Identify correlation clusters among the chosen assets
Answers how closely the chosen assets move together. It reads 90 days of daily candles for each asset, converts them to daily returns, and computes the pairwise Pearson correlation across every pair. The output is a full correlation matrix, the pairs above a fixed threshold, and a high, moderate or low clustering label read from the median pair correlation by a published rule. It shows which of the chosen assets have moved closely together over the window.
Pearson correlation on daily returns, and what it can and cannot support.
Method: 90 daily candles per asset are converted to returns - note that 90 prices give 89 one-period returns, and potentially fewer once assets are aligned on common dates - then for each pair r = cov(X,Y) / (sd(X)*sd(Y)), bounded in [-1, 1]. Whether returns are simple or logarithmic, and how series are aligned by timestamp and cut-off, are part of the method. Correlation is computed on RETURNS rather than prices: correlating price levels can produce spurious correlation when the series are trending or otherwise non-stationary, because such series appear related whether or not they are.
Five properties to carry away.
It measures LINEAR association only; strong non-linear dependence can show a correlation near zero.
It is an ESTIMATE from a finite sample. With 89 observations the sampling error is real, and the value moves as the window rolls even when nothing about the assets has changed.
It is SENSITIVE TO OUTLIERS. A single extreme day - common in crypto - can move a pairwise correlation materially, which is why rank-based alternatives such as Spearman's are sometimes preferred.
It is UNDEFINED when either series has zero variance, as with an exactly flat or heavily rounded stablecoin series, or a stale price feed.
It is not causation, and across many pairs some high values will arise by chance alone - the multiple-comparisons problem.
Correlations among risky assets within a group commonly rise during stress, which weakens measured diversification when it is most wanted. That is a well-supported stylised fact for many groups rather than a universal law: not every pair, and not every episode.
No standalone diversification score is computed, because the matrix has no asset weights. Pairwise correlations alone omit position weights and individual asset volatilities and say nothing about tail dependence, so they are not a measure of portfolio risk in the portfolio-theory sense.
Published reports are readable without an account. The newest is Ether (ETH) and XRP extend a month-long slide as Bitcoin falls below $75,000: trend, momentum and news (Nov 9, 2026). Create an account to run research on the stocks and crypto you choose.