Do funding-rate extremes work?
Funding at 0.05% per eight hours or more (five times the usual 0.01%). Selling on it would have missed gains: a year later the price had changed by a median +166%, against +65.9% for a typical day of the same cycles; 1 of 8 signals was on within 90 days of a cycle top, catching 1 of the 2 tops in its record. With 3 independent signals, chance cannot be ruled out.
Negative funding (shorts paying longs). No reliable edge: a year later the price had changed by a median +69.7%, against +65.9% for a typical day of the same cycles; 3 of 18 signals were on within 90 days of a cycle bottom, catching 2 of the 2 bottoms in its record.
- Funding at 0.05% per eight hours or more (five times the usual 0.01%): 8 signals; 1 of 8 decided was on within 90 days of a cycle top (7 false alarms); it caught 1 of the 2 cycle tops in its record.
- Negative funding (shorts paying longs): 19 signals; 3 of 18 decided were on within 90 days of a cycle bottom (15 false alarms); it caught 2 of the 2 cycle bottoms in its record.
What followed the signals
Funding at 0.05% per eight hours or more (five times the usual 0.01%)
| After | Signals | Median change | Mean change | Share up | All days: median | All days: share up | Same cycles: median |
|---|---|---|---|---|---|---|---|
| 7 days | 8 8 independent | +0.3% | +1.4% | 50% | +0.4% | 52% | +0.6% |
| 30 days | 8 8 independent | +3.2% | +5.5% | 62% | +1.9% | 55% | +2.0% |
| 90 days | 8 5 independent | +14.3% | +33.2% | 62% | +6.5% | 57% | +12.3% |
| 365 days | 8 3 independent | +166% | +180% | 88% | +49.1% | 63% | +65.9% |
Negative funding (shorts paying longs)
| After | Signals | Median change | Mean change | Share up | All days: median | All days: share up | Same cycles: median |
|---|---|---|---|---|---|---|---|
| 7 days | 19 19 independent | −0.4% | +1.0% | 42% | +0.4% | 52% | +0.6% |
| 30 days | 18 18 independent | −2.7% | +1.0% | 39% | +1.9% | 55% | +2.0% |
| 90 days | 18 14 independent | +6.4% | +10.7% | 61% | +6.5% | 57% | +12.3% |
| 365 days | 16 5 independent | +69.7% | +123% | 62% | +49.1% | 63% | +65.9% |
Every signal
Funding at 0.05% per eight hours or more (five times the usual 0.01%)
| First day | Reading | BTC close | 7 days | 30 days | 90 days | 365 days | Nearest cycle top | Near a turn |
|---|---|---|---|---|---|---|---|---|
| 28 Oct 2019 | 0.0549% | $9,222 | +2.2% | −18.3% | −6.7% | +48.0% | 681 days after the 2017 top (16 Dec 2017); price −52.3% from it | False alarm |
| 19 Jan 2020 16 days in signal | 0.0506% | $8,703 | −1.2% | +17.0% | −16.5% | +321% | 659 days before the 2021 top (8 Nov 2021); price −87.1% from it | False alarm |
| 9 May 2020 | 0.0586% | $9,541 | −1.6% | +2.5% | +21.6% | +511% | 548 days before the 2021 top (8 Nov 2021); price −85.9% from it | False alarm |
| 27 Jul 2020 10 days in signal | 0.0620% | $11,043 | +1.7% | +3.8% | +18.1% | +258% | 469 days before the 2021 top (8 Nov 2021); price −83.7% from it | False alarm |
| 22 Nov 2020 68 days in signal | 0.0780% | $18,431 | −1.3% | +29.3% | +203% | +205% | 351 days before the 2021 top (8 Nov 2021); price −72.7% from it | False alarm |
| 21 Oct 2021 4 days in signal | 0.0627% | $62,284 | −2.7% | −4.0% | −33.1% | −69.2% | 18 days before the 2021 top (8 Nov 2021); price −7.8% from it | Hit |
| 2 later signals | Included in Plus |
| Cycle top | BTC close | Signal on within 90 days |
|---|---|---|
| 8 Nov 2021 | $67,549 | Yes: the signal from 21 Oct 2021, which began 18 days before it |
| 6 Oct 2025 | $124,766 | Included in Plus |
Negative funding (shorts paying longs)
| First day | Reading | BTC close | 7 days | 30 days | 90 days | 365 days | Nearest cycle bottom | Near a turn |
|---|---|---|---|---|---|---|---|---|
| 13 Nov 2019 19 days in signal | −0.0007% | $8,774 | −7.8% | −17.2% | +17.0% | +85.9% | 333 days after the 2018 bottom (15 Dec 2018); price +171% from it | False alarm |
| 13 Mar 2020 35 days in signal | −0.1272% | $5,633 | +10.2% | +22.7% | +64.5% | +986% | 454 days after the 2018 bottom (15 Dec 2018); price +74.2% from it | False alarm |
| 15 Sep 2020 15 days in signal | −0.0101% | $10,787 | −2.3% | +6.7% | +78.7% | +346% | 640 days after the 2018 bottom (15 Dec 2018); price +234% from it | False alarm |
