Guide › Testing

Reading the Edge Report

The Edge Report is the result of a walk-forward. It is at the bottom of the Walk-Forward panel and fills in from the first closed fold. A live that came from a plan shows it in its Walk-Forward tab. You can download it as a PDF.

Everything in it is measured on out-of-sample data only: the windows the parameters never saw.

The verdict

The verdict compares how the strategy did on unseen data with how it did on the data it was tuned on. That is WFE (OOS/IS): out-of-sample Sharpe ÷ in-sample Sharpe, the share it kept on unseen data.

WFE (OOS/IS)Verdict
50% or moreRobust
30% to 50%Fragile
Under 30%Overfit
Out-of-sample Sharpe at or under zeroFailed
Cannot be computed (no out-of-sample Sharpe yet, or training Sharpe at or under zero)Inconclusive

Until all the plan's folds have closed, the verdict is marked provisional. How each figure of the report is computed is in The numbers in Luxis.

The eight gates

"EdgeFound Readiness n / 8" counts how many of these the strategy passes.

Performance

  • Sharpe of 1.0 or more
  • Profit factor of 1.3 or more
  • Max drawdown of 15% or less
  • At least 100 out-of-sample trades

Robustness

  • 50% or more of the in-sample result kept
  • Live out-of-sample Sharpe above zero
  • Parameters stable from fold to fold
  • Monte Carlo: even the unlucky 5% of outcomes end above zero

Other sections

  • Fold Ledger: each fold, its winner and its out-of-sample result.
  • Code changes: every change published into the plan.
  • Alpha and Beta, against SPY. Beta 1.0 moves with the market, 0 is independent of it. Alpha is the yearly return left after the market's part.
  • Live Fidelity — Shadow: how closely the live matched its backtest.
  • Caveats: what weakens the result.

How to use it

A strategy that fails a gate is not ready. Few trades and unstable parameters are the two most common reasons, and both mean the result could be luck. See A strategy that lasts.