I Slowed My Trading Bot Down to Four Trades a Year. It Tripled a Backtest, and It Still Failed.

Описание и применение

Автор замедлил торговый бот до четырёх сделок в год, чтобы снизить влияние комиссий. Стратегии показали +200 % дохода за почти пять лет, но не прошли строгий тест Gate 1 из‑за недостаточного количества сделок, низкого Шарпа и превышения лимита прибыли одной сделки. Контрольный тест с случайными входами подтвердил, что результаты не превосходят случайность.

Задача

Fees on a $100 Kraken account at the lowest tier (0.80 % taker + 0.10 % slippage) ate all profits from frequent‑trading strategies.

Результат

The slow‑trade strategies achieved high net returns but failed the gate because of insufficient trade count, Sharpe, and single‑trade profit limits. A matched‑null control showed the strategies were not statistically superior to random timing.

Рабочий процесс

• 1. Define four slow‑trade rules (trend filter, dual momentum, vol target, trend×vol).
• 2. Freeze all parameters and gate rules before coding.
• 3. Backtest each rule on Kraken candle data from Jan 2022 to Sep 2026 with realistic fees.
• 4. Record net return, max drawdown, Sharpe, and trade count.
• 5. Apply Gate 1: require ≥100 closed trades, in‑sample Sharpe ≥1.0, walk‑forward Sharpe, and no single trade >30 % of profit.
• 6. Build a matched‑null control that randomises entry dates but preserves exposure, run 200 simulations.
• 7. Compare strategy performance to null percentiles.
• 8. Identify gate failures and analyze reasons.
• 9. Document findings and plan next steps.

Доказательства

Fees were about 7 % of its gross winnings. Doubling the fees to 1.80 % a fill still left it at +174 %.

Источник

All four failed it, and not narrowly.

Источник

It beat 84 % of them. The rule was 95 %.

Источник

The bug did not change any verdict, in either round.

Источник