Range extremes are just yesterday's trend

Verdict: NO_EDGE · Asset/TF: BTCUSDT 1h and 4h · Sample: chronological 60/20/20

Hypothesis: when close sits at the extreme of a causal 96-bar range (location ≤0.1 or ≥0.9), the next bars either mean-revert toward mid-range or continue as a breakout — more than mid-range bars / more than trend alone.

Q1 / Q2

Both fade (A) and breakout (B) fail on validation for H∈{1,3,6,12}. Mid→extreme events are rare (1h val n=36/48 by side; 4h even thinner). Means near null; P(toward mid) ≈ coin.

Ablation (KEY)

Extremes are almost pure trend labels: P(past 24h return + | high extreme event) ≈ 0.995, P(past 24h − | low) ≈ 0.995. The extreme does not add information beyond the 24h trend sign.

4h discovery looked mildly breakout-ish; validation did not replicate (year signs flip). That would be OVERFIT if cherry-picked.

Trading

Q3 skipped. Exploratory next_open PnL t≈0.4 at 0 bps; dies by 5 bps.

Final Verdict

NO_EDGE

“Price at the edge of the range” is a restatement of recent trend, not an independent fade or breakout signal on this sample. Do not build range-location strategies from these thresholds alone.