okx-codex-trader/reports/short-bias

Short-Bias Strategy Summary

Date: 2026-05-08

Direction Exposure

The currently selected and deployed candidates are too long-biased:

• Remote eth-nextgen-micro observer currently maps nextgen state to long or flat.
• rotation-risk is BTC/ETH long-only rotation plus cash.
• ETH RSI/TWAP candidates are long-only pullback entries.
• Some research modules support short trades, but the promoted live/shadow candidates do not yet use a real short-biased leg.

Results

Intraday Short-Only

Reject.

BTC/ETH 3m/5m/15m short-only tests did not produce a usable candidate. Best row was only +1.41% total with 21.78% drawdown and negative 1y/6m/3m returns.

Swing Short-Only

Keep for follow-up validation.

Best candidate:

• vol_expansion_short-btc-4H-fast20-slow80-entry20-exit10-atr14-stop_atr3.0-take_atr6.0-max_hold120-vol_window120-vol_quantile0.8
• Total return: 285.27%
• Annualized return: 23.75%
• Max drawdown: 24.93%
• Calmar: 0.95
• Trades: 75
• Win rate: 46.67%
• Profit factor: 2.11
• Short trade ratio: 1.0
• 3y/1y/6m/3m returns are all positive.

This is usable as a research candidate, but not yet as a live candidate because drawdown is materially higher than rotation-risk.

Short Overlay

Keep as the most practical next direction.

A small short overlay on top of rotation-risk improved risk behavior:

• Standalone rotation-risk max drawdown: about 7.72%
• With roughly 5% short overlay: max drawdown falls to about 6.50%
• Annualized return falls from about 15.64% to about 14.65% to 15.05%
• Recent 6m improves from about -0.28% to roughly +1.1% to +2.5%

This is not a strong standalone short strategy, but it is useful as a small protection leg during risk-off or high-volatility regimes.

Decision

Do not replace the current long-biased strategy with short-only intraday trading.

Next best step: build a read-only signal stream for rotation-risk + small short overlay, then compare it against plain rotation-risk on live candles before any real short orders are considered.