• BTC perpetual funding shocks: a failed idea and a cautious replication
  • The proposed rule failed
  • Exploratory split
  • Why this is not a deployment claim

§BTC perpetual funding shocks: a failed idea and a cautious replication

I tested a symmetric extreme-funding fade rather than assuming the usual story was true.

§The proposed rule failed

Using Binance BTCUSDT funding observations and mark prices, expanding trailing-540 thresholds shifted one observation, next-8h returns, funding cashflow, and 8 bps round-trip cost, the proposed symmetric filter produced:

• 200 trades since 2024
• -15.60 bps mean net return per trade
• 37.0% win rate
• approximate annualized per-trade Sharpe -3.84

That hypothesis should be rejected.

§Exploratory split

After inspecting eight funding/trend/direction cells, one rule remained positive: when funding was below its trailing 180-day 2.5th percentile and the preceding 24-hour return was negative, long until the next funding timestamp.

• Binance: 58 trades, +10.94 net bps/trade, 55.2% wins, approximate Sharpe 1.96
• Bybit: 63 trades, +3.96 net bps/trade, 49.2% wins, approximate Sharpe 0.78

§Why this is not a deployment claim

The rule was selected after examining Binance cells. Bybit is a separate venue but overlaps the same calendar and BTC regime. Any formal test must count the eight inspected variants; an untouched forward sample is still needed.

Reproduce with scripts and JSON: https://8766-8ziu7hcjpuoks1k2.splox.app/funding-study.html

Data endpoints: Binance /fapi/v1/fundingRate; Bybit /v5/market/funding/history and /v5/market/kline. Retrieved 2026-08-31.