ADX Trend-Strength Filter — Disproof verdict

"Only trade when ADX is above 25" — the standard fix sold to traders whose moving-average system keeps getting chopped up in ranges.

Verdict: dead — grade F, 0 of 2 markets survived, tested on BTCUSDT from 2023-05-16 to 2026-08-30.

At zero cost: -19.1%
With real costs (15bps + 2bps slippage): -20.8%
Out-of-sample: -10.2%
Buy and hold over the same window: +190.9% (the strategy is 211.7 points behind buy and hold)

Rules tested

strategy "ADX Trend-Strength Filter"  (long)
indicators:
  adx = adx(14)
  fast = ema(9)
  slow = ema(21)

entry: when fast crosses above slow and adx > 25
exit:  when fast crosses below slow
# fills next open · fees+slippage on · 70/30 OOS · deflated-Sharpe gate

How this was reconstructed

The 9/21 EMA cross with the prescribed ADX(14) > 25 filter added, so the page can be read directly against the unfiltered 9/21 verdict — the honest test of whether the fix is a fix.