End-of-day

How it works

One row per instrument per trading day: open, high, low, close, volume, often an adjusted close and sometimes the split and distribution factors behind it. It is the oldest shape in the market and the one nearly every card in this catalogue can produce, which is exactly why the unqualified phrase hides so much.

Start with the close, because three different numbers answer to that name.

The auction price. Nasdaq's rule text is the clearest published statement of the mechanism. Under Equity 4, Rule 4754(b), the Nasdaq Closing Cross begins at 4:00:00 p.m. EST, all orders executed in it are executed at the cross price and reported anonymously, and that cross price "will be the Nasdaq Official Closing Price for stocks that participate in the Nasdaq Closing Cross". This is a distinct event with its own print, not the last ordinary trade of the day.

The last print. A security that does not close in an auction still has a close: the most recent trade of the session. That is a legitimate number and a different one, and on an instrument that last traded at 3:12 p.m. it is very different.

A computed midpoint. The same Nasdaq rule shows that an official closing price need not be a traded price at all. Where a Nasdaq-listed exchange-traded product has no closing cross, or its cross trade is less than one round lot, the rule uses the time-weighted average midpoint of the NBBO as the Nasdaq Official Closing Price.

When it is actually available

The bell is not the deadline people assume. Under the same rule, fifteen minutes after the close Nasdaq disseminates via the network processor a trade message setting the Nasdaq Official Closing Price as the official Consolidated Last Sale Price in each Nasdaq-listed security in which one round lot or more executed in the cross and the closing price differs from the consolidated last sale price. So the tape's idea of the close is corrected after the fact, as a matter of published rule.

Everything downstream inherits that, and then adds its own delay. Some vendors publish an end-of-day file within minutes of the session; others build it overnight and post it the following morning; some post a provisional file and replace it. This is the first question to ask of an end-of-day product and the one least often answered on a pricing page.

Whose day it is matters as much. A file is stamped with the venue's trading day, not yours, and two venues in two timezones do not cover the same hours under the same date. Joining one row per symbol per date across markets silently lines up windows that do not overlap.

Whether the file is adjusted, and whether it gets restated

An end-of-day series is adjusted, unadjusted, or both, and the vendors that give you both are a minority. The reason it matters is that adjustment is retroactive. Nasdaq publishes a Daily List of dividends, distributions and splits; every entry that lands changes the adjusted history before it, so an adjusted close is a statement about the series as it stood on the day you downloaded it.

That makes an adjusted end-of-day file a moving record rather than an archive, which is the point-in-time problem in its cheapest form. What each event does to the series is corporate action; why two vendors disagree about the same adjusted close is why adjusted close differs between providers.

Why it matters here

End-of-day is also a licence, not only a shape. It is normally sold as delayed or historical data, which exchanges price as their own product: Nasdaq's US equities price list carries delayed data products at 50 dollars a month per firm with no per-user fee at all, against per-user fees for the real-time equivalent. That is the whole reason an end-of-day API costs a few dollars and a real-time one costs a few hundred, and it is unpacked in what real-time market data costs.

So for any card here that advertises end-of-day data, three questions settle what you are buying. Which close — auction, last print, or the vendor's own snapshot? Adjusted, unadjusted, or both, with the factors? And how long after the session does the file exist? EODHD, Tiingo and Marketstack are built around this shape; Norgate Data and FirstRate Data sell the adjusted-and-unadjusted pairing as the product.

The cheapest mistake is treating the close as a fact rather than a convention. It is the convention the whole catalogue leans on, which is why it is worth knowing which one you have: how to download historical prices in Python is where most readers meet it first.