Beta Lab.

beta is an estimate, not an observation, so the band is the point

Current password Password Confirm password

Specification in force

Asset

Benchmark

How old data counts

Frequency

Window older days are ignored entirely Half-life when a day counts half as much How fast beta may drift left = smoother · right = twitchier Beta type

Returns

Period

from to

Blume adjustment Vasicek shrinkage Teaching mode (known true beta)

Anchor Window how far back each pair is measured

Frequency

Sort by

Add a ticker you don't hold Enter · joins the table and the anchor list

Score weights correlation % beta % alpha %

Reading the chart. With a fixed rolling window a single outlier enters the sample, sits there for the full window, then drops out, producing a step on a date when nothing happened. Turn on teaching mode, set frequency to daily, watch the step, then switch the estimator to EWMA or Kalman and watch it disappear. That artifact is what this tool exists to make visible.

Caveats. Simple returns from split- and dividend-adjusted closes. Confidence intervals assume independent, constant-variance residuals, which equity returns are not, so read them as a floor on the true uncertainty rather than the whole of it. Kalman bands are filter state variance, not frequentist intervals. Risk-free is the 3-month Treasury constant maturity yield. Not investment advice.

Price data by Tiingo.

Change password · Sign out