The Edge That Wasn't
All 81 experiments › #43

#43 — Daily range regime (PDH/PDL)

2026-05-20 (Session 065)
Verdict: negative · ⛔ closed negative — no M5 edge on any param combination
Prior-day high/low as a regime gate — mean-revert inside the range, breakout-follow outside — fails walk-forward in all 128 configurations on all 4 pairs.

Hypothesis — what it tests

Does yesterday's high/low define a tradeable regime at M5 — with price gravitating back to the daily midpoint inside the range, and continuing after a genuine break outside it?

Description

The idea was to use the prior calendar day's high and low (PDH/PDL) as a regime switch: when price sits inside yesterday's range, fade back toward the range midpoint (PDH+PDL)/2; when price closes outside it, follow the breakout with a trailing stop. A Numba-JIT M5 backtest swept 128 configurations per pair (4 minimum-distance filters x 4 mean-reversion exits x 4 breakout trails x 2 session filters) on GBP_JPY, USD_JPY, EUR_JPY and GBP_USD under the project's full backtest-live consistency SOP. Zero of 128 configs passed walk-forward on any pair; the best in-sample result was still -15 pips/day. Diagnosis: FX at M5 scale is momentum-driven, so price above the daily midpoint tends to stay above it (mean-reversion fails), while closes beyond PDH/PDL are dominated by false breakouts (breakout fails). The PDH/PDL signal was judged to have zero predictive value at M5; H4/D1 was noted as possibly different but untested.

Key result

0/128 WF passes on all 4 pairs; best IS -15 p/d (GBP_JPY London-only, 10p min_dist, TP-at-mid, 20p breakout trail); WR 37%, avg trade -2.87 pips

Indicators

Prior-day high/low (PDH/PDL)Daily range midpointTrailing stop

Algorithms

grid searchwalk-forward validationNumba JIT backtest

Code

Interactive version (search, filters, figures): the experiment explorer.

← #42 Grid-trail strategy#44 ATR-adaptive P&F box sweep →
This page is one experiment from the audited record behind the book The Edge That Wasn'tget the book · about the project.