Bitcoin Risk Index

What it measures

How rough the next one to two weeks are likely to be for a long position: the chance of a sharp drop and of wide daily ranges. It is not a price forecast, and it says nothing about where the market goes after that.

Four readings, each as a percentile

Every reading is turned into a percentile of the same coin's trailing 365 days. That is why the score moves: a fixed threshold like "open interest above 50,000 coins" stops meaning anything once the market has grown past it, and it grades ETH on a BTC scale. A percentile cannot go stale.

Then ranked again

The weighted mix of the four is ranked against the trailing year of mixes, and that rank is the score. Averaging percentiles piles up in the middle; ranking the average puts the score back on a scale where 84 means riskier than 84% of the past year's days. Levels are cut on that scale: low below 25, moderate to 55, high to 80, and extreme for the top fifth of days.

What is deliberately not in it

Funding rates. We tested them: at a one-to-two-week horizon, both extreme positive and extreme negative funding were followed by shallower drawdowns than average, for both coins. A crowded trade is a risk on a longer clock than this index keeps. The same test is why an overbought RSI scores low: after the top decile of RSI readings, the next two weeks were calmer than usual, and after the bottom decile they were the roughest of all.

Does it work?

We replayed the index from December 2024 to September 2026, 651 scored days per coin, with every day ranked only against the days before it, and looked at what came next. Candles from July 2024 fill the first rankings; scoring starts once a coin has enough history behind it.

Rank correlation between the score and the next week's realised volatility: 0.44 for BTC, 0.31 for ETH. On the days the first version of this index was live, it managed 0.13 and 0.04 and rated 98% of BTC's days "moderate"; the current index, scored on those same days, manages 0.52 and 0.32.

Data: Binance spot BTC/USDT and ETH/USDT candles, and USDT-margined perpetual open interest. Binance keeps only 30 days of open-interest history, so the replay used Bybit's as a stand-in for that factor; the live score uses Binance, filled in from our own daily snapshots. The factors, which tail of each counts as risky, and the weights were settled on this same period, by judgement rather than a search over settings, so these are in-sample numbers: a description of the past, not a promise about the future. Every figure on this page comes from one replay run, dated 30 September 2026.

The froth reading, kept separate

Risk says how rough the next two weeks are likely to be. It says nothing about how far a move has already run, and a screaming rally scores low on it. Froth is that other reading: how crowded and stretched the market is, as a percentile of the coin's past year, from four inputs. Its levels are cool, warm, hot and frothy, on the same cuts.

What the replay said about froth

We tested froth on the monthly clock: 605 scored days per coin, about 20 independent months. It did not forecast reversals. The frothiest fifth of ETH days returned +15.5% over the next 30 days with a −5% drawdown; the coolest quarter returned −8% with a −16% drawdown. BTC showed no relationship in either direction. One bull cycle cannot test the long unwind that froth is usually blamed for, so read this gauge as how much of the move is already in the price, not as a timing signal. That is also why it is not blended into the risk score.

One follow-up, stated before looking: perhaps the unwind is the dangerous moment, not the peak. Counting days on which froth fell below 60 within two weeks of a reading of 80 or more, BTC had 46, with an average 30-day drawdown of −11.8% against −7.8% on other days; ETH's 29 such days had shallower drawdowns than the rest. A handful of episodes, split between the coins. We keep measuring rather than promising.