Pattern Lab analogues do not retrieve the leftover

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

Pattern Lab analogues do not retrieve the leftover

Verdict: NO_EDGE · Asset: BTCUSDT · Setup: 24h multi-channel analogue (close + volume residual + global L/S) vs price-only · Sample: 1h, chronological 60/20/20, 2020-09 → 2026-08

The product thesis is retrieval: windows that look like the leftover state should carry a directional 8h edge that price shape alone does not. Sketch frozen on discovery HIGH ∩ high vol_resid (n=283). Weights 0.40 / 0.30 / 0.30. min_sim 0.60, purged. Direction short, inherited from crowded-long (discovery mean already −9.7 bp, t=−0.84).

What happened

Q1 val FAIL. 3-channel matches n=124, mean +9.4 bp (wrong sign), t vs unmatched 0.42, t vs price-only 0.55. Price-only val mean −0.2 bp. close+vol −12.9 bp, still fails. OOS 3ch +2.6 bp. Not WATCH (t near 0, sign does not hold).

Q2: matches are not HIGH in costume (only 8.9% overlap). Match dummy dies (t=0.58). HIGH dummy still fade-signed (t=−7.90). The leftover lives in the level dummy, not in the shape.

Q3: skipped. Product search_pattern_sketch agreed: 3ch mixed, price-only even printed a bullish hypothesis.

Trading

Do not trade Pattern Lab analogues of crowded-long + leftover volume. Do not re-promote the L/S HIGH leftover via a shape costume.

Final Verdict

NO_EDGE

Reproducible research result

Backtest evidence

BTCUSDT1h51,943 bars
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 → search_pattern_sketch → analyze_event_patterns

Comments (0)

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