After the Weekend Selloff, Bitcoin Funding Turns Negative: Reading the Derivatives Data

Bitcoin perpetual funding turned slightly negative after the weekend selloff. See funding, open interest, long/short and DVOL data, with reports flagged.

After the weekend selloff in bitcoin, public derivatives data show a clear change in the cost of holding leveraged long positions: funding rates on bitcoin perpetual futures at two large venues, which had been positive on essentially every settlement for two weeks, slipped to roughly zero on November 15 and turned slightly negative at several settlements from November 13 onward. Open interest in dollar terms fell, a measure of implied volatility from Deribit jumped, and the share of accounts holding long positions stayed high. This article sets out those readings from the venues’ public application programming interfaces, explains how to interpret each, and separates what the data show from what has been claimed about liquidations.

The numbers come from three exchanges’ public endpoints. The venues are cited only as data sources. The article does not compare their fees, products or reliability, and it does not draw conclusions about any one venue’s users.

For the underlying contract mechanics, see how perpetual-futures funding and liquidations work. For the scheduled macroeconomic backdrop, see the CPI, PPI and FOMC calendar guide.

Key takeaways

  • On Binance’s BTCUSDT perpetual, the funding rate was positive at every settlement from October 31 through the second settlement of November 15 in the file, and the only negative prints were −0.000247 percent at the third settlement of November 15 and −0.001265 percent at the first settlement of November 16 (values read from the funding history file). Binance settles every eight hours, three times a day, with the first settlement at the day boundary.
  • On OKX’s BTC-USDT perpetual, funding was negative at seven settlements in the file: the second settlement of November 7, the second of November 13, the second of November 14, the first and second of November 15, and the first and second of November 16.
  • The Binance daily-average funding rate fell from about 0.0086 percent per eight hours on November 9 and 10 to about 0.0022 percent on November 15 (calculation from three settlements per day).
  • Binance’s BTCUSDT open interest in dollar terms fell from $7.74 billion on November 5 to $6.11 billion in the latest, incomplete daily record, but open interest measured in bitcoin was roughly flat, at 96,167 BTC on November 5 and 93,524 BTC in the latest record (calculation and file values).
  • Deribit’s BTC volatility index (DVOL) closed at 47.29 on November 14 and reached an intraday high of 54.60 on November 15, from about 40 for most of the preceding two weeks.
  • Binance’s global long-account ratio stayed near 75 percent of accounts (long/short ratio about 3.05) through the weekend, so account-count positioning did not flip even as dollar open interest fell.
  • Published liquidation totals for the weekend, from outlets citing CoinGlass, range from about $1.71 billion to about $2 billion depending on the window and are not reproduced from a primary dataset here.

Funding rates: what the file shows

A perpetual futures contract has no expiry, so exchanges use a periodic payment between long and short holders, the funding rate, to keep the contract price near the spot index. When the rate is positive, longs pay shorts; when it is negative, shorts pay longs. The rates in these files are quoted as a fraction per eight-hour interval, and the tables below express them in percent per eight hours. A rate of 0.01 percent per eight hours is the interest component that Binance’s own premium-index endpoint lists for this contract, and it works out to about 10.95 percent a year if it were paid at every one of the roughly 1,095 eight-hour intervals in a year (calculation). The blog’s earlier guide to funding and liquidations explains the mechanism; this article concerns the readings.

Day (exchange)Binance mean (% per 8h)Binance lowest printOKX mean (% per 8h)OKX lowest print
Nov 90.00860.00590.00580.0021
Nov 100.00860.00590.00650.0046
Nov 110.00750.00660.00800.0046
Nov 120.00620.00350.00760.0061
Nov 130.00510.00130.0025−0.0064
Nov 140.00480.00390.0025−0.0009
Nov 150.0022−0.0002−0.0009−0.0038
Nov 16 (2 of 3 prints)−0.0003−0.0013−0.0009−0.0013

The means and lowest prints are calculations from the settlement files. Binance’s file runs from October 31 through the second settlement of November 16; OKX’s runs from October 27 through the same point. Both venues settle every eight hours, so a day has three prints, except the latest day shown, which has two.

