Build a correlation matrix from aligned returns
Each row must represent the same interval for every asset. Use one frequency and return convention, such as monthly total returns in a common currency. Prices are not returns. This tool does not convert prices, adjust dividends or align separate calendars automatically.
CSV files require a Date, Period or Time first column. JSON may omit that field, in which case rows receive ordinal labels. Names must be distinct, dates or period labels must be unique, and every cell must contain a finite decimal number. Missing cells reject the import rather than shifting later values into the wrong periods.
Inspect the calculation
Pearson correlation is Σ[(x − mean x)(y − mean y)] divided by the square root of Σ(x − mean x)² × Σ(y − mean y)². The NIST correlation reference describes this coefficient. A constant series has zero variance, so its coefficient is undefined, including its self-correlation.
The heatmap displays each value directly; color is supplementary. Rankings sort defined pairs by absolute correlation. Strength labels are display heuristics, not statistical significance tests. The summary uses arithmetic mean and sample standard deviation (n − 1) per supplied period, with no annualization.
Interpret a sample without making a forecast
Positive values describe a tendency to move together linearly in this sample; negative values describe the opposite tendency. A value near zero can coexist with nonlinear dependence. The coefficient says nothing by itself about the size of price moves, expected return or a suitable hedge ratio.
With only two nonconstant observations, correlation is mechanically +1 or −1. More observations do not guarantee stability: outliers, changing market conditions and the chosen window matter. Review the underlying data and compare windows before using correlation as one input to a broader risk assessment.
Keep a copy of your inputs
File parsing and calculation run in the browser; this tool does not send the uploaded returns to an analysis API. It attempts to save data in local storage, which can be unavailable or cleared by browser settings. Export Returns preserves the input table, and Export Matrix saves coefficients with N/A for undefined cells. CSV labels are quoted and formula-like labels are prefixed for spreadsheet safety.
The tool is available without sign-up or a tool fee. It does not select assets, optimize weights or execute trades. Historical negative correlation does not establish a reliable future hedge.