High-confidence signals: backtest brief

Volatility-scaled strategy · Confidence scores 6–7 and 8–10 only

This is a short summary of the same backtest described in the full analysis: each trade risks up to $100, with stop and target distances set from recent volatility. Here we focus on the strongest signal bands—where average results were clearly positive in this sample.

What this shows in plain terms

When the model’s confidence score was in the top bands (6–7 or 8–10), trades had a higher win rate and a better balance of gross wins vs. gross losses (profit factor) than the full sample. Together, these two bands account for of trades in the backtest window ().

The 8–10 band is a small slice—useful as directional evidence, not as proof the edge will repeat. The 6–7 band has more trades, so the averages are somewhat more stable in statistical terms (still not a guarantee out-of-sample).

1. How trades ended (share of wins, losses, time exits)

Each row is 100% of trades in that confidence band. “Time exit” means neither stop nor target was hit before the maximum holding period.

2. Cumulative P&L through trades in each band

Horizontal axis is progress through trades in that band only (chronological order), scaled to 0–100% so both curves span the chart. Dollar amounts are not adjusted for fees or slippage.

Legend: Confidence 6–7 · Confidence 8–10

Same methodology and limitations as the full post: end-of-day prices, no transaction costs, in-sample backtest. For deployment, validate out-of-sample and add realistic costs.