The pattern is a slow fade followed by a break. Through November 12 both venues’ averages stayed between roughly 0.006 and 0.009 percent per eight hours, a level consistent with a market that was paying a small premium to hold long exposure. OKX’s funding rate then dropped to −0.00642 percent at the second settlement of November 13, the lowest print in its file, even though that day’s average of 0.0025 percent was still positive because the other two prints were positive.

By November 15 the OKX daily average was negative, at about −0.0009 percent, and Binance’s was close to zero at 0.0022 percent with one slightly negative print. In annualized terms, the first OKX print of November 16, of −0.0013 percent per eight hours corresponds to about −1.4 percent a year (calculation). That is a trivial cost in absolute terms. Its significance lies in the sign: shorts, not longs, were paying.

Why a tiny negative number matters

The magnitude of these funding rates is small, and no responsible reading would call them extreme. They are nothing like the rates seen in the most crowded phases of a leveraged rally, and nothing like the deeply negative rates seen in a short squeeze. What matters is what the change says about demand for leverage. In the first half of November the funding rate was usually positive on both venues, meaning there was steady demand to hold leveraged longs that exceeded demand to hold shorts. When that premium disappears, either demand for long leverage has dropped, or demand for short exposure has risen, or both, or longs have been forced out. The data cannot separate those explanations; they can only show that the balance changed in the same hours in which the price fell.

One check is available: if funding turned negative because new shorts were opened aggressively, open interest should have risen. It did not. On both venues, dollar open interest fell during the decline, which is the pattern one expects when positions are closed rather than opened. That points toward the weakening of the long premium being driven at least in part by closures, but only open interest measured in coins can finish the argument, and that is where the next section turns.

Open interest: dollars versus bitcoin

Open interest counts outstanding contracts. Measured in dollars, it moves with price even if nobody opens or closes a position, because the same number of bitcoin-denominated contracts is worth less when the price falls. A fall in dollar open interest therefore overstates the amount of position closing when the price has also fallen. Binance’s history file reports both measures for the BTCUSDT perpetual, and comparing them separates the price effect from the position effect.

Day (exchange)Open interest ($ billion)Open interest (BTC)Global long/short account ratioShare of accounts long
Nov 57.7496,1671.3958.2%
Nov 97.1595,5812.8173.7%
Nov 127.1596,6262.6572.6%
Nov 137.23101,1523.1876.1%
Nov 146.7298,3893.2376.4%
Nov 156.2897,9513.0875.5%
Nov 16 (latest, incomplete)6.1193,5243.0575.3%

Between November 5 and the latest record, dollar open interest on this contract fell by about 21.1 percent (calculation), from $7.74 billion to $6.11 billion. Open interest in bitcoin fell by only about 2.7 percent over the same stretch, from 96,167 BTC to 93,524 BTC (calculation). Most of the dollar decline is therefore a price effect: bitcoin fell from about $75,969 (the November 5 candle close) to about $64,400, and the same coin count is worth correspondingly less. The coin-denominated figure even rose to 101,152 BTC on November 13 before falling to 97,950 BTC on November 15 and 93,524 BTC in the latest record, so the large unwinding, if there was one, is concentrated in the last three records, not spread across the month.

OKX’s file reports open interest in dollars across its bitcoin contracts. It shows $3.87 billion on November 12, $3.46 billion on November 13, $3.24 billion on November 14 and $3.03 billion on November 15, a decline of about 21.6 percent over three days (calculation). The file does not report a coin-denominated series, so the same decomposition is not available there. Over the same three days the bitcoin price on Coinbase’s candles fell from $71,370 to $65,315 at the close, about 8.5 percent (calculation), which accounts for a bit more than a third of the OKX dollar decline. The remainder implies fewer contracts outstanding, which is consistent with positions being closed, though the file cannot say whether by choice or by force.

Positioning ratios

Binance’s global long/short account ratio counts the number of accounts holding net long positions against the number holding net short positions, not the size of those positions. The reading of 3.05 on November 16 corresponds to about 75 percent of accounts being net long, essentially unchanged from 3.18 (76 percent) on November 13. OKX’s separate long/short account ratio for bitcoin shows 2.71 in the record stamped November 15, down from 3.20 in the record stamped November 11 but above the roughly 2.1 to 2.2 shown for October 22 to 27 in the file.

