Skip to content
PerpFinder

Funding prediction accuracy

From September 17, 2026 to October 16, 2026, the dYdX BTC funding rate forecast had the lowest mean absolute error in the ledger: 0.002 bps per hour over 702 matched pairs. Its forecast had the same sign as the settled rate in 87% of the pairs. The table covers 3 venues and 2,957 matched pairs.

Some venues publish a forecast of the next funding rate. This page compares each stored forecast with the funding rate at settlement.

Ledger read: 2026-10-16 06:27 UTC · methodology v2 + v3 + v4

3

venues with forecasts and settlements

2,957

matched forecast and settlement pairs

30 days

maximum read period

Accuracy by venue and asset

VenueAssetSettlementsPairsMissingMAE, bps/hBias, bps/hSame signMedian leadCorrectedPeriod (UTC)
dYdXBTC7027020.0%0.00220.000186.8%7.5 min0Sep 17, 2026 to Oct 16, 2026
dYdXETH7026901.7%0.00860.000094.5%7.5 min0Sep 17, 2026 to Oct 16, 2026
Gate.ioBTC87870.0%0.0024-0.000398.9%7.5 min0Sep 17, 2026 to Oct 16, 2026
Gate.ioETH87861.1%0.0027-0.000297.7%7.5 min0Sep 17, 2026 to Oct 16, 2026
HyperliquidBTC7027020.0%0.00270.000199.7%7.5 min203Sep 17, 2026 to Oct 16, 2026
HyperliquidETH7026901.7%0.00240.000299.1%7.5 min191Sep 17, 2026 to Oct 16, 2026

Errors use hourly-normalized rates. One basis point (bps) equals 0.01%. A lower MAE means the forecast was closer to the settled rate. The forecast is the last one that PerpFinder stored before each settlement, so the lead time is short. These figures describe past forecasts. They do not predict future funding.

How PerpFinder measures forecast accuracy

Each realized settlement is compared with the last stored prediction for the same settlement that PerpFinder observed before the settlement time.

Rates are hourly-normalized fractions. Errors show in basis points per hour (1 bps = 0.01%).

The comparison period is shared for each venue and asset: it starts after the first usable stored prediction and ends at the last settlement that a stored prediction targets.

A settlement in that period without a matching prediction counts as missing.

Hyperliquid predictions from methodology versions 1 to 3 stored the start of the running hour as the next settlement. The ledger adds one hour to those rows and labels each corrected pair.

Kraken reports no settlement time. Binance, OKX and Bybit have no predicted rate in the PerpFinder source mapping. These venues have no pairs.

The publication requirements are: 3 venues, 30 settlements per venue, 7 days of settlements, and no more than 30% missing forecasts. The funding methodology gives the rate types, intervals and version history. There is no synthetic backfill: a missed collection run stays missing.

Data downloads

The funding rates overview shows the current and predicted rates.

Terms used on this page

Funding rate
A funding rate is the periodic payment between long and short positions that keeps a perpetual futures price near its index price. When the rate is positive, longs pay shorts.
Funding interval
The funding interval is the time between two funding settlements on one market, for example 1 hour or 8 hours. A rate always belongs to its interval.

Cite this data

Use the page address and the date of the data. Attribute the data as "Data: PerpFinder" with a link to the page.

PerpFinder. "Funding rate prediction accuracy: BTC and ETH by venue." Ledger read October 16, 2026 at 06:27 UTC, settlements to October 16, 2026. https://perpfinder.com/funding-rates/prediction-accuracy
Accuracy by venue and asset (CSV): https://perpfinder.com/api/data/funding-prediction-accuracy?format=csv
Matched pairs (CSV): https://perpfinder.com/api/data/funding-prediction-accuracy?format=csv&detail=pairs

Download: Accuracy by venue and asset (CSV) · Matched pairs (CSV)