Q: What does 'cluster order' do?

Cluster order reorders the matrix so that highly correlated assets sit next to each other using a nearest-neighbor walk. This makes asset groupings (e.g. L1 blockchains, memecoins, stablecoin-adjacents) visually obvious.

Question 1

Why does PythFeeds compute correlation on log returns instead of raw prices?

Accepted Answer

Raw crypto prices all trend together over long windows, which inflates correlations toward +1. Log returns (log(p_t / p_{t-1})) remove the trend and give the actual day-to-day co-movement between assets — the statistically correct input for Pearson and Spearman.

Question 2

What is the difference between Pearson and Spearman correlation?

Accepted Answer

Pearson measures linear correlation between two series and is sensitive to outliers. Spearman ranks the observations first, then computes Pearson on the ranks — this makes it robust to outliers and captures monotonic (non-linear) relationships. Use Spearman when markets have fat-tail days.

Question 3

What does the diversification score mean?

Accepted Answer

Diversification score = 1 − mean(|r|) across all off-diagonal pairs. A score of 100% means zero average correlation (perfect diversification); a score near 0% means everything moves together (no diversification).

Question 4

What is β (beta) vs BTC?

Accepted Answer

Beta is the OLS regression slope of an asset's returns on BTC's returns. β = 1.5 means the asset tends to move 1.5× as much as Bitcoin. β < 0 means inverse movement.

Question 5

What does 'cluster order' do?

Accepted Answer

Cluster order reorders the matrix so that highly correlated assets sit next to each other using a nearest-neighbor walk. This makes asset groupings (e.g. L1 blockchains, memecoins, stablecoin-adjacents) visually obvious.

Trending Now

Watchlist

Correlation Matrix

Correlation Matrix