📈 Sharpe Ratio
Annualized risk-adjusted return of the brain's directional predictions · is it making money
📚 Sharpe ratio tier targets
Annualized SharpeTierInterpretation
< 0losingPredictions cost money on average. Don't trade.
0 – 0.5weakSlightly positive expectancy but unreliable.
0.5 – 1.0fairReal edge but small. Cost of effort vs return is marginal.
1.0 – 1.5goodProfessional-grade. Worth running.
1.5 – 2.5excellentStrong, durable edge. Hedge-fund quality.
> 2.5world-classExtraordinary. Check for overfitting / data leakage.
📐 How this is computed
Per-resolution signed return:
  • If predicted LONG and price went up → positive realized return
  • If predicted LONG and price went down → negative realized return
  • If predicted SHORT and price went down → positive realized return
  • If predicted SHORT and price went up → negative realized return
i.e. "what would a unit bet in the predicted direction have earned?"

Sharpe = mean(signedReturn) / std(signedReturn)

Annualized: short-horizon resolutions are ~10 minutes apart. The trading year has 252 × 6.5 × 60 / 10 ≈ 23,400 such periods. Annualized Sharpe = raw Sharpe × √(periods per year).

Caveats: the Sharpe shown here uses unsigned R-multiples and equal-weighted trades, so it reflects predictor quality, not the actual strategy P&L (which includes size, fees, slippage). Real-trade Sharpe lives in risk-dashboard.html.