Monolith Research
The final out-of-sample pass was deliberately smaller than comfortable. The first 60% of history remained the formation window, the next 20% was used for validation, and the last 20% was held aside as the stricter test. That final slice is the one recorded here. It was not a broad rediscovery exercise; it was a check on whether the selected overlay still behaved after the notebook had already made up its mind.
On the Top-5 cross-asset OU sleeve, the turn-filter variant returned 4.14% on the final holdout, annualised at 6.84%, with a Sharpe of 2.19 and maximum drawdown of -0.93%. The same sleeve without the overlay returned a little more, 4.43%, but with a lower Sharpe of 1.98 and a slightly deeper -0.97% drawdown. The cleaner statistic is the one that matters: less violence in the path, nearly the same destination.
The strongest single spread was UK100/XAU on 4H bars. It returned 11.79% on the final slice, annualised at 19.94%, with a Sharpe of 3.29 and five closed trades. Five trades is not enough to crown anything. It is enough to keep the pair in the file, with the warning written beside it.
This is a research record, not a signal sheet. The useful observation is narrow: the relative-value sleeve has earned the next round of paper observation.