            )?;
            writeln!(
                f,
                "Avg Winner Dur   : {}",
                fmt_duration(ds.avg_winner_duration_secs)
            )?;
            writeln!(
                f,
                "Avg Loser Dur    : {}",
                fmt_duration(ds.avg_loser_duration_secs)
            )?;
            writeln!(
                f,
                "Shortest         : {}",
                fmt_duration(ds.min_duration_secs)
            )?;
            writeln!(
                f,
                "Longest          : {}",
                fmt_duration(ds.max_duration_secs)
            )?;
        }

        // ── Monthly Returns ─────────────────────────────────────────
        if !self.monthly_returns.is_empty() {
            writeln!(f)?;
            writeln!(f, "-- Monthly Returns --")?;
            for mr in &self.monthly_returns {
                writeln!(
                    f,
                    "{:04}-{:02} : {:+.2} ({} trades)",
                    mr.year, mr.month, mr.pnl, mr.trade_count,
                )?;
            }
        }

        // ── Position Summary ────────────────────────────────────────
        if self.total_positions > 0 {
            writeln!(f)?;
            writeln!(f, "-- Position Summary --")?;
            writeln!(f, "Total Positions  : {}", self.total_positions)?;
            writeln!(
                f,
                "Win / Lose       : {} / {}",
                self.winning_positions, self.losing_positions
            )?;
            writeln!(
                f,
                "Position WR      : {:.1}%",
                self.position_win_rate * 100.0
            )?;
        }

        Ok(())
    }
}

// ─── Tests ──────────────────────────────────────────────────────────────────

#[cfg(test)]
mod tests {
    use super::*;
    use crate::evaluation::{EvaluationSection, GroupFilter, PositionFilter};
    use chrono::NaiveDate;
    use std::collections::BTreeSet;

    fn ts(year: i32, month: u32, day: u32, h: u32, m: u32, s: u32) -> NaiveDateTime {
        NaiveDate::from_ymd_opt(year, month, day)
            .unwrap()
            .and_hms_opt(h, m, s)
            .unwrap()
    }

    fn ts_hms(h: u32, m: u32, s: u32) -> NaiveDateTime {
        ts(2026, 1, 1, h, m, s)
    }

    fn make_trade(pnl: f64, close_h: u32) -> TradeResult {
        TradeResult {
            position_id: "p1".into(),
            symbol: "EURUSD".into(),
            side: Side::Buy,
            entry_price: 1.0850,
            exit_price: 1.0850 + pnl,
            size: 1.0,
            pnl,
            open_ts: ts_hms(10, 0, 0),
            close_ts: ts_hms(close_h, 0, 0),
            close_reason: if pnl > 0.0 {
                CloseReason::Target
            } else if pnl < 0.0 {
                CloseReason::Stoploss
            } else {
                CloseReason::Manual
            },
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        }
    }

    #[allow(
        clippy::too_many_arguments,
        reason = "keeping fixture fields explicit is clearer than rewriting the many stable call sites"
    )]
    fn make_trade_full(
        pos_id: &str,
        symbol: &str,
        side: Side,
        pnl: f64,
        open_ts: NaiveDateTime,
        close_ts: NaiveDateTime,
        reason: CloseReason,
        group: Option<GroupId>,
    ) -> TradeResult {
        TradeResult {
            position_id: pos_id.into(),
            symbol: symbol.into(),
            side,
            entry_price: 1.0850,
            exit_price: 1.0850 + pnl,
            size: 1.0,
            pnl,
            open_ts,
            close_ts,
            close_reason: reason,
            group,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        }
    }

    // ── Backward compatibility tests (existing, preserved) ──────────

    #[test]
    fn empty_trade_log() {
        let result = BacktestResult::from_trade_log(10_000.0, vec![]);
        assert_eq!(result.total_trades, 0);
        assert!((result.final_balance - 10_000.0).abs() < f64::EPSILON);
        assert!((result.win_rate - 0.0).abs() < f64::EPSILON);
        assert!((result.max_drawdown - 0.0).abs() < f64::EPSILON);
        // Enhanced fields on empty.
        assert_eq!(result.summary.total_trades, 0);
        assert!(result.duration_stats.is_none());
        assert!(result.monthly_returns.is_empty());
        assert_eq!(result.total_positions, 0);
        assert!((result.position_win_rate - 0.0).abs() < f64::EPSILON);
        assert_eq!(result.streaks.max_consecutive_wins, 0);
        assert_eq!(result.streaks.max_consecutive_losses, 0);
        assert!(result.risk_metrics.sharpe_ratio.is_none());
    }

    #[test]
    fn basic_stats() {
        let trades = vec![
            make_trade(100.0, 11),
            make_trade(-50.0, 12),
            make_trade(200.0, 13),
            make_trade(-30.0, 14),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert_eq!(result.total_trades, 4);
        assert_eq!(result.winning_trades, 2);
        assert_eq!(result.losing_trades, 2);
        assert!((result.total_pnl - 220.0).abs() < f64::EPSILON);
        assert!((result.final_balance - 10_220.0).abs() < f64::EPSILON);
        assert!((result.win_rate - 0.5).abs() < f64::EPSILON);
        assert!((result.profit_factor - 3.75).abs() < f64::EPSILON);
    }

    #[test]
    fn drawdown_calculation() {
        let trades = vec![
            make_trade(100.0, 11),
            make_trade(-200.0, 12),
            make_trade(50.0, 13),
            make_trade(-100.0, 14),
            make_trade(500.0, 15),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!((result.max_drawdown - 250.0).abs() < f64::EPSILON);
        assert_eq!(result.equity_curve.len(), 5);
    }

