#32 — P&F momentum vs counter-trend

2026-05-18 (Session 061)

Verdict: mixed · 🟡 mixed — counter-trend wins but is 10–15× weaker than FIFO-Trends

Head-to-head on Point & Figure entries, counter-trend (fading N down-boxes) beats momentum on 8 of 12 pairs, but the best edge is only ~+5 p/d versus FIFO-Trends' ~70 p/d.

Hypothesis — what it tests

Whether counter-trend P&F entries (fading a run of boxes) beat momentum P&F entries (following a run of boxes), same apparatus, all 12 pairs.

Description

A user-prompted 'try both ways' test: the same Point & Figure entry sweep run with both momentum entries (buy after N up-boxes) and counter-trend entries (buy after N down-boxes), head-to-head on all 12 pairs — 6,912 backtests in total. Counter-trend won on 8 of 12 pairs with higher peak values (best: USD_JPY +5.39 p/d OOS), and the strongest configurations waited for an 80-pip move (10-pip box, entry_n=8) before fading it with a wide 100-pip trail. But the aggregate means were negative for both directions, most top-OOS configs had negative IS (a reliability red flag), and even the best result was 10-15x weaker than the momentum-based FIFO-Trends system (USD_JPY +68.5, GBP_JPY +71.6 p/d). Research closed: both entry directions on P&F bounces are dominated by the existing system.

Key result

Counter-trend wins 8/12 pairs (best USD_JPY +5.39 p/d OOS); aggregate means mom −5.35 / ctr −4.98 p/d; only 10–21% of configs positive; vs FIFO-Trends USD_JPY +68.5 / GBP_JPY +71.6 p/d — 10–15× weaker

Indicators

P&F (Point & Figure) charttrailing stop (TR10, 100-pip trail)

Algorithms

grid search parameter sweepIS/OOS split validationNumba JIT backtest

Code

research/experiments/mean_rev_pnf/backtest_pnf_mean_rev.py

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

← #31 Currency-strength lower-TF viability #33 TR Momentum strategy →

This page is one experiment from the audited record behind the book The Edge That Wasn't — get the book · about the project.