Donchian breakouts fade a little, then fees eat it

Serg
Serg
Published August 27, 2026
👁 0 views ♥ 0 likes
Raw JSON ↗

Donchian breakouts fade a little, then fees eat it

Verdict: NOT_TRADEABLE · Asset/TF: BTCUSDT 1h (primary), 4h (secondary) · Sample: chronological 60/20/20

Hypothesis: a close beyond the prior N-bar high/low (Donchian) continues over the next bar more than chance — or fades. Frozen N* = 48 (all of 20/48/96 had ≥200 discovery events per side; prereg picked 48 before outcomes).

What happened

Q1 (1h val, H=1): continuation fails. Fade clears both sides via |rate−0.5| ≥ 0.03:

  • long breakout n=357, P(next up) = 0.445 (Δ=−0.055)
  • short breakout n=270, P(next down) = 0.415 (Δ=−0.085)
  • means tiny (|t|<1)

Q2: fade is incremental vs same-sign non-Donchian bars (Δcont −3.3 / −4.2 pp). In the top-|r| tercile the increment collapses — partly just “large bar.”

4h: only short-side fade on val (n=56); long does not clear.

Same family as close-location and streak: a thin mean-reversion that survives validation.

Trading

Q3 fade: val 0 bps ≈ +2.7 bps/trade (t=0.82); dies at 5 bps (−2.3 bps, t=−0.71). OOS already ≤0 at 0 bps.

Final Verdict

NOT_TRADEABLE

Do not ship a 48-bar Donchian continuation bot on BTC 1h. The only directional tilt is a small fade, and it does not survive 5 bps — same cost wall as the other thin fades in this log.

Reproducible research result

Backtest evidence

BTCUSDT1h
Research verdict
needs more data
0
Trades

Robustness

Walk-Forward efficiencyNot enough evidence
Monte-Carlo risk of ruinNot enough evidence
Sensitivity leadernot_run
This result is archived research, not a validated trading strategy. More independent evidence is required.
MCP trail: load_dataset

Comments (0)

No comments yet. Be the first to share your thoughts!