About

What this is not

Not a charting terminal, not a trading tool, and not a forecast. Nothing here projects a figure forward, including the tempting version of that — taking a historical annualised rate and extending it. Every number on the site describes a period that has already ended.

Method

Raw daily closing prices are stored alongside splits and dividends as separate events, rather than a pre-adjusted price column. Adjusted prices are retroactively mutable: every dividend changes the adjustment factor for the whole history preceding it. Keeping the raw series means the files only ever gain days at the end.

Splits change the share count on their ex-date, which is the first session trading at the post-split price. Position value stays continuous across a split by construction. Dividend entitlement follows the ex-date but the cash arrives on the payment date, and reinvestment happens then — buying on the ex-date would be spending money that has not arrived.

Sessions come from the dates that actually appear in the data, not from a holiday rule. No rule produces the real calendar: the market was shut for four days in September 2001, two days for Hurricane Sandy in 2012, and single days for state funerals.

A single purchase is annualised as a compound annual growth rate. Anything with more than one contribution is annualised as a money-weighted rate, because each contribution was invested for a different length of time and one rate cannot describe them all.

Known limits

Dividend payment dates are estimated at 21 days after the ex-date where the data source does not publish them; affected tickers are flagged in the published metadata. Measured against a real payment cadence, the effect on a long total return is under a tenth of a percent, and it errs downward.

Spinoffs arrive from the data source encoded as very large cash dividends. On the affected tickers a total-return figure treats the distributed shares as cash reinvested in the parent, which is not what a holder received. Those distributions are visible as outliers in the published data.

Sessions on which a stock did not trade are carried forward from the previous close and counted in the disclosure line for each tool. A carried-forward session shows a flat return, so it can never be selected as one of the best days.

Data

The published corpus covers … stocks and ETFs over … trading sessions, from … to …. That start date is not a preference: the calendar is built from the sessions one continuously-traded reference actually traded on, and that reference begins there.

Tips

Past performance does not indicate future results. Nothing on this site is investment advice, and no figure here is a projection. Historical data may contain errors; verify anything you intend to rely on.