Adjusted Close
A closing price restated for dividends and splits, so the chart reflects the total return an investor actually earned rather than the raw last trade.
Formula
adjusted close = raw close × (product of every adjustment factor after that day) dividend factor = (close the day before the ex-date − dividend) / close the day before the ex-date split factor = 1 / split ratio (e.g. 1/2 for a 2-for-1 split)
Open a long stock chart and you'll sometimes see the price fall off a cliff on a single day. Often nothing bad happened to the company — it just paid a dividend or split its shares. When a $100 stock pays a $2 dividend, its raw closing price drops to around $98 on the ex-dividend date, but that isn't a loss: you received the $2 in cash. Adjusted close bakes those dividend and split effects back into the historical prices so the chart shows the total return instead of an illusion.
The mechanics are simple. Every time a dividend or split occurs, every price before it gets multiplied by an adjustment factor. A 2-for-1 split divides the old prices by two, so a move from $100 to $50 no longer looks like the stock halved. A dividend nudges the earlier prices down by the payout, so the ex-dividend dip doesn't register as a drop in performance. The most recent adjusted close equals the raw close; the two only drift apart as you go further back.
That gap matters when you're scoring predictions honestly. On raw close alone, an "up" call that happens to land on an ex-dividend date can be marked wrong for no real reason, while a "down" call collects a free win from the very same dividend. If you want to check a prediction against real data, always read the return off the adjusted close.
Example
A $100 stock paying a $2 dividend shows a raw ex-date close near $98, about −2%, but adding the $2 you received leaves the total return roughly flat. After a 2-for-1 split the raw price goes $100 → $50, yet on an adjusted basis the whole history is halved too, so the 50% "crash" disappears.
How LDBD uses it
LDBD scores every prediction on returns computed from the adjusted close (adj_close), never the raw close. That way an "up" call on an ex-dividend day isn't punished for the dividend-driven dip, and a "down" call doesn't get a free win from it. Prices come from yahoo-finance2 for stocks and ETFs and CoinGecko for crypto, and everything downstream — the directional log return g and the annualized rate — is built on this adjusted return_pct.
FAQ
How is adjusted close different from the regular closing price?
The regular close is the last price that actually traded at the bell that day. Adjusted close takes that number and retroactively folds in any later dividends and splits, so the historical chart reflects true total return instead of dividend- and split-driven jumps. Recent dates are nearly identical; the two drift apart the further back you look.
Why use adjusted close to calculate returns?
With raw close, the natural price drop on a dividend day looks like a loss. If you don't strip out dividends and splits when comparing long-run returns or prediction accuracy, high-dividend stocks end up looking worse than they really were. A return measured on adjusted close is close to a total return with dividends reinvested.