Pre-registered · Timestamped on Bitcoin · Public
Thousands of trading strategies are sold as if they work. This is the public record of what happens when they are tested the way a statistician would test them.
So far, none have passed. The closest, Vol-scaled cross-sectional momentum, reached 0.75 against the 0.95 it had to beat.
What will be tested, and the score it has to beat, are written down, hashed, and timestamped onto the Bitcoin blockchain before the data is touched. The story cannot be rewritten once the answer is known.
Test a thousand random strategies and the best will look brilliant by chance alone. Every result is scored against how good pure luck would have looked with the same number of tries.
Failures, near misses, and runs that had to be thrown away all stay on the record with the reason attached. A registry that deletes its own mistakes is asking not to be believed.
| ID | Strategy | Best | Luck | Score | Result |
|---|---|---|---|---|---|
| SW-0002 | Z-score reversion + RSI reversion + Bollinger touchCrypto · 1,620 versions | +0.64 | +2.53 | 0.000 | Did not pass |
| SW-0003 | Z-score continuation + RSI continuation + Bollinger breakoutCrypto · 1,620 versions | +1.49 | +2.07 | 0.050 | Void |
| SW-0004 | Z-score continuation + RSI continuation + Bollinger breakoutCrypto · 1,620 versions | — | — | — | Sealed until 2027-09-30 |
| SW-0005 | Moving average crossCrypto · 630 versions | +0.96 | +1.23 | 0.225 | Did not pass |
| SW-0006 | Donchian breakoutCrypto · 432 versions | +1.13 | +1.21 | 0.415 | Did not pass |
| SW-0007 | ATR breakoutCrypto · 540 versions | +1.26 | +1.49 | 0.255 | Did not pass |
| SW-0008 | Time-series momentumCrypto · 540 versions | +1.11 | +2.34 | 0.000 | Did not pass |
| SW-0009 | Volatility-filtered momentumCrypto · 360 versions | +0.96 | +1.80 | 0.010 | Did not pass |
| SW-0010 | Hour-of-day seasonalityCrypto · 576 versions | +0.65 | +5.76 | 0.000 | Did not pass |
| SW-0011 | Moving average crossFX · 945 versions | +0.81 | +1.69 | 0.019 | Did not pass |
| SW-0012 | Donchian breakoutFX · 648 versions | +1.09 | +2.33 | 0.000 | Did not pass |
| SW-0013 | ATR breakoutFX · 810 versions | +1.10 | +2.82 | 0.000 | Did not pass |
| SW-0014 | Time-series momentumFX · 810 versions | +1.10 | +3.77 | 0.000 | Did not pass |
| SW-0015 | Volatility-filtered momentumFX · 540 versions | +0.88 | +3.41 | 0.000 | Did not pass |
| SW-0016 | Hour-of-day seasonalityFX · 864 versions | +0.79 | +13.99 | 0.000 | Did not pass |
| SW-0017 | Fair value gapCrypto · 540 versions | +0.54 | +2.27 | 0.000 | Did not pass |
| SW-0018 | Order blockCrypto · 540 versions | +0.35 | +3.50 | 0.000 | Did not pass |
| SW-0019 | Liquidity sweepCrypto · 540 versions | +0.64 | +1.68 | 0.000 | Did not pass |
| SW-0020 | Break of structureCrypto · 540 versions | +1.69 | +1.71 | 0.482 | Did not pass |
| SW-0021 | Engulfing candleCrypto · 540 versions | +0.55 | +4.44 | 0.000 | Did not pass |
| SW-0022 | Inside bar breakoutCrypto · 540 versions | +0.41 | +1.44 | 0.002 | Did not pass |
| SW-0023 | Fair value gapFX · 810 versions | +0.65 | +4.63 | 0.000 | Did not pass |
| SW-0024 | Order blockFX · 810 versions | +0.87 | +7.46 | 0.000 | Did not pass |
| SW-0025 | Liquidity sweepFX · 810 versions | +0.85 | +2.50 | 0.000 | Did not pass |
| SW-0026 | Break of structureFX · 810 versions | +0.80 | +2.49 | 0.000 | Did not pass |
| SW-0027 | Engulfing candleFX · 810 versions | +0.68 | +8.39 | 0.000 | Did not pass |
| SW-0028 | Inside bar breakoutFX · 810 versions | +0.78 | +2.40 | 0.000 | Did not pass |
| SW-0029 | Cross-sectional momentumCrypto · 90 versions | +0.00 | +0.00 | 0.500 | Void |
| SW-0030 | Cross-sectional reversalCrypto · 90 versions | +0.00 | +0.00 | 0.500 | Void |
| SW-0031 | Vol-scaled cross-sectional momentumCrypto · 90 versions | +0.00 | +0.00 | 0.500 | Void |
| SW-0032 | Cross-sectional momentumCrypto · 90 versions | +0.98 | +0.64 | 0.809 | Void |
| SW-0033 | Cross-sectional reversalCrypto · 90 versions | +0.29 | +0.65 | 0.177 | Void |
| SW-0034 | Vol-scaled cross-sectional momentumCrypto · 90 versions | +0.99 | +0.62 | 0.823 | Void |
| SW-0035 | Cross-sectional momentumCrypto · 90 versions | — | — | — | Sealed until 2029-06-30 |
| SW-0036 | volume_breakout + obv_trend + volume_climaxCrypto · 1,080 versions | +0.86 | +7.92 | 0.000 | Did not pass |
| SW-0039 | Cross-sectional momentumCrypto · 45 versions | +0.97 | +0.86 | 0.605 | Did not pass |
| SW-0040 | Cross-sectional reversalCrypto · 45 versions | -0.20 | +1.01 | 0.001 | Did not pass |
| SW-0041 | Vol-scaled cross-sectional momentumCrypto · 45 versions | +1.04 | +0.77 | 0.745 | Did not pass |
Best is the top version's return adjusted for risk. Luck is what that version would have scored if none of them worked. Score is how likely the result is real rather than selection, and it has to reach 0.95. Open any row for the full record.
These were run with purpose-written code rather than the sweep engine. They carry no returns matrix, no deflated Sharpe ratio, no probability of backtest overfitting and no Bitcoin timestamp, and they are not counted in any figure above. They are published because the measurements are clean and the findings useful, and kept apart because calling them sweeps would break the one guarantee this registry makes.
EXT-0038 · 144 versions
NO EDGE FOUND
On the exploration half all 15 highest-mean variants were LONG, not short, despite medians of -2.7% to -18.9%: the positive mean comes entirely from the upper tail. Only 5 of 72 short variants had both mean and median positive.
EXT-0039 · 84 versions
NO EDGE FOUND
Zero of 84 combinations positive. All 14 symbols negative, most with t between -3 and -7.
EXT-0040 · 480 versions
NO EDGE FOUND, MECHANISM CONFIRMED
Zero of 24 surviving combinations positive; 292 of 480 were discarded by the sanity bound.