    #[test]
    fn all_winners() {
        let trades = vec![make_trade(100.0, 11), make_trade(200.0, 12)];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!((result.win_rate - 1.0).abs() < f64::EPSILON);
        assert!(result.profit_factor.is_infinite());
        assert!((result.max_drawdown - 0.0).abs() < f64::EPSILON);
    }

    // ── SubsetStats tests ───────────────────────────────────────────

    #[test]
    fn subset_stats_basic() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(-50.0, 12);
        let t3 = make_trade(200.0, 13);
        let t4 = make_trade(-30.0, 14);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4];

        let s = SubsetStats::from_trades(&refs);
        assert_eq!(s.total_trades, 4);
        assert_eq!(s.winning_trades, 2);
        assert_eq!(s.losing_trades, 2);
        assert_eq!(s.breakeven_trades, 0);
        assert!((s.total_pnl - 220.0).abs() < f64::EPSILON);
        assert!((s.gross_profit - 300.0).abs() < f64::EPSILON);
        assert!((s.gross_loss - 80.0).abs() < f64::EPSILON);
        assert!((s.win_rate - 0.5).abs() < f64::EPSILON);
        assert!((s.profit_factor - 3.75).abs() < f64::EPSILON);
        assert!((s.avg_win - 150.0).abs() < f64::EPSILON);
        assert!((s.avg_loss - 40.0).abs() < f64::EPSILON);
        assert!((s.win_loss_ratio - 3.75).abs() < f64::EPSILON);
        // expectancy = 0.5*150 - 0.5*40 = 55
        assert!((s.expectancy - 55.0).abs() < f64::EPSILON);
    }

    #[test]
    fn subset_stats_all_winners() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(200.0, 12);
        let refs: Vec<&TradeResult> = vec![&t1, &t2];

        let s = SubsetStats::from_trades(&refs);
        assert_eq!(s.losing_trades, 0);
        assert!((s.avg_loss - 0.0).abs() < f64::EPSILON);
        assert!(s.win_loss_ratio.is_infinite());
        assert!(s.profit_factor.is_infinite());
    }

    #[test]
    fn subset_stats_all_losers() {
        let t1 = make_trade(-100.0, 11);
        let t2 = make_trade(-200.0, 12);
        let refs: Vec<&TradeResult> = vec![&t1, &t2];

        let s = SubsetStats::from_trades(&refs);
        assert_eq!(s.winning_trades, 0);
        assert!((s.avg_win - 0.0).abs() < f64::EPSILON);
        assert!((s.win_loss_ratio - 0.0).abs() < f64::EPSILON);
        assert!((s.profit_factor - 0.0).abs() < f64::EPSILON);
    }

    #[test]
    fn subset_stats_empty() {
        let s = SubsetStats::from_trades(&[]);
        assert_eq!(s.total_trades, 0);
        assert!((s.total_pnl - 0.0).abs() < f64::EPSILON);
        assert!((s.win_rate - 0.0).abs() < f64::EPSILON);
        assert!((s.expectancy - 0.0).abs() < f64::EPSILON);
    }

    #[test]
    fn subset_stats_largest_win_loss() {
        let t1 = make_trade(50.0, 11);
        let t2 = make_trade(200.0, 12);
        let t3 = make_trade(-30.0, 13);
        let t4 = make_trade(-100.0, 14);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4];

        let s = SubsetStats::from_trades(&refs);
        assert!((s.largest_win - 200.0).abs() < f64::EPSILON);
        assert!((s.largest_loss - 100.0).abs() < f64::EPSILON);
    }

    #[test]
    fn subset_stats_breakeven_trades() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(0.0, 12);
        let t3 = make_trade(-50.0, 13);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3];

        let s = SubsetStats::from_trades(&refs);
        assert_eq!(s.breakeven_trades, 1);
        assert_eq!(s.winning_trades, 1);
        assert_eq!(s.losing_trades, 1);
    }

    // ── StreakStats tests ───────────────────────────────────────────

    #[test]
    fn streaks_alternating() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(-50.0, 12);
        let t3 = make_trade(100.0, 13);
        let t4 = make_trade(-50.0, 14);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4];

        let s = StreakStats::from_trades(&refs);
        assert_eq!(s.max_consecutive_wins, 1);
        assert_eq!(s.max_consecutive_losses, 1);
    }

    #[test]
    fn streaks_consecutive_wins() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(50.0, 12);
        let t3 = make_trade(80.0, 13);
        let t4 = make_trade(-50.0, 14);
        let t5 = make_trade(100.0, 15);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4, &t5];

        let s = StreakStats::from_trades(&refs);
        assert_eq!(s.max_consecutive_wins, 3);
        assert_eq!(s.max_consecutive_losses, 1);
    }

    #[test]
    fn streaks_consecutive_losses() {
        let t1 = make_trade(-10.0, 11);
        let t2 = make_trade(-20.0, 12);
        let t3 = make_trade(-30.0, 13);
        let t4 = make_trade(-40.0, 14);
        let t5 = make_trade(100.0, 15);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4, &t5];

        let s = StreakStats::from_trades(&refs);
        assert_eq!(s.max_consecutive_wins, 1);
        assert_eq!(s.max_consecutive_losses, 4);
    }

    #[test]
    fn streaks_all_winners() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(200.0, 12);
        let t3 = make_trade(300.0, 13);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3];

        let s = StreakStats::from_trades(&refs);
        assert_eq!(s.max_consecutive_wins, 3);
        assert_eq!(s.max_consecutive_losses, 0);
        assert_eq!(s.current_streak, 3);
    }

    #[test]
    fn streaks_empty() {
        let s = StreakStats::from_trades(&[]);
        assert_eq!(s.max_consecutive_wins, 0);
        assert_eq!(s.max_consecutive_losses, 0);
        assert_eq!(s.current_streak, 0);
    }

