Live prices are currently unavailable — the exchange feed could not be reached and no recent cached reading is held.

Bitcoin Data GuideBitcoin Reference

Research · Altcoins

The ICP–BTC correlation record

Correlation between two cryptoassets is usually high, and that is the least interesting thing about it. The number is unstable when the series is short, it says nothing about direction of influence, and it is largely a restatement of the fact that both assets trade in the same market.

Last reviewed 2026-09-21Source: CoinGecko daily price series for ICP and BTC; method described belowNo correlation coefficient is asserted on this page; the method and its instability are described instead.

Dataset, period and method

The dataset is the daily closing price series for ICP and for BTC published by CoinGecko. The period runs from ICP's listing in May 2021 to the review date. The method a correlation study would use is to take the two series of daily returns, compute the Pearson correlation coefficient between them, and report it alongside a rolling window to show how the estimate moves over time.

This page does not report a coefficient. That is a deliberate editorial decision rather than an omission. A correlation estimate computed over a series this short is not stable enough to be worth publishing as a figure, and a number presented without its window, its sample size and its confidence interval invites exactly the over-reading this page exists to prevent. What the page does instead is explain what the estimate would and would not mean.

The distinction that matters throughout is between correlation, association and causation. Correlation is a numerical summary of how two series moved together in a particular window. Association is the weaker claim that they are related in some way. Causation is the claim that one produced the other. A correlation coefficient can support the first, sometimes the second, and never the third on its own.

Why a young series supports weaker conclusions

What a correlation estimate depends on, and how ICP's short history affects each dependency.
What the estimate depends onEffect of ICP's short history
Sample sizeA coefficient computed from a few hundred daily observations has a wide confidence interval; the point estimate can move substantially with a handful of days added or removed.
Window choiceA 30-day, 90-day and 365-day window over the same period can give materially different readings, and there is no principled reason to prefer one.
Regime stabilityThe relationship between two assets can change with market conditions. A series that spans one contraction and one recovery cannot show whether the relationship is stable across regimes.
OutliersA single extreme day — a listing-week move, an exchange event — can dominate a short sample and produce a coefficient that describes that day rather than the period.

Last reviewed 2026-09-21Source: Methodological note; no coefficient is computed or reportedThe factors listed are properties of correlation estimation, not findings about ICP.

The practical consequence is that a reader who sees a single correlation figure for ICP and BTC should ask three questions before doing anything with it: over what window, from how many observations, and how much does it move when the window changes? If the answer to the third is "a lot", the figure is describing the window rather than the relationship.

There is a second reason to be cautious that has nothing to do with sample size. Cryptoassets trade in the same venues, are quoted in the same base currency, and are held by overlapping sets of participants. A common response to a change in market-wide risk appetite will produce a positive correlation between almost any two of them, without either asset influencing the other. That is a shared-factor explanation, and it is the default explanation for a high reading rather than an exotic one.

What a high reading would and would not mean

A high correlation would tell a reader that the two series tended to move in the same direction on the same days during the window measured. That is genuinely useful for one purpose: it is a rough statement about diversification, because two assets that move together provide less offset to each other than two that do not.

It would not tell the reader which asset moved first, whether one responded to the other, or whether both responded to something else. Correlation is symmetric — the coefficient between A and B is the same as between B and A — so it cannot by construction carry information about direction. Any claim that Bitcoin "led" ICP, or that ICP "followed" Bitcoin, requires evidence this statistic does not contain.

It would also not be stable in the way a reader might assume. A coefficient estimated over one window and applied to the next is a forecast, and forecasts of correlation are not more reliable than forecasts of returns. The page on ICP during major BTC moves takes the event-study approach instead, which is a different and narrower question.

Limitations

The comparison period begins at ICP's listing, which is also the period of its most extreme price discovery. Early trading in a newly listed token is not representative of its later behaviour, and including it in a correlation estimate weights the result toward the listing period.

The price series are venue-aggregated. Different exchanges quote different prices, and the aggregation method affects both series. This site's exchange price differences page covers why that matters generally.

Finally, correlation is a statement about returns, not levels. Two assets can be highly correlated in daily returns while their price levels diverge over years, because correlation says nothing about the average return of either. A reader who wants to know how the two records compare in level terms should read the four-asset comparison, which states the differing listing dates as a limitation.

Sources and references

The price series come from CoinGecko. The methodological points about correlation estimation are standard statistical properties rather than findings about either asset.

  • ICP historical price data. CoinGecko, Internet Computer (ICP): the daily series and the listing date that bounds the comparison.
  • Bitcoin historical price data. CoinGecko, Bitcoin (BTC): the daily series used as the reference leg.
  • Bitcoin's own volatility and drawdown record. Bitcoin Data Guide, Risk & Volatility: how Bitcoin's own risk measures are constructed from the daily record.

Related reading