okx-codex-trader/docs/live-strategy-status-2026-05-18.md

Live Strategy Status - 2026-05-18

Current live service:

• bb-squeeze-executor.service
• Strategy: BB squeeze live l48-bw960-q25 both vc006
• Symbol: ETH-USDT-SWAP
• Bar: 15m
• Execution mode: live OKX candles, loop polling every 1 second, only newly confirmed candles are processed.
• Status after the manual close on 2026-05-18: service stopped until the exchange-side stop-loss path is deployed and checked.

Current Decision

Keep the live strategy unchanged.

Recent exploration did not find a replacement that is better across enough windows. The current live BB squeeze strategy is weak over the latest 7d/14d windows, but it remains stronger than the newly tested alternatives over the 30d/90d windows and has usable trade frequency.

Do not increase risk until a replacement has positive 30d and 90d performance after fees with enough trades.

Live Risk Control Requirement

Every live opening order must include an OKX exchange-side stop-loss attachment.

The executor uses the strategy stop-loss percentage and submits it through attachAlgoOrds on the opening market order:

• long stop trigger: entry_reference * (1 - STOP_LOSS_PCT)
• short stop trigger: entry_reference * (1 + STOP_LOSS_PCT)
• stop order price: -1, meaning market execution after trigger

The live strategy still has no fixed take-profit. Its intended exit remains the strategy-side middle-band exit or stop-loss. Adding a fixed take-profit would change the traded strategy and requires separate backtesting.

Recent Data State

Local OKX cache was refreshed for low timeframes:

• ETH 3m: through 2026-05-17 09:15 UTC
• ETH 5m: through 2026-05-17 09:10 UTC
• BTC 3m: through 2026-05-17 09:15 UTC
• BTC 5m: through 2026-05-17 09:15 UTC
• ETH/BTC 15m: through 2026-05-17 06:45 UTC

Data validation passed for ETH/BTC 3m, 5m, and 15m: no duplicate timestamps, no gaps, valid OHLCV rows, and cache roundtrip checks passed.

Recent Exploration Read

Market state:

• ETH has been materially weaker than BTC over the latest 30d.
• BTC/ETH same-bar correlation remains high, but BTC lead/lag is not strong enough to use as the main alpha.
• Low-bandwidth squeeze is present, but recent squeeze follow-through is weak and needs fast confirmation.

Explored directions:

• False-breakout mean reversion: rejected. Best 5m ETH short is only near flat over 3y.
• BTC regime router v2: rejected. Trade frequency is high, but fee-adjusted returns remain negative.
• 3m/5m recent squeeze and BTC lead/lag candidates: not live-ready. Some 30d/14d/7d rows are positive, but 90d remains negative.

Acceptable next action:

• Keep current live BB squeeze executor.
• Add only read-only observation for promising 3m/5m candidates if signal telemetry is needed.
• Reconsider strategy replacement only after the observation candidate has at least 30-50 forward signals and positive 30d/90d after fees.

Primary reports:

• reports/live-recent/recent-strategy-report.md
• reports/recent-regime/recent-regime-exploration-summary.md
• reports/recent-regime/recent-regime-low-timeframe-refresh-summary.md