    #[test]
    fn streaks_breakeven_resets() {
        let t1 = make_trade(100.0, 11);
        let t2 = make_trade(200.0, 12);
        let t3 = make_trade(0.0, 13); // breakeven resets
        let t4 = make_trade(100.0, 14);
        let refs: Vec<&TradeResult> = vec![&t1, &t2, &t3, &t4];

        let s = StreakStats::from_trades(&refs);
        assert_eq!(s.max_consecutive_wins, 2); // not 3
        assert_eq!(s.current_streak, 1);
    }

    // ── RiskMetrics tests ───────────────────────────────────────────

    #[test]
    fn sharpe_ratio_positive() {
        // Consistent small wins should produce positive Sharpe.
        let trades: Vec<TradeResult> = (0..20)
            .map(|i| {
                make_trade_full(
                    &format!("p{}", i),
                    "EURUSD",
                    Side::Buy,
                    10.0 + (i as f64),
                    ts(2026, 1, 1, 10, 0, 0),
                    ts(2026, 1, 1 + (i as u32 / 5), 11 + (i as u32 % 12), 0, 0),
                    CloseReason::Target,
                    None,
                )
            })
            .collect();
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!(result.risk_metrics.sharpe_ratio.is_some());
        assert!(result.risk_metrics.sharpe_ratio.unwrap() > 0.0);
    }

    #[test]
    fn sharpe_ratio_insufficient_data() {
        let trades = vec![make_trade(100.0, 11)];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!(result.risk_metrics.sharpe_ratio.is_none());
    }

    #[test]
    fn sortino_ratio_no_downside() {
        let trades = vec![make_trade(100.0, 11), make_trade(200.0, 12)];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        // No negative returns → Sortino undefined.
        assert!(result.risk_metrics.sortino_ratio.is_none());
    }

    #[test]
    fn calmar_ratio_zero_drawdown() {
        let trades = vec![make_trade(100.0, 11), make_trade(200.0, 12)];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        // No drawdown → Calmar undefined.
        assert!(result.risk_metrics.calmar_ratio.is_none());
    }

    #[test]
    fn max_drawdown_duration_recovered() {
        // Win, then lose (creates drawdown), then win big (recovers).
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                -200.0,
                ts(2026, 1, 1, 11, 0, 0),
                ts(2026, 1, 2, 11, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                300.0,
                ts(2026, 1, 2, 11, 0, 0),
                ts(2026, 1, 5, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        // Drawdown from day 1 11:00 (peak after first trade) to day 5 11:00 (recovery).
        assert!(result.risk_metrics.max_drawdown_duration_secs.is_some());
        let dur = result.risk_metrics.max_drawdown_duration_secs.unwrap();
        assert!(dur > 0);
    }

    #[test]
    fn max_drawdown_duration_unrecovered() {
        // Win then lose — never recovers.
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                -200.0,
                ts(2026, 1, 1, 11, 0, 0),
                ts(2026, 1, 5, 11, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!(result.risk_metrics.max_drawdown_duration_secs.is_some());
        // Duration should span from peak_ts (day 1 11:00) to last trade (day 5 11:00) = 4 days.
        let dur = result.risk_metrics.max_drawdown_duration_secs.unwrap();
        assert_eq!(dur, 4 * 86400);
    }

    // ── DurationStats tests ─────────────────────────────────────────

    #[test]
    fn duration_stats_basic() {
        let t1 = make_trade_full(
            "p1",
            "EURUSD",
            Side::Buy,
            100.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 12, 0, 0),
            CloseReason::Target,
            None,
        );
        let t2 = make_trade_full(
            "p2",
            "EURUSD",
            Side::Buy,
            -50.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 14, 0, 0),
            CloseReason::Stoploss,
            None,
        );
        let refs: Vec<&TradeResult> = vec![&t1, &t2];

        let ds = DurationStats::from_trades(&refs).unwrap();
        assert_eq!(ds.min_duration_secs, 7200); // 2h
        assert_eq!(ds.max_duration_secs, 14400); // 4h
        assert_eq!(ds.avg_duration_secs, 10800); // 3h
        assert_eq!(ds.avg_winner_duration_secs, 7200);
        assert_eq!(ds.avg_loser_duration_secs, 14400);
    }

    #[test]
    fn duration_stats_single_trade() {
        let t1 = make_trade_full(
            "p1",
            "EURUSD",
            Side::Buy,
            100.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 11, 0, 0),
            CloseReason::Target,
            None,
        );
        let refs: Vec<&TradeResult> = vec![&t1];

        let ds = DurationStats::from_trades(&refs).unwrap();
        assert_eq!(ds.avg_duration_secs, 3600);
        assert_eq!(ds.min_duration_secs, 3600);
        assert_eq!(ds.max_duration_secs, 3600);
    }

    #[test]
    fn duration_stats_empty() {
        assert!(DurationStats::from_trades(&[]).is_none());
    }

    #[test]
    fn duration_stats_winner_vs_loser() {
        // Winners held shorter, losers longer.
        let t1 = make_trade_full(
            "p1",
            "EURUSD",
            Side::Buy,
            100.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 10, 30, 0),
            CloseReason::Target,
            None,
        );
        let t2 = make_trade_full(
            "p2",
            "EURUSD",
            Side::Buy,
            -50.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 16, 0, 0),
            CloseReason::Stoploss,
            None,
        );
        let refs: Vec<&TradeResult> = vec![&t1, &t2];

        let ds = DurationStats::from_trades(&refs).unwrap();
        assert!(ds.avg_winner_duration_secs < ds.avg_loser_duration_secs);
    }

