Does the breakout rule break in the 2022 bear?
The full sample looks like a pass. 2022’s bear and 2023’s chop still lose money.
Anyone can look good on the easy years. So I break it on purpose first. 947 trades clear the full-sample bar, but 71 trades in 2022 average −0.04R and 2023 Q3 averages −0.18R. Those windows fail.
Rule name: Setup breakout_rr, gate rs_sector_resid, pass bar DSR ≥ 0.95.
Full sample
| metric | value |
|---|---|
| trades | 947 |
| ExpR | 0.2313 |
| SR(R) | 0.58 |
| DSR | 1.00 (pass ≥ 0.95: True) |
| trials | 12 |
Subperiod stress
| window | trades | avg R | sum R | Sharpe | destroy_ok |
|---|---|---|---|---|---|
| full | 947 | 0.231 | 219.06 | 0.58 | True |
| bear_2022 | 71 | −0.04 | −2.85 | −0.11 | False |
| sideways_2023Q3 | 36 | −0.182 | −6.56 | −0.51 | False |
| sideways_2018 | 0 | — | 0 | — | — |
Full-sample DSR passing does not mean the book is regime-proof. 2022 and 2023Q3 fail the destroy check. The Aug 27 breakout paper restart uses these gates anyway, with that failure in view.
Factor note (same day)
A PnL decomposition on 943 residual-gate trades vs SPY and vs SPY+XLK printed R² ≈ 0.011. The annualized-α number from that regression is not a CAGR and is not published here. Low R² is the usable sentence: this R stream is not mostly a hidden SPY/XLK bet.
Paper loop ≠ this 947-trade history.