1,620Versions testedevery combination, fixed in advance
+1.49Best versionreturn adjusted for risk, per year
+2.07Luck alonewhat the best scores if nothing works
0.050Scorehas to reach 0.950
Result against chance
If the best version cannot beat what luck alone produces from the same number of tries, there is nothing there. The third bar is the smallest real edge this test could have detected: anything below it is invisible to the test, and that limit is stated rather than hidden.
Spread across every version
Every version tested, worst to best. The box holds the middle half. Showing only the best bar is how noise gets sold as skill. 47% of versions finished above zero.
Why this record is void
Not an independent test. The continuation families were exact position-wise inversions of SW-0002's reversion families, so every statistic returned as the precise arithmetic negative of the original to four decimal places: gross Sharpe -0.2901 became +0.2901, z of -5.542 became +5.542. An exact inversion is an identity, not a replication. Kept rather than deleted: removing a sweep because it turned out uninformative is the selective reporting this registry exists to document.
Notes
The one thing it does establish: gross Sharpe is antisymmetric under position inversion, so any symmetric bias — costs, spread, microstructure — contributes zero to it. Combined with the shuffled-null result, that rules out lookahead as the source of the SW-0002 effect.
Sealed on the Bitcoin blockchain before the test ran. sha256 d2f13ee4cdaa0f54423ec3e15bead9886e6ba8af099e4bbe0f89c3223304169e