BTC/USDT 1H
3-class directional model validated on out-of-sample data with bootstrap significance testing.
Out-of-Sample Results
Metrics below are from backtesting on data the model never saw during training — not live trading.
1.36
Profit Factor
Every $1 risked returns $1.36
2.67:1
Risk-to-Reward
Winners 2.67x larger than losers
+$86
Expectancy
Avg profit per trade in backtesting
< 0.001
p-value
Statistically significant edge
How It Was Tested
Walk-forward cross-validation — the gold standard for time-series model testing.
5 Sequential Folds
Data is split chronologically into 5 expanding windows. Each fold trains on past data and tests on future data — never the reverse.
Embargo Period
A time gap is enforced between training and test data to prevent any information leakage across the boundary.
No Lookahead Bias
The model is never exposed to future data during training. Predictions are evaluated exactly as they would be in live trading.
Rolling Stability
The model was independently trained and tested on 3 non-overlapping 1-year windows to verify the edge persists across different market conditions.
Year 1
Mar 2023 — Feb 2024
Year 2
Feb 2024 — Feb 2025
Year 3
Feb 2025 — Feb 2026
All 3 windows were profitable. 2 out of 3 reached statistical significance (p < 0.05). The most recent year shows the strongest edge.
Live Performance
Want the full history with live updates? Open the live dashboard
Out-of-sample metrics are from backtesting on unseen data. Live performance metrics are from real-time signal tracking. Past performance is not indicative of future results. This is not financial advice. Full disclaimer