Direction edge experiment 72k+

Experiment Summary

Random Forest · 30s direction target

completed

Train rows: 4419
Test rows: 1895
Default confidence: 0.8
Accuracy: 0.3789
Progress stage: -
Started: 2026-06-30 12:06:38.580422+02:00
Heartbeat: -
Completed: 2026-06-30 12:06:57.099864+02:00

Training Dataset Diagnostics

Eligibility funnel captured when this experiment trained.

Training diagnostics unavailable

This experiment artifact does not include training diagnostics.

Class Metrics

Class	Precision	Recall	F1
long	0.38	0.3847	0.3823
short	0.3507	0.5083	0.4151
flat	0.5048	0.2043	0.2909

Confusion Matrix

Actual \ Predicted	Long	Short	Flat
Long	277	388	55
Short	277	336	48
Flat	175	234	105

Feature Importance

Top 20 features reported by the trained model.

cross_exchange_spread_percent

0.026138

ask_wall_score

0.025839

cross_exchange_price_deviation_percent

0.025359

mid_price_delta

0.025347

wall_shift_score

0.024931

candle_return_3

0.024767

candle_momentum_5

0.023843

order_book_imbalance_20

0.023437

imbalance_5_delta

0.023191

bid_wall_score

0.023068

capture_latency_ms

0.022897

imbalance_10_delta

0.022863

spread_percent

0.022182

order_book_imbalance_10

0.02213

ask_depth_5_delta

0.021879

volume_24h

0.021692

bid_depth_10_delta

0.021251

ask_depth_10_delta

0.021032

price_change_percent_24h

0.020824

order_book_imbalance_5

0.020642

Probability Diagnostics

Maximum class probabilities on the held-out rows used by shadow backtesting.

Probability diagnostics unavailable

No labeled held-out rows remain for diagnostics.

Run Shadow Backtest

Historical simulation on held-out test rows only. No orders or live predictions.

Shadow Backtest Results

Expectancy values are percentages per accepted historical signal.

Edge	Confidence	Signals	Long / Short	Win rate %	Avg win %	Avg loss %	Raw expectancy %	Net expectancy %	Max drawdown %	Profit factor	Fees + slip %	Created
Weak / Insufficient	0.8	0	0 / 0	-	-	-	-	-	-	-	0.15	2026-06-30 12:07