    // ── MonthlyReturn tests ─────────────────────────────────────────

    #[test]
    fn monthly_returns_single_month() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 5, 10, 0, 0),
                ts(2026, 1, 10, 10, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                -30.0,
                ts(2026, 1, 12, 10, 0, 0),
                ts(2026, 1, 15, 10, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let monthly = compute_monthly_returns(&trades, 10_000.0);
        assert_eq!(monthly.len(), 1);
        assert_eq!(monthly[0].year, 2026);
        assert_eq!(monthly[0].month, 1);
        assert!((monthly[0].pnl - 70.0).abs() < f64::EPSILON);
        assert_eq!(monthly[0].trade_count, 2);
    }

    #[test]
    fn monthly_returns_multi_month() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 5, 10, 0, 0),
                ts(2026, 1, 10, 10, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                200.0,
                ts(2026, 2, 5, 10, 0, 0),
                ts(2026, 2, 10, 10, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                -50.0,
                ts(2026, 3, 5, 10, 0, 0),
                ts(2026, 3, 10, 10, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let monthly = compute_monthly_returns(&trades, 10_000.0);
        assert_eq!(monthly.len(), 3);
        assert_eq!(monthly[0].month, 1);
        assert_eq!(monthly[1].month, 2);
        assert_eq!(monthly[2].month, 3);
    }

    #[test]
    fn monthly_returns_ending_balance() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 5, 10, 0, 0),
                ts(2026, 1, 10, 10, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                200.0,
                ts(2026, 2, 5, 10, 0, 0),
                ts(2026, 2, 10, 10, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let monthly = compute_monthly_returns(&trades, 10_000.0);
        assert!((monthly[0].ending_balance - 10_100.0).abs() < f64::EPSILON);
        assert!((monthly[1].ending_balance - 10_300.0).abs() < f64::EPSILON);
    }

    // ── Per-breakdown tests ─────────────────────────────────────────

    #[test]
    fn per_symbol_breakdown() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "XAUUSD",
                Side::Buy,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                200.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 13, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert_eq!(result.per_symbol.len(), 2);

        let eu = result.per_symbol.get("EURUSD").unwrap();
        assert_eq!(eu.total_trades, 2);
        assert!((eu.total_pnl - 300.0).abs() < f64::EPSILON);

        let xau = result.per_symbol.get("XAUUSD").unwrap();
        assert_eq!(xau.total_trades, 1);
        assert!((xau.total_pnl - -50.0).abs() < f64::EPSILON);
    }

    #[test]
    fn per_side_breakdown() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Sell,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                200.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 13, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert_eq!(result.long_stats.total_trades, 2);
        assert_eq!(result.short_stats.total_trades, 1);
        assert!((result.long_stats.total_pnl - 300.0).abs() < f64::EPSILON);
        assert!((result.short_stats.total_pnl - -50.0).abs() < f64::EPSILON);
    }

    #[test]
    fn per_close_reason_breakdown() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                80.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 13, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p4",
                "EURUSD",
                Side::Buy,
                30.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 14, 0, 0),
                CloseReason::TrailingStop,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        assert_eq!(result.per_close_reason.len(), 3);
        // Sorted by count descending — Target (2), then SL (1) and Trailing (1).
        assert_eq!(result.per_close_reason[0].reason, CloseReason::Target);
        assert_eq!(result.per_close_reason[0].count, 2);
        assert_eq!(result.per_close_reason[1].reason, CloseReason::Stoploss);
        assert_eq!(result.per_close_reason[2].reason, CloseReason::TrailingStop);
        assert!((result.per_close_reason[0].percentage - 0.5).abs() < f64::EPSILON);
    }

    #[test]
    fn per_group_breakdown() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                Some("momentum".into()),
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Buy,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                Some("reversion".into()),
            ),
            make_trade_full(
                "p3",
                "EURUSD",
                Side::Buy,
                200.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 13, 0, 0),
                CloseReason::Target,
                Some("momentum".into()),
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        assert_eq!(result.per_group.len(), 2);
        let mom = result.per_group.get("momentum").unwrap();
        assert_eq!(mom.total_trades, 2);
        assert!((mom.total_pnl - 300.0).abs() < f64::EPSILON);
        let rev = result.per_group.get("reversion").unwrap();
        assert_eq!(rev.total_trades, 1);
    }

    #[test]
    fn per_group_empty_when_no_groups() {
        let trades = vec![make_trade(100.0, 11), make_trade(-50.0, 12)];
        let result = BacktestResult::from_trade_log(10_000.0, trades);
        assert!(result.per_group.is_empty());
    }

    // ── PositionSummary tests ───────────────────────────────────────

    #[test]
    fn position_summary_single_close() {
        let t1 = make_trade_full(
            "p1",
            "EURUSD",
            Side::Buy,
            100.0,
            ts(2026, 1, 1, 10, 0, 0),
            ts(2026, 1, 1, 12, 0, 0),
            CloseReason::Target,
            None,
        );
        let refs: Vec<&TradeResult> = vec![&t1];

        let ps = PositionSummary::from_trades(&refs);
        assert_eq!(ps.position_id, "p1");
        assert_eq!(ps.close_count, 1);
        assert!((ps.net_pnl - 100.0).abs() < f64::EPSILON);
        assert!(ps.is_winner());
        assert!(!ps.is_loser());
    }

