A few days ago we asked ourselves: what if we used funding rates as a trading signal for BTC?
The logic seems compelling. Funding rates on perpetual futures are the cost of keeping a long position open. When they're strongly positive, longs are paying shorts — that happens when the market is overheated, with too many people betting on rising prices. When they go negative, the dynamic reverses.
From this you can derive two signals:
BUY signal: Extreme negative funding → shorts dominating → contrarian buy. Filter: Extreme positive funding → longs overheated → block existing buy signals.
Sounds reasonable. So we tested it.
The Method
Dataset: Daily BTC funding rates from September 2019 to June 2026. That's 2,462 trading days, averaged across Binance and OKX.
We computed a 30-day rolling Z-score. This normalizes the signal across market regimes — "extremely high" always means the same thing regardless of whether we're in a bull or bear market.
Z-score distribution:
- Z > +1.5 (extreme longs): 9% of days
- Z < −1.5 (extreme shorts): 8.3% of days
- Neutral zone: 82.7% of days
Rare enough to be meaningful — or so we thought.
We tested two scenarios:
Scenario A — Standalone strategy: Buy when Z-score first crosses below −1.5. Sell when it returns above 0, or after 30 days.
Scenario B — Filter: Take buy signals from another strategy — but block them when Z-score is above +1.5.
The Results
Scenario A produced 93 trades over 7 years:
- Win rate: 53.8%
- Avg return per trade: +1.14%
- Avg hold time: 5 days
Technically slightly positive. After fees, probably zero or negative. Not a viable strategy.
Scenario B is where it gets surprising:
| Condition | Avg 30d Return | Win Rate |
|---|---|---|
| Normal funding (allowed) | +4.23% | 54.1% |
| High funding, Z > 1.5 (blocked) | +6.95% | 62.3% |
| Very high funding, Z > 2.5 | +9.87% | 64.5% |
The filter would have blocked exactly the best entry points.
Why?
Because high funding rates don't signal reversals — they accompany strong trends.
In a real bull market, funding stays elevated for weeks. The market keeps punishing short sellers while the trend continues. The mean-reversion effect we were counting on exists — but it's too weak and too delayed to be visible within a 30-day window.
High funding = the trend is running. Not an exit signal.
What We Take Away
Funding rate Z-score is not a useful signal or filter for BTC entries in this form.
One could argue: combine it with trend context. Block entries only when funding is high and momentum divergence is visible. But at that point that's essentially a momentum-divergence signal — the funding component contributes little.
We're shelving this idea for now.
The most important thing here isn't the result. It's the process: formulate an idea, build a concrete implementation, test it on real data, report honestly.
Sometimes the answer is: doesn't work. That's worth just as much as a positive result — it saves time and protects against false confidence.
Methodology: Daily funding rates (Binance + OKX average), BTC spot prices, 30-day rolling Z-score. No transaction costs modeled. Not financial advice.