Research · test #7
Does rising open interest plus extreme funding predict reversals? No edge
The idea: extreme funding shows a crowded side, and open interest that keeps rising shows the crowd is still adding leverage - so the squeeze should be bigger. We already tested funding on its own and found no edge. Does adding open interest fix it? We wrote the rules down first and ran them on 522 Binance coins. It does not: with open interest added, the result is the same as funding alone.
In short: 1.79 million funding payments with open interest known, 2024-01-01 to 2026-10-20. Short when funding is +0.05% per 8 hours or more and open interest rose 10% or more in 24 hours: 3,553 trades, 57% winners, average -0.34% after costs and funding. Long when funding is -0.05% or lower with the same open interest rise: 7,297 trades, 40% winners, average +0.07%. Funding alone: -0.37% and +0.10%.
The rules (fixed before the run)
- Coins: the 522 crypto perpetuals on Binance USDT-M from our funding test (tokenized stocks and commodities excluded) that were trading on 21 October 2026.
- Open interest: Binance public 5-minute history, the last value no older than 30 minutes at the payment and 24 hours earlier. Payments without it are skipped (3,166 of 1.80 million).
- Result = price move in the trade's direction + every funding payment during the hold - 0.15% for fees and slippage. One trade per coin and side per 24 hours.
- Control 1: the same position at every payment with open interest known (random samples of the same size). Control 2: funding alone, the same 0.05% threshold on the same payments, so the only difference is the open interest condition.
- Edge requires: average above zero, t above 2 and p below 0.05 on the first part (to 30 June 2025) and the second part (from 1 July 2025) separately, beating "funding alone", with at least 30 trades in each part.
This is the second run of the same rules. The first used only the 20 largest coins and ended undecided: 48 shorts and 55 longs in 2.8 years, only 3 and 18 on the second part - too few to judge. We extended it to all coins with a new written record before loading the data.
Result: no edge on either side
| 24-hour hold | Trades | Coins | Average | Median | Winners | t | vs funding alone, p |
|---|---|---|---|---|---|---|---|
| Short, 1/24-6/25 | 1,454 | 269 | -0.61% | +0.69% | 54% | -1.73 | 0.88 |
| Short, 7/25-10/26 | 2,099 | 198 | -0.16% | +2.13% | 59% | -0.31 | 0.23 |
| Short, all | 3,553 | 412 | -0.34% | +1.38% | 57% | -1.03 | 0.47 |
| Long, 1/24-6/25 | 2,089 | 289 | +0.17% | -1.02% | 44% | 0.69 | 0.43 |
| Long, 7/25-10/26 | 5,208 | 460 | +0.03% | -1.93% | 38% | 0.13 | 0.63 |
| Long, all | 7,297 | 479 | +0.07% | -1.65% | 40% | 0.40 | 0.58 |
No line comes close to t of 2. The long also "beats" the plain pool of all payments (p 0.035 overall), but only because that pool, every long on every payment, loses money; the long's own average is indistinguishable from zero.
Open interest adds nothing
For each case the bar with the open interest condition and the bar without it are almost the same height. On the second part the short is -0.16% with open interest and -0.40% without, but with t of -0.31 that gap is noise; it still loses money. Funding alone fired 22,186 times for shorts and 28,978 for longs; adding "open interest up 10%" kept 16% and 25% of them.
Other thresholds
| Rule (24-hour hold) | Trades | Average | t | Winners |
|---|---|---|---|---|
| Short, funding +0.05%, open interest +5% | 6,279 | -0.48% | -2.26 | 55% |
| Short, funding +0.05%, open interest +10% (tested) | 3,553 | -0.34% | -1.03 | 57% |
| Short, funding +0.05%, open interest +20% | 1,748 | -0.55% | -0.98 | 58% |
| Short, funding +0.10%, open interest +10% | 1,797 | -0.72% | -1.31 | 57% |
| Long, funding -0.05%, open interest +5% | 10,506 | +0.06% | 0.44 | 42% |
| Long, funding -0.05%, open interest +10% (tested) | 7,297 | +0.07% | 0.40 | 40% |
| Long, funding -0.05%, open interest +20% | 4,557 | +0.03% | 0.10 | 38% |
| Long, funding -0.10%, open interest +10% | 5,736 | +0.05% | 0.24 | 39% |
Stricter open interest or funding did not rescue either side. These rows are descriptive; only the tested row counts for the verdict.
What else we looked at
- The 72-hour exit. Shorts were worse (-1.24%, median +3.02%: wins are common, the losses are large). Longs averaged +0.62% (t 1.91) with a median of -2.63% and 40% winners - a few large rebounds carry the average. It is not the main exit and does not enter the verdict.
- Concentration. The shorts come from 412 coins, the most from small coins such as SIREN (45 trades) and AIOT (44); no single coin dominates.
What this test does not say
- Coins delisted before 21 October 2026 are missing. Those coins may have had these conditions more often, so the result for them could differ. We cannot say by how much.
- Entries are at the hourly close. In a fast move, slippage can be larger than the 0.15% we deducted, especially on small coins.
- Trades on the same coin follow each other and are not independent; the number of coins is shown next to every line.
- Open interest from Binance only; the market-wide open interest, and a different window than 24 hours, were not tested.
Data and method
Rules were committed before any data was loaded, and the code before the run; the code and every trade are kept with the project. Data: Binance public futures funding history, hourly candles and 5-minute open interest. The same approach as in our other tests: a random control from the same coins and period, two time periods checked separately, costs included. Live data: open interest, funding rates on four exchanges, funding heatmap. Explained: what open interest really shows.
FAQ
Does rising open interest make extreme funding a better signal?
No. With funding of at least 0.05% per 8 hours in either direction and open interest up 10% or more in 24 hours, a 24-hour short averaged -0.34% after costs and funding (3,553 trades) and a long +0.07% (7,297). Funding alone gave -0.37% and +0.10%.
Is rising open interest with high funding a short signal?
Not in our data. 57% of the shorts won, but the average was -0.34%, and -0.16% on the second half of the data. A few squeezes outweighed the many small wins.
Does rising open interest with negative funding mean a squeeze is coming?
We found no tradable edge. A 24-hour long won 40% of the time and averaged +0.07% after costs (t 0.40); on the second half of the data +0.03%.
Why was the first test of this idea undecided?
It used only 20 large coins and found 48 shorts and 55 longs in 2.8 years - too few to decide. The same rules on all 522 coins gave 3,553 and 7,297 trades.