J3P v2.4 — Regime-Filter Matrix (WIDE 52 names, ~14d)
Each row adds a filter to the v2.4 engine (shorts + Dale-faithful trend + CBO).
⚠️ Directional only — one choppy 14-day window (kti cap); judge on deeper data once the historical adapter is in. Filters: Dale/Shannon/VCET-grounded.
options & sources → ·
registry →
Verdict: Best so far: BEST-HELPERS: grade A/B + Shannon-40 at -0.120R. Still negative — filtering reduced the bleed but did not create edge on this window.
| Variant | Trades | Win% | Avg R | Long n/R | Short n/R | P&L |
|---|
| v2.3 baseline (long-only, no filter) | 951 | 35.2% | -0.194 | 930/-0.199 | 21/+0.018 | $-18,420 |
| v2.4 + shorts | 1329 | 33.3% | -0.182 | 930/-0.199 | 399/-0.144 | $-24,196 |
| + rotation-block (true break only) | 283 | 22.6% | -0.207 | 152/-0.257 | 131/-0.148 | $-5,864 |
| + market-regime (trade w/ SPY&QQQ) | 572 | 30.9% | -0.225 | 372/-0.226 | 200/-0.222 | $-12,853 |
| + Shannon 40-MA direction | 972 | 33.0% | -0.174 | 663/-0.195 | 309/-0.129 | $-16,918 |
| + grade A/B only | 547 | 26.9% | -0.130 | 300/-0.160 | 247/-0.093 | $-7,097 |
| COMBINED: rot-block + market + A/B | 127 | 18.9% | -0.193 | 53/-0.157 | 74/-0.218 | $-2,470 |
| BEST-HELPERS: grade A/B + Shannon-40 | 413 | 27.4% | -0.120 | 246/-0.147 | 167/-0.079 | $-4,950 |
"Long n/R" and "Short n/R" show how each side performs — the v2.4 short side (CBO-short + Dale-trend-short) is new. The COMBINED row is the "only trade when everything lines up" test (true break + market aligned + A/B grade). Fewer trades is expected and desired if the kept ones are better.
Reading this
- Did filtering help? Compare each filter's Avg R to the v2.3 baseline.
- Do shorts help? Look at the Short n/R column vs Long n/R.
- Best combo is highlighted — but on 14 days treat it as a hint, not proof.