Methodology

Gold Beta estimates how much a stock's daily returns co-move with its benchmark, GLD for gold-focused names or SLV for names whose primary revenue is silver, using ordinary least squares (OLS) regression on aligned daily return series. Benchmark assignment is an explicit, human-maintained list keyed on each company's primary metal, not a keyword match on free-text sector data, and it is stored on the ticker rather than re-inferred on every page load.

For each stock, we regress stock return on benchmark return over the trailing 252 trading sessions of split- and dividend-adjusted closes. The slope is Gold Beta; the coefficient of determination (R-squared) indicates how much of the stock's day-to-day variation is explained by benchmark moves over that window.

Before regressing, we inner-join the stock and benchmark series on trading date (a date missing from either side is dropped, never zero-filled or forward-filled), drop days with zero or missing stock volume, drop days where the stock's return was exactly 0.0 while the benchmark moved (a stale price rather than a real no-move day), and drop single-day returns beyond 400 percent as unadjusted-split or bad-tick data. Large but plausible moves are kept and only logged.

We surface beta together with R-squared so you can judge reliability. Below an R-squared of 0.10, or with too little clean history or too many dropped days, we do not publish a beta at all: the stock shows as not computable with a reason. Between 0.10 and 0.20 the beta publishes but is flagged low confidence. We never apply a bias, floor, multiplier, or winsorization to make the published distribution look a particular way.