Methodology

In one sentence: for each pair of stocks, Correlworks computes the Pearson correlation of their daily split-adjusted closing prices over one fixed period of about five years (June 22, 2020 – June 18, 2025), then shows each stock's coefficients against all other stocks as a distribution.

What is correlated

Closing price levels, not daily returns. Each pair gets one coefficient computed over the whole period (about 1,255 trading days). The stock-page distributions are not a rolling window.

How to read this: correlation of price levels largely reflects shared long-term trends. Two stocks that both rose over 2020–2025 will show a high positive correlation even if their day-to-day moves are unrelated, and a stock that rose will show a negative correlation with one that fell. It is a different measure from the return correlation used in most portfolio-risk models.

Price data

Daily price bars (open, high, low, close, volume) from Polygon.io, adjusted for stock splits. Company names, sectors (SIC description) and exchange codes also come from Polygon.io reference data.

Universe

About 6,800 US-listed securities, taken from the Nasdaq.com stock list. The list covers stocks listed on NASDAQ, the NYSE and NYSE American, as well as SPACs, units and warrants. The number of pairs differs by stock and is shown on each stock page (up to 6,809).

From coefficients to a distribution

• Coefficients are counted in 0.2-wide bins from −1.0 to +1.0.
• Stock pages display a 0.1-wide grid. Each 0.2 bin is split evenly across its two 0.1 rows, which is why neighbouring rows show nearly identical counts. This will be replaced by true 0.1 bins in the data update.
• Strong positive means ≥ 0.7; strong negative means ≤ −0.7; weak means between −0.3 and 0.3.
• The average correlation shown on stock pages is estimated from bin midpoints.

The NVDA time-series chart

The NVDA chart on the homepage is a separate dataset: for each trading day, the Pearson correlation of daily closing prices between NVDA and each other stock over the previous 30 trading days (a 30-day rolling window), shown as a daily distribution.

Data date and updates

Last data date: June 18, 2025. A data update is in progress; a regular update schedule has not been set yet. Every stock page shows the date of its data.

Limitations

• Correlations are historical. They do not predict future co-movement and say nothing about cause.
• Thinly traded securities (SPACs, warrants, units) can show extreme coefficients.
• For information only — not investment advice.

Questions about the method: correlworks@gmail.com