    #[test]
    fn position_summary_multiple_closes() {
        // TP1 wins, then SL loses — net positive.
        let t1 = TradeResult {
            position_id: "p1".into(),
            symbol: "EURUSD".into(),
            side: Side::Buy,
            entry_price: 1.0850,
            exit_price: 1.0900,
            size: 0.5,
            pnl: 25.0,
            open_ts: ts(2026, 1, 1, 10, 0, 0),
            close_ts: ts(2026, 1, 1, 11, 0, 0),
            close_reason: CloseReason::Target,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let t2 = TradeResult {
            position_id: "p1".into(),
            symbol: "EURUSD".into(),
            side: Side::Buy,
            entry_price: 1.0850,
            exit_price: 1.0830,
            size: 0.5,
            pnl: -10.0,
            open_ts: ts(2026, 1, 1, 10, 0, 0),
            close_ts: ts(2026, 1, 1, 14, 0, 0),
            close_reason: CloseReason::Stoploss,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let refs: Vec<&TradeResult> = vec![&t1, &t2];
        let ps = PositionSummary::from_trades(&refs);

        assert_eq!(ps.close_count, 2);
        assert!((ps.entry_price - 1.0850).abs() < f64::EPSILON);
        assert!((ps.net_pnl - 15.0).abs() < f64::EPSILON);
        assert!((ps.original_size - 1.0).abs() < f64::EPSILON);
        assert!(ps.is_winner());
        assert_eq!(
            ps.close_reasons,
            vec![CloseReason::Target, CloseReason::Stoploss]
        );
        assert_eq!(ps.duration_seconds, 4 * 3600); // 10:00 to 14:00
    }

    #[test]
    fn position_summary_weights_changing_close_time_entry_basis() {
        let first_partial_close = TradeResult {
            position_id: "scaled".into(),
            symbol: "TEST".into(),
            side: Side::Buy,
            entry_price: 100.0,
            exit_price: 110.0,
            size: 1.0,
            pnl: 10.0,
            open_ts: ts_hms(10, 0, 0),
            close_ts: ts_hms(11, 0, 0),
            close_reason: CloseReason::Target,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let close_after_scale_in = TradeResult {
            position_id: "scaled".into(),
            symbol: "TEST".into(),
            side: Side::Buy,
            entry_price: 120.0,
            exit_price: 125.0,
            size: 3.0,
            pnl: 15.0,
            open_ts: ts_hms(10, 0, 0),
            close_ts: ts_hms(12, 0, 0),
            close_reason: CloseReason::Manual,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let trades = [&first_partial_close, &close_after_scale_in];

        let summary = PositionSummary::from_trades(&trades);

        assert!((summary.entry_price - 115.0).abs() < f64::EPSILON);
        assert!((summary.avg_exit_price - 121.25).abs() < f64::EPSILON);
    }

    #[test]
    fn position_summary_weighted_prices_conserve_pnl() {
        let first_partial_close = TradeResult {
            position_id: "scaled".into(),
            symbol: "TEST".into(),
            side: Side::Buy,
            entry_price: 100.0,
            exit_price: 110.0,
            size: 1.0,
            pnl: 10.0,
            open_ts: ts_hms(10, 0, 0),
            close_ts: ts_hms(11, 0, 0),
            close_reason: CloseReason::Target,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let close_after_scale_in = TradeResult {
            position_id: "scaled".into(),
            symbol: "TEST".into(),
            side: Side::Buy,
            entry_price: 120.0,
            exit_price: 125.0,
            size: 3.0,
            pnl: 15.0,
            open_ts: ts_hms(10, 0, 0),
            close_ts: ts_hms(12, 0, 0),
            close_reason: CloseReason::Manual,
            group: None,
            commission: 0.0,
            swap: 0.0,
            gross_pnl: None,
        };
        let trades = [&first_partial_close, &close_after_scale_in];

        let summary = PositionSummary::from_trades(&trades);
        let pnl_from_close_rows = trades
            .iter()
            .map(|trade| (trade.exit_price - trade.entry_price) * trade.size)
            .sum::<f64>();
        let pnl_from_summary =
            (summary.avg_exit_price - summary.entry_price) * summary.original_size;

        assert!((summary.net_pnl - pnl_from_close_rows).abs() < f64::EPSILON);
        assert!((pnl_from_summary - pnl_from_close_rows).abs() < f64::EPSILON);
    }

    #[test]
    fn position_win_rate_differs_from_trade_win_rate() {
        // Position p1: TP1 +$25, SL -$10 → net +$15 (position is a winner)
        // Trade-level: 1 win, 1 loss → 50% win rate
        // Position-level: 1 winner / 1 total → 100% win rate
        let trades = vec![
            TradeResult {
                position_id: "p1".into(),
                symbol: "EURUSD".into(),
                side: Side::Buy,
                entry_price: 1.085,
                exit_price: 1.090,
                size: 0.5,
                pnl: 25.0,
                open_ts: ts(2026, 1, 1, 10, 0, 0),
                close_ts: ts(2026, 1, 1, 11, 0, 0),
                close_reason: CloseReason::Target,
                group: None,
                commission: 0.0,
                swap: 0.0,
                gross_pnl: None,
            },
            TradeResult {
                position_id: "p1".into(),
                symbol: "EURUSD".into(),
                side: Side::Buy,
                entry_price: 1.085,
                exit_price: 1.083,
                size: 0.5,
                pnl: -10.0,
                open_ts: ts(2026, 1, 1, 10, 0, 0),
                close_ts: ts(2026, 1, 1, 14, 0, 0),
                close_reason: CloseReason::Stoploss,
                group: None,
                commission: 0.0,
                swap: 0.0,
                gross_pnl: None,
            },
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        // Trade-level.
        assert_eq!(result.total_trades, 2);
        assert!((result.win_rate - 0.5).abs() < f64::EPSILON);

        // Position-level.
        assert_eq!(result.total_positions, 1);
        assert_eq!(result.winning_positions, 1);
        assert!((result.position_win_rate - 1.0).abs() < f64::EPSILON);
    }

    fn completed_position(
        position_id: &str,
        pnl: f64,
        epsilon: f64,
        open_ts: NaiveDateTime,
        close_ts: NaiveDateTime,
    ) -> CompletedPosition {
        let reason = if pnl > 0.0 {
            CloseReason::Target
        } else {
            CloseReason::Stoploss
        };
        let close = CloseEvent::new(
            position_id,
            0,
            "ES",
            Side::Buy,
            close_ts,
            1.0,
            100.0 + pnl,
            pnl,
            reason,
        );
        CompletedPosition::from_close_events(
            position_id,
            "ES",
            Side::Buy,
            open_ts,
            close_ts,
            1.0,
            100.0,
            None,
            None,
            Vec::new(),
            vec![close],
            None,
            None,
            epsilon,
        )
    }

