#44 — ATR-adaptive P&F box sweep

Hypothesis — what it tests

Does a volatility-adaptive P&F box size (ATR x multiplier, frozen per column) adapt to regimes better than a fixed 5-pip box?

Description

A test of whether the FIFO-Trends Point & Figure charts should size their boxes dynamically from volatility (box = atr_mult x Wilder ATR, re-anchored and frozen at each new column) instead of using a fixed 5-pip box. The backtest swept 3,600 configs per pair (3 ATR periods x 5 multipliers x 3 reversals x 5 n_min x 8 exits) on 4 pairs through the standard 3-stage pipeline (IS walk-forward, Monte Carlo, OOS). Every pair's best adaptive config converged to mult=0.5 — roughly the 5-pip fixed box for JPY pairs — meaning ATR added no adaptive value, and OOS results were worse than fixed boxes on all three JPY pairs. The one apparent gain, GBP_USD (+9.5 p/d), turned out to be a sub-5-pip box discovery rather than an ATR benefit, and it spawned the follow-up small-box sweep (entry 45). Verdict: 5 pips is the noise floor; adaptivity doesn't beat it.

Key result

Indicators

Algorithms

Code