Methodology
A stock price looks like the simplest data point on the site. It isn't: a stock split or a reused ticker symbol can each quietly turn a correct chart into a wrong one.
When a stock splits, the reported closing price drops by the split factor overnight. Nvidia's 10-for-1 split: $1,208.88 on June 7, 2024, then $121.79 on June 10. Read as a raw series, that's a 90% crash that never happened.
We store the price the market actually reported and apply split ratios at the moment the chart is drawn, so it renders as a smooth line. Storing adjusted prices as facts would be a trap: the next split retroactively changes every adjusted price before it, forcing a rewrite of years of history. Computing fresh each time, one new row per split is enough, permanently.
Some providers pre-adjust the prices they hand back. Adjusting an already-adjusted price again would corrupt the whole history before the split. So incoming data is checked for the split-day cliff before it's trusted as raw — if the expected cliff isn't there, the feed already adjusted it, and we don't ingest it.
Symbols get renamed and recycled — a symbol dropped by a delisted company is eventually handed to someone else entirely. A price history keyed on symbol text alone would, at some point in the tape, silently become a different company. We key every price to a specific, reviewed exchange listing, and check each symbol against the exchange's current records before trusting it.
When a symbol can't be verified against a live, unambiguous listing, we show no chart rather than guess — a chart built on a guessed match looks exactly as confident as a correct one, and you'd have no way to tell the difference. The same caution applies after publication: a price that would create an unexplained split-sized cliff is held for review instead of published, and a published price later found to be bad is removed from the live chart.
See a split-adjusted chart in practice on Nvidia's security page — or check which stocks tracked managers agree on right now.