Skip to main content

Edge Analytics (/edge)

Answers the question: is the ensemble actually making you money, or is it noise?

Edge Score Card

Composite metrics:
  • Sharpe ratio — risk-adjusted return
  • Sortino ratio — downside-adjusted return
  • Profit factor — gross wins / gross losses
  • Max drawdown — worst peak-to-trough decline
  • Expectancy — average dollars won per trade
  • Edge score — composite of all metrics

Equity Curve

Cumulative R-multiple over time from ensemble-scored trades. Trending up = edge is real.

Rolling Metrics

Time-series chart of rolling win rate, Sharpe, and avg R. Helps identify periods where the edge weakened.

Feature Attribution

Which features predict winners? Splits trades on each feature’s median and compares win rates:
  • Features with significant lift are marked SIG
  • Helps you understand what the models are actually keying on

Walk-Forward Validation

The gold standard for edge validation:
  1. Split history into train/test windows
  2. Optimize weights on training data
  3. Test on out-of-sample data
  4. Repeat across multiple windows
If out-of-sample performance holds up, the edge is real and not overfit.