    #[test]
    fn automatic_provider_report_applies_or_within_and_and_between_filters() {
        let mut matching_es =
            completed_position("es-long", 1.0, 0.001, ts_hms(9, 0, 0), ts_hms(10, 0, 0));
        matching_es.group = Some("trend".into());

        let mut matching_nq =
            completed_position("nq-long", 2.0, 0.001, ts_hms(10, 0, 0), ts_hms(11, 0, 0));
        matching_nq.symbol = "NQ".into();
        matching_nq.group = Some("trend".into());

        let mut wrong_side =
            completed_position("es-short", 3.0, 0.001, ts_hms(11, 0, 0), ts_hms(12, 0, 0));
        wrong_side.side = Side::Sell;
        wrong_side.group = Some("trend".into());

        let mut wrong_group =
            completed_position("es-other", 4.0, 0.001, ts_hms(12, 0, 0), ts_hms(13, 0, 0));
        wrong_group.group = Some("countertrend".into());

        let artifacts = FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                ..ExecutionMetadata::default()
            },
            completed_positions: vec![matching_es, matching_nq, wrong_side, wrong_group],
            ..FutureBacktestArtifacts::default()
        };
        let result = BacktestResult::from_future_artifacts_with_options(
            artifacts,
            EvaluationOptions {
                sections: BTreeSet::from([
                    EvaluationSection::Coverage,
                    EvaluationSection::PositionPerformance,
                ]),
                filter: PositionFilter {
                    symbols: vec!["ES".into(), "NQ".into()],
                    sides: vec![PositionSide::Long],
                    groups: vec![GroupFilter::Named("trend".into())],
                    close_reasons: vec!["Target".into(), "Manual".into()],
                    ..PositionFilter::default()
                },
                ..EvaluationOptions::default()
            },
        );
        let evaluation = result
            .provider_evaluation
            .expect("FutureQuote result includes provider evaluation");
        let coverage = evaluation.coverage.expect("coverage requested");
        let performance = evaluation
            .position_performance
            .expect("position performance requested");

        assert_eq!(coverage.provided_positions, 4);
        assert_eq!(coverage.selected_positions, 2);
        assert_eq!(coverage.filtered_out_positions, 2);
        assert_eq!(performance.position_count, 2);
        assert_eq!(performance.total_outcome.value, Some(3.0));
        assert!(evaluation.r_metrics.is_none());
    }

    #[test]
    fn future_position_statistics_exclude_partially_closed_open_campaigns() {
        let completed =
            completed_position("completed", -10.0, 0.001, ts_hms(9, 0, 0), ts_hms(11, 0, 0));
        let mut partial = CloseEvent::new(
            "still-open",
            0,
            "ES",
            Side::Buy,
            ts_hms(12, 0, 0),
            0.5,
            110.0,
            100.0,
            CloseReason::Target,
        );
        partial.remaining_size = Some(0.5);
        let open = OpenPositionSnapshot {
            position_id: "still-open".into(),
            symbol: "ES".into(),
            side: Side::Buy,
            open_ts: Some(ts_hms(10, 0, 0)),
            average_entry_price: 100.0,
            remaining_size: 0.5,
            realized_pnl: 100.0,
            ..OpenPositionSnapshot::default()
        };
        let artifacts = FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                pnl_epsilon: 0.001,
                ..ExecutionMetadata::default()
            },
            close_events: vec![completed.close_events[0].clone(), partial],
            completed_positions: vec![completed],
            open_positions: vec![open],
            ..FutureBacktestArtifacts::default()
        };

        let result = BacktestResult::from_future_artifacts(artifacts);

        assert_eq!(result.total_trades, 2);
        assert_eq!(result.trade_log.len(), 2);
        assert_eq!(result.close_events.len(), 2);
        assert!(
            result
                .trade_log
                .iter()
                .any(|row| row.position_id == "still-open")
        );
        assert_eq!(result.total_positions, 1);
        assert_eq!(result.winning_positions, 0);
        assert_eq!(result.losing_positions, 1);
        assert_eq!(result.position_win_rate, 0.0);
        assert_eq!(result.positions.len(), 1);
        assert_eq!(result.positions[0].position_id, "completed");
        assert_eq!(result.positions[0].net_pnl, -10.0);
        assert_eq!(result.streaks.max_consecutive_wins, 0);
        assert_eq!(result.streaks.max_consecutive_losses, 1);
        assert_eq!(result.streaks.current_streak, -1);
        let duration = result
            .duration_stats
            .expect("one completed campaign has duration stats");
        assert_eq!(duration.avg_duration_secs, 2 * 3600);
        assert_eq!(result.monthly_returns.len(), 1);
        assert_eq!(result.monthly_returns[0].trade_count, 1);
        assert_eq!(result.monthly_returns[0].pnl, -10.0);
    }