Two cautions apply. First, an account ratio treats a one-bitcoin account and a thousand-bitcoin account as equal, and the large accounts matter more for liquidation cascades. Second, a ratio that stays high after a large price drop does not show that no one was liquidated; it shows that a large majority of accounts are still on the long side, perhaps because many small long accounts were not liquidated, perhaps because new longs entered at lower prices. One reading is that the ratio shows an unchanged appetite to buy dips among small accounts. Another is that it shows many accounts still exposed to further declines. The data are silent on which is right.

Volatility: implied against realized

Deribit’s DVOL index is an implied-volatility measure built from bitcoin options prices, quoted as an annualized percentage. The file’s daily records give an open, high, low and close.

Day (exchange)OpenHighLowClose
Nov 1039.2840.6639.1539.45
Nov 1139.4540.2938.5638.92
Nov 1238.9240.6938.4240.07
Nov 1340.0741.5838.8439.87
Nov 1439.8749.3139.7747.29
Nov 1547.2854.6043.3343.95
Nov 1643.9544.0841.7742.57

DVOL spent November 10 to 13 between about 38 and 42 and then jumped on November 14, when its high was 49.31 and its close 47.29. On November 15 the high reached 54.60, the highest value in the file window, and the close fell back to 43.95; the latest, incomplete November 16 record shows 42.57. Implied volatility thus rose by roughly a third at its peak, from about 40 to 54.6 (calculation), and then retraced roughly three-quarters of that jump within a day, a pattern that fits a market repricing a spike and then partly reassuring itself.

Realized volatility, computed from Coinbase daily closes, tells a different story. The standard deviation of the 31 daily log returns from October 12 through November 11, annualized by the square root of 365, is about 35.8 percent (calculation). The same measure over the seven returns from November 9 through November 15 is about 57.1 percent, and over the four returns from November 12 through November 15 it is about 66.5 percent (calculations). A realized figure from four observations is very noisy and should not be given weight on its own.

But the comparison raises a pointed question: with realized volatility running well above the implied peak of 54.6 over the last four sessions, options were not priced for as much movement as the market had just delivered. The implied index looks forward and the realized figure looks back, so a gap is not an error, but it is informative about how surprised the options market was.

A note on noise: the November 7 print

A negative funding print does not by itself signal stress, and the OKX file contains an example. At the second settlement of November 7, a Saturday, OKX’s rate was −0.00038 percent, a slightly negative figure, while the Coinbase candle for that day shows the quietest session in the period, with a range of about 0.9 percent and a volume of 2,438 BTC. The next settlement that day was positive at 0.0065 percent. A single slightly negative print in a quiet market is what small, temporary imbalances between the perpetual price and the index look like. What distinguishes November 13 to 16 is persistence and context: five negative OKX prints in the eight settlements from November 14 through November 16, in a market whose price had just fallen more than 20 percent from its November 4 high.

How to reproduce these figures

Anyone can recheck the numbers in this article, and the steps are worth listing because a few pitfalls cause most errors. First, retrieve the endpoint and read the timestamp field: Binance and OKX report millisecond Unix timestamps, which must be converted to a calendar date, using the venue’s day convention, before tables are built. Second, convert funding rates from a fraction to a percent by multiplying by 100; the raw value 0.0001 is 0.01 percent.

Third, group the settlements by that calendar date and average them, remembering that the latest date is partial. Fourth, for open interest, use the coin-denominated field when you want to know whether positions were opened or closed, and the dollar field when you want to know exposure at current prices. Fifth, for DVOL, remember that the four numbers in each record are open, high, low and close in that order. The record stamping conventions for daily open-interest buckets are not documented in the retrieved files, so a daily record is read here as the bucket carrying that date and not as a point-in-time reading at a specific hour.

The conclusions above are robust to a shift of one bucket in either direction, because the pattern spans several days, but exact day-to-day comparisons are not.

