The Edge That Wasn't
All 81 experiments › #35

#35 — FIFO-Trends S5 trail resolution

2026-05-21 (Session 068)
Verdict: negative · ⛔ closed negative — box quantization IS the edge
Replacing FIFO-Trends' coarse P&F-box high-water mark with a 5-second price HWM flipped EUR_USD from +45.2 to −69.2 p/d — proving the 5-pip box step is the strategy's noise filter, not a limitation.

Hypothesis — what it tests

Whether 5-second (S5) trail monitoring with a raw-price high-water mark beats the slow M5 P&F box-level trail on FIFO-Trends.

Description

A natural 'improvement' to the FIFO-Trends P&F trend-follower: instead of the high-water mark advancing only in 5-pip Point & Figure box steps (M5 polling), track the actual price HWM every 5 seconds so the trail hugs intra-box peaks. The backtest compared the two trail modes directly on EUR_USD (Feb 2024 → May 2026 M5/S5 overlap) and EUR_JPY (Oct 2025 → Apr 2026). The result was decisively negative: the S5 price-HWM trail collapsed EUR_USD from +45.2 p/d OOS at 60% win rate (baseline) to −69.2 p/d at 19% win rate. With the HWM advancing on every 5-second tick, the trail becomes too tight for normal tick noise and stops out nearly every trade. The lesson inverted the premise: the P&F box quantization deliberately filters sub-5-pip noise and is the edge mechanism itself — the strategy requires coarse-grained HWM tracking.

Key result

EUR_USD OOS: M5 P&F-level trail +45.2 p/d / 60.0% WR vs S5 price-HWM trail −69.2 p/d / 19.0% WR (avg win +11.0p→+10.2p, avg loss −6.6p→−4.5p)

Indicators

P&F (Point & Figure, 5-pip box)high-water-mark trailing stop

Algorithms

A/B backtest comparison (two fill/trail modes, one code path)IS/OOS split validation

Code

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

← #34 S5 velocity/acceleration sweep#36 FIFO-Trends bar-close fill correction →
This page is one experiment from the audited record behind the book The Edge That Wasn'tget the book · about the project.