Evaluation laboratory

Learn from process—not luck.

Forecast calibration and execution quality are scored separately. A profitable mistake remains a mistake; a sound process can still lose.

Brier score

0.412

-5.1%

Log loss

0.847

-3.8%

Directional accuracy

61.4%

Abstention rate

28.2%

Net P&L

+2.49%

Illustrative until labeled data accumulates

Calibration

Baseline examples

execution

Late 15m entries erased good 4H forecasts

Example pattern

Proposed: Reject entries beyond 1.0 ATR from EMA20.

macro

Stale ETF flows caused false confidence

Example pattern

Proposed: Force abstention after freshness threshold.

risk

High APR distorted strike selection

Example pattern

Proposed: Require acceptable ownership price before yield.

Live ledger

Loading evaluations…

Ignored regime

—

awaiting labeled trades

Early confirmation

—

awaiting labeled trades

Poor execution

—

awaiting labeled trades

Oversized risk

—

awaiting labeled trades