    #[test]
    fn future_trade_reconstruction_uses_each_close_inventory_basis() {
        let mut first = CloseEvent::new(
            "campaign",
            0,
            "ES",
            Side::Buy,
            ts_hms(11, 0, 0),
            1.0,
            110.0,
            10.0,
            CloseReason::Manual,
        );
        first.entry_price = Some(100.0);
        let mut final_close = CloseEvent::new(
            "campaign",
            1,
            "ES",
            Side::Buy,
            ts_hms(12, 0, 0),
            2.0,
            130.0,
            40.0,
            CloseReason::Manual,
        );
        final_close.entry_price = Some(110.0);
        let completed = CompletedPosition::from_close_events(
            "campaign",
            "ES",
            Side::Buy,
            ts_hms(10, 0, 0),
            ts_hms(12, 0, 0),
            3.0,
            320.0 / 3.0,
            None,
            None,
            vec![],
            vec![first.clone(), final_close.clone()],
            None,
            None,
            crate::artifacts::DEFAULT_PNL_EPSILON,
        );
        let result = BacktestResult::from_future_artifacts(FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                ..ExecutionMetadata::default()
            },
            close_events: vec![first, final_close],
            completed_positions: vec![completed],
            ..FutureBacktestArtifacts::default()
        });

        assert_eq!(result.trade_log[0].entry_price, 100.0);
        assert_eq!(result.trade_log[1].entry_price, 110.0);
        assert_eq!(result.total_pnl, 50.0);
    }

    #[test]
    fn future_partial_tp_then_sl_is_one_breakeven_for_campaign_analytics() {
        let open_ts = ts(2026, 1, 31, 22, 0, 0);
        let partial_ts = ts(2026, 1, 31, 23, 0, 0);
        let close_ts = ts(2026, 2, 1, 2, 0, 0);
        let partial_tp = CloseEvent::new(
            "campaign",
            0,
            "ES",
            Side::Buy,
            partial_ts,
            0.5,
            150.0,
            50.0,
            CloseReason::Target,
        );
        let final_sl = CloseEvent::new(
            "campaign",
            1,
            "ES",
            Side::Buy,
            close_ts,
            0.5,
            50.0,
            -50.0,
            CloseReason::Stoploss,
        );
        let completed = CompletedPosition::from_close_events(
            "campaign",
            "ES",
            Side::Buy,
            open_ts,
            close_ts,
            1.0,
            100.0,
            None,
            None,
            Vec::new(),
            vec![partial_tp.clone(), final_sl.clone()],
            None,
            None,
            0.001,
        );
        assert_eq!(completed.outcome, NetPnlOutcome::Breakeven);
        let result = BacktestResult::from_future_artifacts(FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                pnl_epsilon: 0.001,
                ..ExecutionMetadata::default()
            },
            close_events: vec![partial_tp, final_sl],
            completed_positions: vec![completed],
            ..FutureBacktestArtifacts::default()
        });

        // Close-event fields remain legacy-compatible.
        assert_eq!(result.total_trades, 2);
        assert_eq!(result.winning_trades, 1);
        assert_eq!(result.losing_trades, 1);

        // Campaign analytics count only the completed net outcome.
        assert_eq!(result.total_positions, 1);
        assert_eq!(result.winning_positions, 0);
        assert_eq!(result.losing_positions, 0);
        assert_eq!(result.streaks.max_consecutive_wins, 0);
        assert_eq!(result.streaks.max_consecutive_losses, 0);
        assert_eq!(result.streaks.current_streak, 0);
        let duration = result
            .duration_stats
            .expect("one completed campaign has duration stats");
        assert_eq!(duration.avg_duration_secs, 4 * 3600);
        assert_eq!(duration.min_duration_secs, 4 * 3600);
        assert_eq!(duration.max_duration_secs, 4 * 3600);
        assert_eq!(duration.avg_winner_duration_secs, 0);
        assert_eq!(duration.avg_loser_duration_secs, 0);
        assert_eq!(result.monthly_returns.len(), 1);
        assert_eq!(result.monthly_returns[0].year, 2026);
        assert_eq!(result.monthly_returns[0].month, 2);
        assert_eq!(result.monthly_returns[0].trade_count, 1);
        assert_eq!(result.monthly_returns[0].pnl, 0.0);
        assert_eq!(result.monthly_returns[0].ending_balance, 10_000.0);
    }

    #[test]
    fn future_position_and_provider_statistics_use_configured_breakeven_outcome() {
        let completed =
            completed_position("tiny", 0.0005, 0.001, ts_hms(9, 0, 0), ts_hms(11, 0, 0));
        assert_eq!(completed.outcome, NetPnlOutcome::Breakeven);
        let artifacts = FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                pnl_epsilon: 0.001,
                ..ExecutionMetadata::default()
            },
            close_events: completed.close_events.clone(),
            completed_positions: vec![completed],
            ..FutureBacktestArtifacts::default()
        };

        let result = BacktestResult::from_future_artifacts(artifacts);
        let performance = &result
            .provider_evaluation
            .as_ref()
            .expect("future reports include provider evaluation")
            .position_performance
            .as_ref()
            .expect("position performance requested");

        assert_eq!(result.total_positions, 1);
        assert_eq!(result.winning_positions, 0);
        assert_eq!(result.losing_positions, 0);
        assert_eq!(performance.wins, 0);
        assert_eq!(performance.losses, 0);
        assert_eq!(performance.breakeven, 1);
        assert_eq!(performance.total_outcome.value, Some(0.0005));
        assert_eq!(performance.gross_positive.value, Some(0.0));
    }

    #[test]
    fn legacy_from_trade_log_keeps_exact_zero_position_classification() {
        let result = BacktestResult::from_trade_log(10_000.0, vec![make_trade(0.0005, 11)]);

        assert_eq!(result.total_positions, 1);
        assert_eq!(result.winning_positions, 1);
        assert_eq!(result.losing_positions, 0);
        assert_eq!(result.position_win_rate, 1.0);
    }

    // ── Integration tests ───────────────────────────────────────────

    #[test]
    fn full_report_matches_summary() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Sell,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p3",
                "XAUUSD",
                Side::Buy,
                200.0,
                ts(2026, 1, 2, 10, 0, 0),
                ts(2026, 1, 2, 13, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        // Summary should match top-level fields.
        assert!((result.summary.total_pnl - result.total_pnl).abs() < f64::EPSILON);
        assert_eq!(result.summary.total_trades, result.total_trades);
        assert_eq!(result.summary.winning_trades, result.winning_trades);
        assert_eq!(result.summary.losing_trades, result.losing_trades);
        assert!((result.summary.win_rate - result.win_rate).abs() < f64::EPSILON);
        assert!((result.summary.profit_factor - result.profit_factor).abs() < f64::EPSILON);
    }