What has been said about liquidations

Liquidations were widely cited as a driver of the weekend move. The retrieved reports give different totals. One article, citing CoinGlass, said roughly $1.96 billion of crypto positions were liquidated on November 15, with bitcoin accounting for nearly $750 million, about $600.5 million of it longs. Another, citing CoinGlass data via Bloomberg, put the total at about $2 billion in the 24 hours to early Sunday, with $774 million in bitcoin and $488 million in ether. A third, on Monday, said about $1.71 billion over 24 hours, of which $1.41 billion was longs, affecting around 260,000 traders. The windows differ, the exact datasets differ, and the figures are not reconcilable from the articles alone.

This article did not retrieve CoinGlass data, and the one public liquidation feed it queried, from OKX, returned only the most recent minutes of forced orders, which cannot reconstruct the weekend. The figures are therefore cited as reported and should be treated as order-of-magnitude indications: about $1.7 billion to $2 billion of forced closures across crypto over roughly a day. For scale, Binance’s BTCUSDT open interest was between about $6 billion and $7.7 billion in dollar terms over the period shown, so a bitcoin liquidation total of $750 million would be roughly a tenth of that one contract’s open interest if all of it occurred there, which is unlikely and cannot be checked from the data (calculation, illustrative only).

Reading the signals together

SignalWhat the data showWhat it is consistent withWhat it does not show
Funding ratePositive through Nov 12; slightly negative at several settlements from Nov 13 on OKX and Nov 15 on BinanceReduced demand for long leverageWhether shorts were added
Open interest ($)Down about 21% on Binance and 22% on OKX over the weeks and days shownPrice effect plus some position closingWhich side closed
Open interest (BTC), BinanceRoughly flat through Nov 15, lower in the latest recordMost of the dollar decline is priceWhether new positions replaced closed ones
Account long ratioAbout 75% long, little changedMany small accounts still longPosition size
Implied volatilitySpike to 54.6, partial retrace to about 42.6Options repriced for larger movesDirection of the next move

The table’s open-interest row compares the Binance decline from $7.74 billion (November 5) to $6.11 billion, a fall of about 21 percent, and the OKX decline from $3.87 billion (November 12) to $3.03 billion (November 15), about 22 percent. The windows differ, as the earlier paragraphs state, so the percentages are not directly comparable and are listed here for orientation only.

A practical reading for traders

For a trader who does not use leverage, the lesson of this data is mostly about conditions. A funding rate near zero after a long positive stretch tells you that the carry cost of holding long exposure through perpetuals has momentarily disappeared; it says nothing about direction. For a trader who does use leverage, the more important lesson is the volatility numbers: a realized volatility more than half again above the month’s average means that stop distances and position sizes that were reasonable three weeks ago may now be too tight or too large. Daily ranges of 6 to 7 percent can pass through a 5 percent stop in minutes. A rule of thumb that is easy to apply is to scale position size inversely to the recent realized range, which can be read from the same public candle data used in this analysis.

Nothing in the data supports a forecast. The files cover only a few weeks, which is too short to say what a move in funding or volatility has historically preceded. What the data allow is a statement of condition: leverage was being reduced, volatility was being repriced, and the premium for holding long exposure had vanished at the end of a violent week.

Sources and research limits

Funding histories are from Binance’s futures funding-rate endpoint and OKX’s public funding-rate history. Binance open interest and long/short ratio are from its open-interest history and global long/short account ratio endpoints. OKX open interest and account ratio are from its contracts open-interest endpoint and long/short account ratio endpoint. DVOL is from Deribit’s volatility-index endpoint. Realized volatility uses Coinbase’s daily candles. Liquidation totals are as reported in BitcoinWorld, FX Leaders and Business Korea, each of which cites CoinGlass or Bloomberg.

Limits: all derivatives figures are venue-specific and cover one contract per venue except OKX’s aggregate open interest. The latest daily records are incomplete. Account ratios count accounts, not size. Funding rates, volatility and open interest do not establish cause. No liquidation total in this article was independently verified.

Disclaimer: This content was generated with the assistance of artificial intelligence (AI) and has been reviewed by our editorial team. It is intended for informational purposes only and should not be construed as financial, investment, or legal advice. Cryptocurrency investments involve significant risk.
KCEX BLOGKCEX BLOG
Previous 12 hours ago
Next 6 days ago

Related Posts

SHARE
TOP