Zilla Arena record
The six month record, market by market
Six frontier models each received an identical brief per market: design one trading rule from a visible training window, on our candles, with our backtest engine as the referee. One design per model per market was locked, and a sealed six month scoring window was then opened exactly once. No design was rescored, swapped, or quietly dropped.
These pages exist so the claim can be checked rather than admired. Every market is shown, including the one where every design lost. Every entrant is shown, including the ones that failed. Where a result depends on a caveat, the caveat is printed at the same size as the result.
Sealed window Feb 13 2026 to Aug 14 2026 · 150 configurations searched on training before the seal broke · spread charged · commission and swap not charged
Every position was priced at a whole bitcoin instead of a hundredth of one, so this board measured a sizing bug rather than a market.
Re-run on the corrected engine. On the first re-run the best design came out at +2.4% with no account wiped. The run stays in the records; the claim comes down until the numbers are measured again.
Every position was twenty times too large, so the two accounts that went to zero were the sizing rule, not the metal.
Re-run on the corrected engine, where the smallest lot is 50 ounces rather than 1,000. The run stays in the records; the claim comes down until the numbers are measured again.
Every page shows the full field, including the models that lost, the samples too small to trust, and the checks that failed. The machine records live in the repository under backend/arena_matrix_out, one file per market, and the pages are regenerated from them.