| 19 May 2021 28 days in signal | −0.0146% | $36,772 | +6.9% | −2.6% | +21.5% | −17.6% | 551 days before the 2022 bottom (21 Nov 2022); price +133% from it | False alarm |
| 24 Sep 2021 | −0.0041% | $42,848 | +12.4% | +42.0% | +18.6% | −55.8% | 423 days before the 2022 bottom (21 Nov 2022); price +172% from it | False alarm |
| 12 Jan 2022 47 days in signal | −0.0002% | $43,920 | −5.1% | −3.5% | −8.7% | −57.1% | 313 days before the 2022 bottom (21 Nov 2022); price +179% from it | False alarm |
| 15 Aug 2022 14 days in signal | −0.0013% | $24,102 | −11.2% | −16.0% | −32.3% | +21.0% | 98 days before the 2022 bottom (21 Nov 2022); price +52.9% from it | Hit |
| 27 Oct 2022 12 days in signal | −0.0003% | $20,295 | −0.4% | −18.9% | +13.7% | +67.1% | 25 days before the 2022 bottom (21 Nov 2022); price +28.8% from it | Hit |
| 11 later signals | Included in Plus |
| Cycle bottom | BTC close | Signal on within 90 days |
|---|---|---|
| 21 Nov 2022 | $15,760 | Included in Plus |
| 30 Jun 2026 | $58,532 | Included in Plus |
Method and sources
Signals. Top signal: funding at 0.05% per eight hours or more (five times the usual 0.01%); Bottom signal: negative funding (shorts paying longs). The levels are the ones commonly quoted for this indicator, fixed before any result was looked at. A signal starts on the first day the condition holds after it has been off for at least 30 days, so a run of days counts once (the same rule as the signal lab).
Point in time. Each day is the mean of the funding rates Binance settled on its BTCUSDT perpetual that day, as an eight-hour rate; nothing later enters a day. Binance is the venue whose public funding history reaches back to September 2019. Daily readings are closed UTC days from 10 September 2019 to 26 October 2026; the open day is never used.
Outcomes. The change of the BTC daily close from the close of the signal's first day to the close 7, 30, 90 and 365 days later, against the same change from every day the indicator has a reading (the base rate). Signals whose windows overlap share the same price moves, so the tables also count independent signals: windows that do not overlap, counted from the first. Recent signals wait for their outcomes.
Cycle tops and bottoms. A cycle top is an all-time-high close followed by a fall of at least 75% before any higher close; the latest all-time high starts the current cycle. A cycle bottom is the lowest close between two tops; the current cycle's bottom is its lowest close so far, so it can still move. A signal is a hit when a turn of its kind came while it was on or within 90 days either side (from 90 days before its first day to 90 days after its last), and a false alarm otherwise; a signal still on, or ended less than 90 days ago, waits. The verdict compares each signal with the median change of every day in the same four-year halving cycle, so the outsized early years neither flatter nor penalise later signals.
Plans. The chart needs Plus: the statistics cover its whole history for everyone, the signals listed and drawn stop at 31 December 2022 without Plus, and today's reading comes with Plus.
Sources. Our own Bitcoin node and BTC/USD price record, through the published chart linked below. Past signals describe what happened; they are not a forecast and nothing here is investment advice.
Questions
Do funding-rate extremes work?
How accurate are funding-rate extremes?
Have funding-rate extremes given false signals?
Are funding-rate extremes signalling now?
How is this backtest calculated?
Each signal is the first day Bitcoin funding meets its level after at least 30 days without it, using only data that existed that day. We measure the BTC price change 7, 30, 90 and 365 days later against the same change from every day and from the days of the same four-year cycle, and the distance to the nearest cycle top or bottom.
Open the Bitcoin funding rate chartAll free tools
Signals and outcomes are calculated by CryptoKeyhole from our own Bitcoin node and price data, using only data that existed on each day. Past signals describe what happened; they are not a forecast, and nothing here is investment advice.