Why do old prices look "wrong" on charts?

A curious learner checks a fact: "Apple traded around $500 before its 2020 split" (Course 1). But one chart shows 2020 prices near $125, and another shows them near $121. Is one of them lying?

Neither. They're adjusted, and adjusted differently — and understanding this is the difference between reading data and being fooled by it. (We build market data for a living; consider this unit the house specialty.)

Three versions of one day

Take Apple's last session before the split, 28 August 2020. That single day has three honest numbers:

• Raw (unadjusted) — what the ticker actually printed: a close of $499.23. Apple really did trade near $500 that day.
• Split-adjusted — pre-split history divided by 4, so split day stops looking like a fake 75% "crash": $124.81. Most price charts default to this one.
• Total-return adjusted — the same series with every later dividend folded back in as well: about $121 as at mid-2026, roughly 3% lower again, because Apple has paid dividends since. This is what an "adjusted close" in a data table usually means — and it is the only one of the three that will read differently when you look it up, for the reason two paragraphs below.

Adjusting exists because a split moves the price tag, not your wealth (Course 1's pizza), and an unadjusted long-term chart is littered with cliffs that never happened to anyone's money. But "adjusted" is not one thing. Always note which of the three you're holding.

Dividends: the subtler adjustment

Ex-dividend drops (Unit 3) create the same illusion in miniature — price dips, but holders received cash. Dividend adjustment folds those payouts back in, which is why a "total return" line can rise over decades where the raw price line looks flat.

How big does that get? Take TLT, a bond ETF that has never split, so every bit of the gap is dividends: its 3 January 2006 close was $91.78 raw and, when this was written, $47.57 adjusted. TLT distributes monthly, so that second figure is rewritten twelve times a year — it had already moved to $47.38 a few weeks later. The raw one has not moved since 2006 and never will. The raw line sits nearly twice as high as the adjusted one, with no split anywhere in the story.

One consequence nobody expects: each new dividend re-applies the adjustment, so the adjusted history of a dividend payer changes every quarter. "Why did my old prices move since last week?" usually has that answer.

Which version does the data hand you?

In EODHD's daily price data the open, high, low and close come back raw and unadjusted — that's why 28 August 2020 reads 499.23 — while a separate adjusted close carries both the splits and the dividends. Same row, two conventions, both correct. Elsewhere you'll most often meet the split-adjusted middle version, which is exactly how three sources can disagree with none of them being wrong.

The last session before that split and the first one after it — the phantom cliff is in the raw close and nowhere else:

Why this will matter beyond charts

Any calculation over history — returns, comparisons, and especially the backtests waiting in the Quant & Coding path — silently assumes adjusted data. Run a backtest on raw prices and every split in your sample becomes a phantom crash your "strategy" brilliantly avoids or catastrophically hits. Garbage in, confident garbage out.

In the data

The adjustment reaches beyond prices. The splits and dividends behind it — the corporate actions — are recorded too, and the old dividends are restated along with the prices. Here is Apple's first dividend of the modern era, August 2012:

The first amount is that dividend restated onto today's share count, after the 2014 and 2020 splits; the second is the cheque holders actually banked in 2012. A dividend history pulled today is not what anyone received at the time. Volume is restated too, which surprises people who expect a share count to be sacred: in the split table above, the volume of 28 August 2020 is the printed count scaled up by the split that followed. That makes long volume averages safe and makes one particular multiplication dangerous — that volume times the raw close claims $94bn of turnover in a session, while the same volume times the adjusted close gives about $23bn. Mixing an adjusted number with a raw one puts the split back into an answer you thought you had cleaned.

Try it now

1. Find a split first: Apple's are all listed below, each with its date and its ratio of new shares to old. Then go back to the two sessions either side of the 2020 one, in the table above, and read the two prices side by side: the close is the raw tape, the adjusted close carries the splits and dividends.

2. Spot the phantom cliff in the raw close — the crash that no investor ever experienced — and measure it: the last close before the split divided by the first close after. Set it beside the ratio in the split table, four new shares for one old. Your division lands close to that ratio without matching it, and the difference is not an error: it is the real move the stock made that session, which the adjustment keeps and the split does not.
3. Prove that to yourself on the adjusted close: before against after gives you that move on its own, a percentage of a few points at most. Multiply it by the split ratio and you land back on your raw division, to the decimal.
4. Now find the number that catches people out. Multiply the volume by the raw close on the day before the split, then on the day after: the turnover appears to collapse by roughly the split factor overnight. Do it again with the adjusted close and the two days sit side by side. Volume was adjusted along with the prices, so pairing it with the raw close is what manufactured the cliff — this time in your own arithmetic rather than in the chart.
5. Say the rule once: "long-horizon math needs adjusted prices; the day's actual tape is raw." You now out-read most chart users on the internet.