📊 SAR v6 Tri-Mode: Full 6-Way Comparison
Generated 2026-06-22 10:55 UTC | 41 windows × 33 pairs | 3.5 years Jan 2023 – Jun 2026
⚠️ v6 with shorts is the worst performer overall.
Adding trend-short entries destroyed v4’s profitability.
Shorts in crypto SAR clusters suffer violent squeeze losses that exceed floor stops in single candles.
Aggregate 3.5-Year Results
Realistic Capital
| Strategy | Tot PnL | Avg/Win | Trades | Win/Loss/0 |
|---|
| sar_cluster_v1 | -5,316.7 | -129.7 | 2996 | +16/-25/0=0 |
| sar_cluster_v2 | -2,417.6 | -59.0 | 1917 | +20/-21/0=0 |
| sar_cluster_v3 | -1,310.6 | -32.0 | 1528 | +15/-17/0=9 |
| sar_cluster_v4_long | 1,421.2 | 34.7 | 339 | +13/-19/0=9 |
| sar_cluster_v5 | 1,421.2 | 34.7 | 339 | +13/-19/0=9 |
| sar_cluster_v6 | -5,195.7 | -126.7 | 346 | +18/-14/0=9 |
Unlimited Capital
| Strategy | Tot PnL | Avg/Win | Trades | Win/Loss/0 |
|---|
| sar_cluster_v1 | -4,160.3 | -101.5 | 9128 | +16/-25/0=0 |
| sar_cluster_v2 | -529.1 | -12.9 | 3014 | +20/-21/0=0 |
| sar_cluster_v3 | -373.9 | -9.1 | 2360 | +15/-17/0=9 |
| sar_cluster_v4_long | 2,463.8 | 60.1 | 1026 | +12/-20/0=9 |
| sar_cluster_v5 | 2,463.8 | 60.1 | 1026 | +12/-20/0=9 |
| sar_cluster_v6 | -4,273.4 | -104.2 | 992 | +17/-15/0=9 |
Where v6 Bleeds vs v4 (Realistic $100K)
| Window | v4 PnL | v6 PnL | Delta | Trades v4 | Trades v6 |
|---|
| 20241022-20241221 | 3496 | -3752 | -7248 | 17 | 19 |
| 20240126-20240326 | 2920 | -2964 | -5884 | 13 | 13 |
| 20231028-20231227 | 2747 | -2787 | -5535 | 13 | 13 |
| 20230101-20230302 | 2292 | -2858 | -5150 | 11 | 14 |
| 20230928-20231127 | 2016 | -2497 | -4513 | 12 | 13 |
| 20250619-20250818 | 1273 | -2111 | -3384 | 10 | 11 |
| 20240922-20241121 | 1274 | -1021 | -2295 | 10 | 10 |
| 20260316-20260515 | 547 | -1002 | -1549 | 10 | 11 |
| 20230829-20231028 | 653 | -339 | -992 | 10 | 10 |
| 20250520-20250719 | 409 | -568 | -976 | 10 | 10 |
Findings
- v4_long (no shorts) is the only profitable variant over the full period (+$1,421 realistic).
- Gate + PCA + dynamic stops successfully filtered out most chop.
- v6 shorts fire in windows where regime flips to bear, but crypto bear moves are asymmetric — sharp drops followed by violent reversals. The -10% floor stop is gapped through in single 15m candles during short squeezes.
- Range logic (v5 = v4) never fires because low-variance chop is nearly nonexistent in 33-pair crypto PCA.
- Shorts with regime gating are still lethal — the PCA regime detector has 160-candle (~40h) lag, so bear flips are detected late, and shorts enter AFTER most of the drop is over, just in time for the bounce.