backtest
Regime gate matrix — bear and sideways
base vs rs_sector vs flow vs full vs SIT_FLAT on breakout_rr across stress windows.
Setup: breakout_rr · min_rr 2.0 · universe n=38
Question: which gate set survives bear and sideways tapes?
Bear 2022 (selected)
| Gate | Trades | WR% | ExpR | Port% | MDD |
|---|---|---|---|---|---|
| base | 111 | 32.4 | −0.162 | −17.2% | −27.4% |
| rs_sector | 74 | 41.9 | 0.080 | 5.5% | −16.6% |
| flow | 39 | 43.6 | 0.092 | 3.3% | −8.9% |
| market | 67 | 22.4 | −0.434 | −25.6% | −25.6% |
| full | 12 | 33.3 | −0.346 | −4.1% | −5.9% |
| SIT_FLAT | 0 | — | — | 0.0% | 0.0% |
Sideways 2018 (selected)
| Gate | Trades | WR% | ExpR | Port% | MDD |
|---|---|---|---|---|---|
| base | 228 | 39.5 | 0.078 | 16.9% | −33.6% |
| rs_sector | 125 | 48.0 | 0.307 | 44.8% | −17.3% |
| flow | 64 | 40.6 | 0.095 | 5.6% | −15.3% |
| full | 40 | 42.5 | 0.171 | 6.7% | −12.7% |
| SIT_FLAT | 0 | — | — | 0.0% | 0.0% |
Sideways 2023 Q3 (selected)
Most gate sets negative; SIT_FLAT control correctly stays flat. rs_sector still negative (−25.4% port) — regime sit is sometimes mandatory.
Winner pattern
Per-window winners favor rs_sector or flow in choppy/bear slices — not the heaviest full gate. Default paper uses rs_sector accordingly.
Source
Summarized from reports/regime_sit_compare_bear_sideways.md.