Strategy Lab
How the live Phase 37 meta-portfolio picks sleeves dynamically.
What you see on /lab
/lab shows the current live meta-portfolio. It is dynamic — every walk-forward rebalance, the orchestrator re-ranks the entire sleeve catalog on train-period Sharpe, drops the bottom 5 (except hedge sleeves), and allocates Sharpe-weighted to the top-N.
When market regime shifts (e.g. BTC breaks down through its 200-day MA), low-Sharpe sleeves get dropped and crash/hedge sleeves get more weight automatically.
How to read a sleeve row
Each sleeve row shows: - Train Sharpe — Sharpe ratio over the most recent 180-day train window - OOS contribution — cumulative out-of-sample PnL contributed - Weight — current allocation % - Max hold / signal type — the underlying rule (Donchian breakout, EMA cross, momentum)
Why some sleeves look bad standalone but get allocation
Hedge sleeves (short-only vol-spike, e.g. SHORT-VOL-SPIKE-BTC) have low standalone Sharpe but provide uncorrelated PnL during BTC regime-off periods. They're exempt from the "drop bottom 5" filter for that reason.