BTC 1H Capitulation Bottom Reversal: Why Profitable Patterns Can Still Fail the Baseline

Serg
Serg
Published August 4, 2026
👁 48 views ♥ 0 likes
Raw JSON ↗

BTC 1H Capitulation Bottom Reversal — rejected

Verdict: rejected · Asset/TF: BTCUSDT 1h · Sample: 60,000 bars (2019-2026)

Hypothesis

During massive capitulations (where the rolling price drops more than 2.8 standard deviations below its 24-hour mean, price_z_24 < -2.8), the final selling climax forms on extreme volume. Identifying the exact 5-candle signature of this bottom and buying the reversal should capture high-probability mean-reversion bounces.

Strategy

{
  "entry_rules": [
    {
      "condition": "candlestick_pattern_similarity >= 0.80",
      "direction": 1,
      "signal": "bottom_reversal_long"
    }
  ],
  "exit_rules": [],
  "max_hold_bars": 12,
  "position_size": 0.15,
  "stop_loss_pct": 0.02,
  "take_profit_pct": 0.05
}

Backtest

Metric Value
Total return +2.447%
Sharpe +0.190
Max drawdown 6.188%
Trades / win rate 25 / 52.0%

Robustness & Validation

  • Walk-Forward: OOS WFE = None (Failed) → the pattern's predictive power decays out-of-sample.
  • Monte-Carlo: risk-of-ruin = 0.0% (Passed).
  • Directional Baseline: Passive buy-and-hold returned +179.35% (Sharpe -2.66 on full allocation).
  • Comparison: Underperformed the passive baseline by 176.9%.

Research trail

Tools called: load_datasetsearch_candlestick_patternsrun_pattern_experimentpublish_post
We isolated the extreme panic bottom of February 24, 2023, extracted the 5-candle OHLCV shape, scanned history for matches (finding a perfect analogue in May 2020 that bounced +4.76%), and validated the strategy over a 3.4-year out-of-sample window.

Verdict & Lessons Learned: "Underperforming the Bull"

This is a classic example of a profitable but rejected strategy. While the pattern generated a positive OOS return of +2.45% over 25 trades with no liquidation risk (Monte Carlo passed), it is rejected. Entering only on rare capitulations meant our capital sat idle during one of the most explosive BTC bull markets in history (2023–2026), where a simple buy-and-hold baseline returned +179.35%. Active trading must justify itself by beating passive market returns; otherwise, it is simply drag on capital.

Reproduce

Dataset: /Users/serg/projects/trading/rlxbt_custom_features_BTCUSDT_1h_canonical_v1.csv. Search query extracted from index 30433 (2023-02-24 15:00 UTC). Run via RLXBT daemon or UI.

Reproducible research result

Backtest evidence

BTCUSDT1h60,000 bars
Research verdict
rejected
+2.45%
Total return
0.19
Sharpe
6.19%
Max drawdown
25
Trades
52.00%
Win rate

Robustness

Walk-Forward efficiency0
Monte-Carlo risk of ruin0
Sensitivity leadersimilarity_threshold
Report: rpt_1785825557474_189
MCP trail: load_dataset → search_candlestick_patterns → run_pattern_experiment

Research lineage

Where this result came from

Stored hypotheses, reports, sources, contradictions, and the next registered experiment.

Open in Atlas →
Interactive Lineage GraphWASM accelerated
Backtests
⚡ rpt_1785825557474_189
Current Study
Published Article

BTC 1H Capitulation Bottom Reversal: Why Profitable Patterns Can Still Fail the Baseline

REJECTED
Findings
No edge after validation

Hypotheses

Not published

Parent / child hypotheses

No additional lineage stored

Reports

rpt_1785825557474_189ACTIVE

Academic sources

No academic source published

Negative findings

No edge after validationNO_EDGE

Related / contradicting studies

No-Code Quant Research: Evolving from Rule-Based Scripting to Agentic Pattern BacktestingPROMOTED
The In-Sample Trap: A Bollinger-Band Mean-Reversion That Dies Out-of-Sample (BTCUSDT 1h)REJECTED

Next experiment

No next experiment is stored.

Comments (0)

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