#68 — First-touch H4 low-volume reversion

Hypothesis — what it tests

Whether the first touch of a fresh H4 swing level mean-reverts (rehearsals deplete resting orders), and whether thin tick-volume at the touch (no breakout conviction) is the discriminating filter.

Description

Born from the user's reasoning that repeated touches consume resting orders, so the FIRST touch of a fresh H4 swing high/low should revert — and that a touch arriving on thin volume signals no breakout conviction. v1 (plain fade of touches≤1 with a ~48h hold, net spread) was already encouraging — positive in all three calendar thirds, stationary 2021→2026 — but not deploy-grade (per-trade MC P≈0.41, fat tails). The v2 upgrade the same day added an ATR target/stop (2·ATR each, 48h cap) and split trades by tick-volume at the touch: the low-volume half delivers OOS +9.49 pips/trade, WR 54%, MC P(≤0)=0.018, walk-forward positive in all 3 thirds (+2.9/+2.3/+14.7), 7/12 pairs OOS-positive (GBP_JPY flips −16 to +32 under the low-vol filter), while high-volume touches fail (−6.28p, P=0.88 — heavy volume means a genuine break). The user's third idea (sharper approaches revert more) did not hold. It fits the only surviving regime in the project — contrarian at structure with a multi-day hold that amortizes the spread — was deployed as the fx-first-touch-paper container, and a later trait-miner refined a +32p conviction-size overlay. Notably, the recipe does NOT generalize to previous-day-high/low levels — it is the level type, not the timeframe. Caveats: p=0.018 from one pass, and 5/12 pairs don't work.

Key result

Indicators

Algorithms

Code