1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
//! Seasonal Z-Score — how far the current bar's return sits from the historical
//! mean return of bars in the *same hour of day*, in standard deviations.
use crate::calendar::civil_from_timestamp;
use crate::ohlcv::Candle;
use crate::traits::Indicator;
const HOURS: usize = 24;
/// Seasonal Z-Score keyed on hour of day.
///
/// For every bar the indicator forms the simple return `close / previous_close - 1`
/// and compares it to the running mean and standard deviation of all prior
/// returns that fell in the *same* local hour (the wall-clock hour of
/// [`Candle::timestamp`](crate::Candle) shifted by `utc_offset_minutes`). The
/// output is `(return - hour_mean) / hour_std`. A bucket needs at least two prior
/// samples before it can emit; a bucket with zero historical variance reports
/// `0.0`. The per-hour statistics use Welford's online algorithm.
///
/// # Example
///
/// ```
/// use wickra_core::{Candle, Indicator, SeasonalZScore};
///
/// let day = 24 * 3_600_000;
/// let mut z = SeasonalZScore::new(0);
/// // Same hour each day so they share a bucket; close grows then jumps.
/// for (i, close) in [100.0, 101.0, 103.0].iter().enumerate() {
/// z.update(Candle::new(*close, *close, *close, *close, 1.0, i as i64 * day).unwrap());
/// }
/// // Fourth same-hour sample has two priors in the bucket -> emits a z-score.
/// let out = z.update(Candle::new(110.0, 110.0, 110.0, 110.0, 1.0, 3 * day).unwrap());
/// assert!(out.is_some());
/// ```
#[derive(Debug, Clone)]
pub struct SeasonalZScore {
utc_offset_minutes: i32,
prev_close: Option<f64>,
count: [u64; HOURS],
mean: [f64; HOURS],
m2: [f64; HOURS],
last: Option<f64>,
}
impl SeasonalZScore {
/// Construct a Seasonal Z-Score indicator with the given UTC offset (minutes).
pub const fn new(utc_offset_minutes: i32) -> Self {
Self {
utc_offset_minutes,
prev_close: None,
count: [0; HOURS],
mean: [0.0; HOURS],
m2: [0.0; HOURS],
last: None,
}
}
/// Configured UTC offset in minutes.
pub const fn utc_offset_minutes(&self) -> i32 {
self.utc_offset_minutes
}
/// Most recent z-score if a populated bucket has produced one.
pub const fn value(&self) -> Option<f64> {
self.last
}
fn z_for(&self, hour: usize, ret: f64) -> Option<f64> {
if self.count[hour] < 2 {
return None;
}
let variance = self.m2[hour] / (self.count[hour] - 1) as f64;
if variance > 0.0 {
Some((ret - self.mean[hour]) / variance.sqrt())
} else {
Some(0.0)
}
}
fn accumulate(&mut self, hour: usize, ret: f64) {
self.count[hour] += 1;
let delta = ret - self.mean[hour];
self.mean[hour] += delta / self.count[hour] as f64;
let delta2 = ret - self.mean[hour];
self.m2[hour] += delta * delta2;
}
}
impl Indicator for SeasonalZScore {
type Input = Candle;
type Output = f64;
fn update(&mut self, candle: Candle) -> Option<f64> {
let civil = civil_from_timestamp(candle.timestamp, self.utc_offset_minutes);
let hour = civil.hour as usize;
let result = if let Some(prev) = self.prev_close {
let ret = if prev == 0.0 {
0.0
} else {
candle.close / prev - 1.0
};
let z = self.z_for(hour, ret);
self.accumulate(hour, ret);
z
} else {
None
};
self.prev_close = Some(candle.close);
if result.is_some() {
self.last = result;
}
result
}
fn reset(&mut self) {
self.prev_close = None;
self.count = [0; HOURS];
self.mean = [0.0; HOURS];
self.m2 = [0.0; HOURS];
self.last = None;
}
fn warmup_period(&self) -> usize {
2
}
fn is_ready(&self) -> bool {
self.last.is_some()
}
fn name(&self) -> &'static str {
"SeasonalZScore"
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::traits::BatchExt;
use approx::assert_relative_eq;
const DAY: i64 = 24 * 3_600_000;
fn c(close: f64, ts: i64) -> Candle {
Candle::new(close, close, close, close, 1.0, ts).unwrap()
}
#[test]
fn metadata_and_accessors() {
let z = SeasonalZScore::new(120);
assert_eq!(z.utc_offset_minutes(), 120);
assert_eq!(z.name(), "SeasonalZScore");
assert_eq!(z.warmup_period(), 2);
assert!(!z.is_ready());
assert!(z.value().is_none());
}
#[test]
fn no_output_until_bucket_has_two_priors() {
let mut z = SeasonalZScore::new(0);
// Each bar shares the same hour bucket (same time-of-day, daily spacing).
