Skip to the guide
CryptoKeyhole
Log inSign up

Leverage gauge backtest

What followed every hourly reading of the leverage gauge: how far the price moved over the next 24 and 72 hours after each band, coin by coin, against all hours.

After an extreme reading

1.20×

median 24-hour move against the same coin's usual move

Coins graded

30

Coins with 72 hours of prices after their readings

Hours graded

1,174,308

Since 2020

Graded on

Size

Moves in either direction; not direction

Hours graded to seven days ago; recalculated every hour. Last update 2 November 2026, 16:16 UTC

Every coin together

Across 29 coins, the median move in the 24 hours after an extreme reading was 1.20 times the same coin's usual 24-hour move, and 0.92 times after a low one. That is the size of the move only: the gauge is not graded on direction.

Every coin together: moves after each band as a multiple of the same coin's usual move
BandHoursCoinsMedian move, next 24hNext 72hRealised volatility, next 24hBig 24h move
Low243,661300.92×0.94×0.92×17%
Elevated514,613300.97×0.97×0.97×19%
High301,860301.05×1.05×1.06×21%
Extreme114,174291.20×1.14×1.21×26%

Bitcoin and Ether

Bitcoin

For Bitcoin, after extreme readings the median move over the next 24 hours was 1.7%, 1.28 times the 1.3% of all hours; after high readings the median move over the next 24 hours was 1.4%, 1.02 times the 1.3% of all hours. Moves are counted in either direction: the gauge says how stretched positions are, not which way the price breaks.

Bitcoin: what followed each band, 2 Sep 2020 to 23 Oct 2026
BandHoursSpellsMedian move, next 24hMedian move, next 72hRealised volatility, next 24hFollowed by a big 24h move
Low11,5051,8471.3%2.5%40.7%18%
Elevated25,7183,8401.3%2.4%41.7%19%
High12,1812,5361.4%2.7%45.8%22%
Extreme3,6746141.7%3.8%53.3%28%
All hours53,078–1.3%2.5%42.9%20%

Ethereum

For Ethereum, after extreme readings the median move over the next 24 hours was 2.3%, 1.25 times the 1.8% of all hours; after high readings the median move over the next 24 hours was 1.8%, 1.00 times the 1.8% of all hours. Moves are counted in either direction: the gauge says how stretched positions are, not which way the price breaks.

Ethereum: what followed each band, 2 Sep 2020 to 23 Oct 2026
BandHoursSpellsMedian move, next 24hMedian move, next 72hRealised volatility, next 24hFollowed by a big 24h move
Low8,2671,4371.8%3.5%55.7%20%
Elevated22,6663,6961.8%3.2%54.2%18%
High15,8033,3381.8%3.3%56.6%19%
Extreme6,3421,1682.3%4.4%66.3%27%
All hours53,078–1.8%3.5%56.4%20%

Coin by coin

Median 24-hour move after each band, as a multiple of the coin's usual move
CoinHoursLowHighExtremeUsual 24h move
Bitcoin53,0780.97×1.02×1.28×1.3%
Ethereum53,0780.96×1.00×1.25×1.8%
Chainlink51,1800.96×1.04×1.15×2.6%
Litecoin51,1800.86×1.05×1.28×2.0%
Bitcoin Cash50,1060.86×1.10×1.27×2.0%
Cardano47,9820.89×1.07×1.26×2.3%
Polkadot47,9820.93×1.03×1.17×2.4%
Aave46,7990.92×1.01×1.17×2.7%
XRP46,7660.84×1.15×1.19×1.8%
Dogecoin46,3180.97×1.10×1.13×2.2%
BNB45,6380.88×1.09×1.20×1.4%
Stellar44,3120.90×1.05×1.13×1.9%
TRON44,3060.91×1.09×1.14×1.1%
Avalanche43,9350.90×1.14×1.34×2.6%
Near43,3120.96×1.02×1.21×3.0%
Solana43,2400.89×1.06×1.25×2.5%
Shiba Inu42,9900.93×1.04×1.20×2.1%
Hedera Hashgraph42,9760.94×1.07×1.27×2.4%
Uniswap42,1361.02×0.98×1.22×2.5%
The Graph41,9830.93×1.01×1.08×2.8%
Lido DAO36,5360.86×1.08×1.15×3.0%
Aptos34,3600.98×1.01×1.05×2.6%
Arbitrum30,7131.01×0.97×1.03×2.6%
Sui29,7290.91×1.07×1.18×2.8%
Pepe29,7291.00×1.00×1.07×3.1%
Quant28,6160.91×1.06×1.26×2.0%
Ethena21,7020.92×1.05×1.03×3.8%
POL (ex-MATIC)17,9450.94×1.01×1.24×2.2%
Ondo13,4420.90×1.08×1.25×2.7%
Toncoin2,2390.95×1.05×0.90×1.5%

How the record is measured

For every sealed hour with a gauge reading and the coin's hourly closes (its Binance USDT spot market), the absolute price change from the hour's close to 24 and 72 hours later, and the realised volatility over those hours (the root of the summed squared hourly log returns, scaled to a year), grouped by the band the hour read. Each band is compared with all hours of the same coin. A big move is one in the coin's largest fifth of 24-hour moves over the hours graded. Hours overlap, so the hours are not independent; spells (runs of consecutive hours in a band) count separate episodes. Every coin together: for each band, the median over coins (with at least 100 hours in the band) of the band's median move divided by the coin's own all-hours median. Point in time: each reading uses only data up to its hour. The last seven days are left out. Moves are counted in either direction: the gauge says how stretched positions are, not which way the price breaks. The record and the current readings are free to reuse under CC BY 4.0, crediting CryptoKeyhole with a link.

Questions

Does a high leverage reading predict a crash?

Not a crash in particular. High and extreme readings were followed by larger moves than usual, in either direction; the record does not grade direction, and our research into the same positioning data found it does not tell up from down.

How big were the moves after extreme readings?

Across 29 coins, the median move in the 24 hours after an extreme reading was 1.20 times the same coin's usual 24-hour move, and 0.92 times after a low one. That is the size of the move only: the gauge is not graded on direction.

Why measure size and not direction?

Leverage is fuel. Stretched positions make forced buying or selling more likely when the price moves, which shows in the size of the move, whichever way it goes.

Is the record point in time?

Yes: each hour's reading uses only data up to that hour, and its outcome is the price after it. The last seven days are left out of the public record.

Crypto leverage gaugeAll free tools

Readings and outcomes are calculated by CryptoKeyhole from the public market data we collect, using only data that existed at each hour. The record describes what followed in the past; it is not a forecast.