Module normal 

2D grid-search trade optimizer for normal-distribution markets.

Maximises net expected value (EV − collateral) under the trader’s belief, subject to collateral ≤ budget. Both EV and cost are reported in chain units — i.e. λ-scaled with λ = k / ‖p‖₂, matching helpers.cairo:50-176 (scaled_verify_minimum_with_lambda).

λ-scaling (v0.1.3)

Pre-v0.1.3 the inner grid loop mixed units: the EV was λ-scaled (Gaussian-product integral × λ_g) but the cost from normal_collateral was the unscaled −d_min. Two bugs surfaced by Reviewer B’s audit (docs/REVIEW_ITEM3_DRIVER_B.md §6):

• Budget filter mismatch: coll > input.budget compared an unscaled cost against the user’s real-XP budget. With λ ≈ 200×–266× at typical (k, σ), the filter let trades pass that were 200× over budget at the chain charge.
• Mixed-units candidate selection: best_net = ev − coll ranked candidates on incompatible units, so the “best” pick was wrong.

v0.1.3 routes the inner cost through the same λ-scaling optimize_quote_offline applies (crates/deadeye-sdk/src/normal.rs): chain_cost = max(0, λ_f · f(x*) − λ_g · g(x*)) evaluated at the audited stationary point from normal_collateral. Both the budget filter and the candidate selector run on chain_cost, restoring unit consistency with expected_value.

Structs

NormalOptimizationInput

Inputs to optimize_normal_trade.

NormalOptimizationResult

Output of optimize_normal_trade.

OptimizerConstraints

Tunable bounds on the optimizer’s policy region.

Functions

optimize_normal_trade

Picks the highest-net-EV trade in the policy region.