Does price move the hour before funding? 8 coins, 2020–2026

A Binance perpetual settles funding every 8 hours: when the rate is positive, longs pay shorts; when it is negative, shorts pay longs. A common idea follows from it. When funding is high, some longs close shortly before the settlement rather than pay, so the price should sag in the last hour — and a short in that hour should catch it. With negative funding the same logic says shorts cover, and the price should lift.

This study tests that hour directly on eight coins over almost seven years. The short answer: before costs there was no consistent sag after high funding, only a small and unstable lift after negative funding, and trading costs larger than that lift.

It is a different question from funding as a signal, which asked whether the funding level predicts the next one to seven days and held positions for a week, paying and collecting funding along the way. Here a position lives for one hour and closes before the settlement, so the only thing measured is the price move that the settlement is supposed to cause.

Data and method

• Market: Binance USDT-M perpetuals BTC, ETH, SOL, XRP, DOGE, BNB, ADA and LINK, 1h candles, 2020-01-01 → 2026-10-01 (SOL and DOGE from their first perpetual candles in 2020).
• Settlements: in the platform's funding history these eight coins settle at 00:00, 08:00 and 16:00 UTC; the only exception seen was SOL, which settled every two hours for several days in 2022.
• The trade: the rule is checked on the close of the 06:00, 14:00 and 22:00 UTC candle, fills at the next open (07:00, 15:00, 23:00), and flatBeforeFunding closes it at the last close before the settlement. The position never holds through a settlement, so it neither pays nor receives funding — what is measured is the price move of that one hour. 1x leverage, isolated margin, 95% of the account per trade.
• The funding filter: the platform's FUNDING value is the last settled rate — the one paid eight hours before the settlement the trade is waiting for; the next rate is not known in advance in this data. Thresholds were fixed before the first run: above 0.03% or above 0.05% per 8 hours → short; below 0 → long. A fourth rule is the control: buy every one of these hours, with no filter.
• Costs: first all four rules as unsaved dry runs with zero fees and slippage, to see the raw move. Then saved runs of the two main rules with a 0.02% fee per side and no slippage — the maker rate, which is optimistic for orders filled at the open — and the negative-funding rule on BTC and ETH again at the 0.05% taker fee plus 0.01% slippage.
• How much was tried: 4 rules × 8 coins over the whole period, three of them again for each half of it. Nothing was tuned after seeing a result.

High funding: no selling before the settlement

Each cell is the compounded return of the rule, the t-statistic of its average trade and the number of trades.

With funding above 0.03%, the short made money before costs on 3 of 8 coins — BTC, ETH and SOL — and above 0.05% also on 3 of 8. Its best t-statistic was 0.92; nothing was close to significant in favour of the idea. Against it, XRP and DOGE lost about 60% before costs: on those two coins the price tended to rise into the settlement, not fall.

Split in two halves, the picture contradicts itself. In 2020–2022 the 0.03% short had a negative average trade on 6 of 8 coins; in 2023–2026 a positive one on 6 of 8 — but with t-statistics of 1.62 at most, on 78 to 153 trades per coin, because funding above 0.03% became rare after the 2021 mania. A rule whose sign flips with the period is not an effect to trade.

Negative funding: a small lean that fees remove

The mirror rule looked better before costs: buying the hour before settlement when funding was negative ended ahead on 7 of 8 coins, with t-statistics of 2.06 on ETH and 1.98 on ADA. With 8 coins and 4 rules tested, one or two values near 2 are what chance alone tends to produce, so the halves matter more:

Two things stand out. The pre-settlement hour itself was good for the altcoins in 2020–2022 and turned negative on 7 of 8 coins after — a large part of the old profit was the hour, not the funding. And in 2023–2026 the negative-funding hour stayed positive on all 8 coins while the plain hour was negative on 7. That is the most interesting result here, and still a weak one: no t-statistic above 1.84.

Then the costs. A trade lasts one hour and pays the fee twice, and the average move was smaller than that:

At the maker rate the negative-funding long kept a profit on 3 of 8 coins and the high-funding short lost on all 8. At the taker rate with slippage, the long lost about half the account on both coins tried:

Backtest

Long the hour before funding when funding < 0 (maker fees)

ETHUSDT · 1h · Futures · 2020-01-01 → 2026-10-01

Return

+15.97%

Max drawdown

18.35%

Sharpe

0.25

Win rate

51.2%

Trades

1,019

Open the report →

Backtest

Long the hour before funding when funding < 0 (taker fees)

ETHUSDT · 1h · Futures · 2020-01-01 → 2026-10-01

Return

-46.56%

Max drawdown

53.06%

Sharpe

-0.77

Win rate

44.2%

Trades

1,019

Open the report →

Backtest

Short the hour before funding when funding > 0.03% (maker fees)

BTCUSDT · 1h · Futures · 2020-01-01 → 2026-10-01

Return

-6.82%

Max drawdown

20.68%

Sharpe

-0.09

Win rate

49.8%

Trades

615

Open the report →

The BTC version of the long at taker costs ended at -53.03%.

Conclusion

• High funding did not make the price sag in the hour before settlement on these eight perpetuals in 2020–2026: the short made money before costs on 3 of 8 coins, and its sign flipped between the two halves of the period.
• Negative funding came with a small lift in that hour — positive before costs on 7 of 8 coins, and on all 8 in 2023–2026 — but the lift was smaller than two fees, even at the maker rate.
• Much of what looked like a funding effect in 2020–2022 was the hour itself, which was simply good for altcoins then and not since.

More on funding: funding as a signal over days rather than an hour and when in the day bitcoin moves. How to read the cards: backtest metrics explained.