assert!(z.update(c(100.0, 0)).is_none()); // first: no return
assert!(z.update(c(101.0, DAY)).is_none()); // return #1 -> bucket has 0 priors
assert!(z.update(c(102.0, 2 * DAY)).is_none()); // return #2 -> bucket has 1 prior
// return #3 -> bucket has 2 priors -> emits.
assert!(z.update(c(104.0, 3 * DAY)).is_some());
assert!(z.is_ready());
}
#[test]
fn z_score_matches_manual_welford() {
let mut z = SeasonalZScore::new(0);
// Returns into one hourly bucket: r1 = 0.01, r2 = 0.02, r3 = 0.03.
z.update(c(100.0, 0));
z.update(c(101.0, DAY)); // r1 = 0.01
z.update(c(103.02, 2 * DAY)); // r2 = 0.02
// Priors {0.01, 0.02}: mean 0.015, sample std = sqrt(((.005)^2*2)/1).
let mean = 0.015;
let std = (((0.01_f64 - mean).powi(2) + (0.02 - mean).powi(2)) / 1.0).sqrt();
let r3 = 0.03;
let expected = (r3 - mean) / std;
let close = 103.02 * (1.0 + r3);
let out = z.update(c(close, 3 * DAY)).unwrap();
assert_relative_eq!(out, expected, epsilon = 1e-9);
}
#[test]
fn zero_variance_bucket_reports_zero() {
let mut z = SeasonalZScore::new(0);
// Constant return into the bucket -> variance 0 -> z = 0.
z.update(c(100.0, 0));
z.update(c(110.0, DAY)); // r1 = 0.10
z.update(c(121.0, 2 * DAY)); // r2 = 0.10
let out = z.update(c(133.1, 3 * DAY)).unwrap(); // r3 = 0.10
assert_relative_eq!(out, 0.0);
}
#[test]
fn zero_prev_close_uses_zero_return() {
let mut z = SeasonalZScore::new(0);
z.update(c(0.0, 0)); // prev close 0
z.update(c(0.0, DAY)); // ret = 0 (guarded), bucket sample
z.update(c(0.0, 2 * DAY)); // ret = 0, bucket now 2 priors
let out = z.update(c(0.0, 3 * DAY)).unwrap();
assert_relative_eq!(out, 0.0);
}
#[test]
fn reset_clears_state() {
let mut z = SeasonalZScore::new(0);
for i in 0..4 {
z.update(c(100.0 + f64::from(i), i64::from(i) * DAY));
}
z.reset();
assert!(!z.is_ready());
assert!(z.value().is_none());
assert!(z.update(c(100.0, 4 * DAY)).is_none());
}
#[test]
fn batch_equals_streaming() {
let candles: Vec<Candle> = (0..50)
.map(|i| c(100.0 + f64::from(i % 9), i64::from(i) * 3 * 3_600_000))
.collect();
let mut a = SeasonalZScore::new(0);
let mut b = SeasonalZScore::new(0);
assert_eq!(
a.batch(&candles),
candles.iter().map(|x| b.update(*x)).collect::<Vec<_>>()
);
}
}