MMT

Optimize · Find better weights and thresholds

Two questions: (1) what lens-weight mix maximises composite IC vs forward S&P returns? (2) at what entry threshold does each strategy hit its best Sharpe? Both run against cached backtest data — no look-ahead, no source re-fetches.

Backtest cadence: every 14d · IC horizon: 20d

Lens-weight optimization

Random simplex search (~2,000 candidates) over weights ≥ 0 summing to 1. Sign-aware variant additionally tries every per-lens flip combination — reveals which lenses the data wants to be flipped.

Searching simplex (~2,000 candidates × 64 sign vectors)… takes 30–60 s on first load.

Threshold sweep

Sweeps the entry threshold for each strategy and reports the full risk-return curve. The optimum is picked by Sharpe; ties by CAGR. Buy-and-hold Sharpe shown as the dashed reference.

Long when score ≥ T
Sweeping thresholds for Directional…
connecting…