Compare returns, not price levels

Correlation measures how two series vary together. For instruments with different price scales, returns are generally more meaningful than raw closes. The interval, return convention, lookback, and timestamp alignment all affect the result.

A coefficient near +1 indicates strong positive co-movement in the selected sample, near -1 indicates inverse co-movement, and near zero indicates little linear relationship in that sample. It does not establish causality.

Beta answers a different question

Beta describes sensitivity to a selected benchmark and is commonly expressed as covariance with the benchmark divided by benchmark variance. Correlation is scale-free co-movement; beta also reflects relative movement magnitude. Similar correlation can coexist with different beta.

The benchmark matters. Comparing a stock with a broad equity index asks a different question than comparing it with a sector ETF, another asset, or a futures contract.

Coverage and regime matter

Different trading hours, missing bars, holidays, and stale observations can create false precision. Matching timestamps and reporting usable pair count makes thin calculations visible. Splitting recent and longer windows can show instability, but shorter samples are noisier.

Correlation, beta, and standardized spread distance are descriptive sample statistics. Relationships can change abruptly, particularly during stress, and none is an instruction to hedge or trade a pair.