Monolith Research
The profit pass tested the tempting ideas first: Grey sizing, spread-health filters, volatility filters, dynamic sizing, and the simple turn check. The winner was not the cleverest mechanism. It was the one that asked a blunt question before entry: has the spread already begun moving back toward zero?
On the held-out 40% test window, the baseline Top-5 OU sleeve carried an 8.73% annualised return, Sharpe 1.62, and maximum drawdown of -1.20%. The turn-filter variant kept the return almost intact at 8.60%, lifted Sharpe to 1.83, and reduced maximum drawdown to -1.00%. The health filter went slightly lower on drawdown at -0.97%, but gave up more return and did not improve the ratio as cleanly.
Grey sizing produced the highest annualised return in the pass, 9.56%, but it did so by accepting more path risk and a worse drawdown profile. That makes it interesting, not first. The turn filter is quieter: less return-chasing, better discipline, almost no extra machinery.
The note goes into the archive as a preference for small mechanical improvements over decorative complexity. The next question is not whether the filter looks elegant in the notebook. It is whether the same restraint survives forward paper observation.