    #[test]
    fn per_symbol_sums_to_overall() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "XAUUSD",
                Side::Buy,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
            make_trade_full(
                "p3",
                "GBPUSD",
                Side::Sell,
                80.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 13, 0, 0),
                CloseReason::Target,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        let sym_total_trades: usize = result.per_symbol.values().map(|s| s.total_trades).sum();
        let sym_total_pnl: f64 = result.per_symbol.values().map(|s| s.total_pnl).sum();

        assert_eq!(sym_total_trades, result.total_trades);
        assert!((sym_total_pnl - result.total_pnl).abs() < 1e-10);
    }

    #[test]
    fn per_side_sums_to_overall() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "EURUSD",
                Side::Sell,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        let side_trades = result.long_stats.total_trades + result.short_stats.total_trades;
        let side_pnl = result.long_stats.total_pnl + result.short_stats.total_pnl;

        assert_eq!(side_trades, result.total_trades);
        assert!((side_pnl - result.total_pnl).abs() < 1e-10);
    }

    #[test]
    fn display_does_not_panic_with_new_fields() {
        // Empty.
        let r1 = BacktestResult::from_trade_log(10_000.0, vec![]);
        let _ = format!("{}", r1);

        // Single trade.
        let r2 = BacktestResult::from_trade_log(10_000.0, vec![make_trade(100.0, 11)]);
        let _ = format!("{}", r2);

        // Mixed.
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                Some("grp1".into()),
            ),
            make_trade_full(
                "p2",
                "XAUUSD",
                Side::Sell,
                -50.0,
                ts(2026, 1, 2, 10, 0, 0),
                ts(2026, 1, 2, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let r3 = BacktestResult::from_trade_log(10_000.0, trades);
        let output = format!("{}", r3);
        assert!(output.contains("Backtest Result"));
        assert!(output.contains("Risk Metrics"));
        assert!(output.contains("Side Breakdown"));
    }

    #[test]
    fn serialized_breakdown_maps_use_stable_key_order() {
        let trades = vec![
            make_trade_full(
                "z",
                "ZZZ",
                Side::Buy,
                1.0,
                ts_hms(9, 0, 0),
                ts_hms(11, 0, 0),
                CloseReason::Target,
                Some("z-group".into()),
            ),
            make_trade_full(
                "a",
                "AAA",
                Side::Buy,
                1.0,
                ts_hms(9, 0, 0),
                ts_hms(12, 0, 0),
                CloseReason::Target,
                Some("a-group".into()),
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        let symbols = serde_json::to_string(&result.per_symbol).expect("symbols serialize");
        let groups = serde_json::to_string(&result.per_group).expect("groups serialize");
        assert!(symbols.find("AAA").unwrap() < symbols.find("ZZZ").unwrap());
        assert!(groups.find("a-group").unwrap() < groups.find("z-group").unwrap());
    }

    #[test]
    fn mtm_output_summary_flows_from_artifacts_and_defaults_for_old_results() {
        let summary = MtmOutputSummary {
            policy: crate::mtm::MtmOutputPolicy::None,
            observed_points: 12,
            retained_points: 0,
            omitted_points: 12,
        };
        let result = BacktestResult::from_future_artifacts(FutureBacktestArtifacts {
            execution: ExecutionMetadata {
                initial_balance: 10_000.0,
                ..ExecutionMetadata::default()
            },
            mtm_output_summary: summary,
            ..FutureBacktestArtifacts::default()
        });
        assert_eq!(result.mtm_output_summary, summary);

        let mut json = serde_json::to_value(&result).unwrap();
        json.as_object_mut().unwrap().remove("mtm_output_summary");
        let restored: BacktestResult = serde_json::from_value(json).unwrap();
        assert_eq!(restored.mtm_output_summary, MtmOutputSummary::default());
    }

    #[test]
    fn serde_roundtrip_enhanced_result() {
        let trades = vec![
            make_trade_full(
                "p1",
                "EURUSD",
                Side::Buy,
                100.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 11, 0, 0),
                CloseReason::Target,
                None,
            ),
            make_trade_full(
                "p2",
                "XAUUSD",
                Side::Sell,
                -50.0,
                ts(2026, 1, 1, 10, 0, 0),
                ts(2026, 1, 1, 12, 0, 0),
                CloseReason::Stoploss,
                None,
            ),
        ];
        let result = BacktestResult::from_trade_log(10_000.0, trades);

        let json = serde_json::to_string(&result).unwrap();
        let restored: BacktestResult = serde_json::from_str(&json).unwrap();

        assert_eq!(restored.total_trades, result.total_trades);
        assert!((restored.total_pnl - result.total_pnl).abs() < f64::EPSILON);
        assert_eq!(restored.summary.total_trades, result.summary.total_trades);
        assert_eq!(restored.positions.len(), result.positions.len());
        assert_eq!(
            restored.per_close_reason.len(),
            result.per_close_reason.len()
        );
        assert_eq!(restored.monthly_returns.len(), result.monthly_returns.len());
    }

    // ── fmt_duration helper tests ───────────────────────────────────

    #[test]
    fn fmt_duration_basic() {
        assert_eq!(fmt_duration(0), "0m");
        assert_eq!(fmt_duration(300), "5m");
        assert_eq!(fmt_duration(3600), "1h 0m");
        assert_eq!(fmt_duration(3660), "1h 1m");
        assert_eq!(fmt_duration(86400), "1d 0h 0m");
        assert_eq!(fmt_duration(90061), "1d 1h 1m");
    }
}