SD1H4H — Best Validated Version

One row per market × engine × strategy combo (default 1H/4H pair, same universe as the Strategy Priority Ranking page), PLUS one additional row for every validated mechanics/entry/sizing finding on top of a base combo -- currently 5 gold overlays, of which only lookback_10 still clears the 2026-09-01 holdout-respecting bar (Tokyo / atrstop_15 / no_event_bias / conviction sizing are full-sample historical rows — see All Analyses), all in the green Filter column below (a Tokyo/Asian session entry filter, two alternate stop mechanisms, one structure-detection param change, and a conviction-sizing scheme built on the session filter). Each variant row is its own ranked entry, layered on top of but NOT replacing its base combo's own (always-unfiltered/unmodified) baseline row -- the delta line underneath each variant row shows what changed. Every other row's "None (no validated filter)" means candidate filters WERE tested (a VWAP entry-confirmation filter under three different momentum definitions, a VIX daily regime filter, and the session filter swept across all 6 markets) and did not survive walk-forward, not that nothing was tried. See FINDINGS.md for the full writeup of every validated result and HANDOFF.md for every negative one. "Survives costs" mirrors the Trading Costs methodology (Vantage FX / Bybit rate cards). Ranked by the same risk-adjusted score the Priority Ranking page uses -- (MC probability of profit − MC risk of ruin), scaled down for samples under 200 trades, any combo that doesn't survive costs ranked last -- computed from each row's OWN numbers (i.e. a variant row is ranked on its own filtered/modified stats, its base combo's row is ranked on baseline). Not sorted by Monte Carlo median net profit alone: that ignores risk of ruin and drawdown entirely, so a high-median/high-ruin-risk combo would misleadingly outrank a lower-median/low-risk one. Avg R / Total R / PF are the single REALIZED historical result from this strategy's actual trade history. MC Med. PF / MC Med. Avg R are different: the MEDIAN across 2,000 bootstrap-resampled alternate trade sequences (same methodology as Prob. Profit / Risk of Ruin / Median DD, all four Monte Carlo columns share the same simulation) -- a forward-looking view of what's typical if this strategy's own historical trade distribution repeats, not a different or better estimate of the past. The two Avg R figures are usually close (same underlying trade pool) but can diverge on thin samples, where a skewed win/loss distribution pulls the bootstrap median away from the raw historical mean.