Most pages about stock indicators argue over which one to put on top of a chart. RSI or MACD, Bollinger Bands or moving averages. Almost none of them ask a more basic question: what happened to the prices before the indicator ever saw them?
Here's the short version. Every number an indicator prints on a single-stock chart comes from a price series that your data vendor has already rewritten. A split divides every historical price. A dividend adjustment shifts the whole earlier series down. A trading halt deletes bars that the indicator's period count still expects. So the RSI you read today for a date two years back is not the RSI anyone saw on that date, and a signal you can spot in the history may never have been visible to a person trading it live.
This post walks through the three events, shows a worked example with small numbers you can follow by hand, and ends with a two-minute check to run on your own chart. It's not a ranking of stock technical indicators. If you want that, our guide to indicators for US stock trading covers the picks. This is about the layer underneath them.
Why the series matters more than the indicator
An indicator is a function. Feed it closes, highs, lows and volumes, and it returns a number. The function is deterministic, so the only way two people get different readings on the same bar is by feeding it different inputs. When one trader's chart shows RSI at 78 and another's shows 13 for the same stock on the same day, the indicator isn't broken. The series is different.
The standard bar format is OHLCV data: open, high, low, close, volume. Each bar is a snapshot of what traded in that period. For a stock that never splits, never pays a dividend and never halts, the snapshot never needs editing. Real stocks do all three, and that's where the trouble starts.
There's a second consequence people miss. Backtests run on the rewritten series. If you test a stock strategy over five years, and the stock split in year three, the vendor's choice (raw or adjusted) decides what your entry signals looked like in years one and two. You can get a clean equity curve out of a series that couldn't have produced those trades in real time.
Method: how we built the example
We built one short, invented price series and ran the same 14-period RSI over it under three treatments: raw, split-adjusted and dividend-adjusted. The series is illustrative. It isn't any real security's history, and no claim here is about how a real stock behaved. We chose 15 closes so that a 14-period RSI has exactly 14 changes to average, which keeps the arithmetic short enough to do on paper. The methodology is plain Wilder-style RSI with a simple average for the first window: average gain divided by average loss, converted with RSI = 100 − 100 / (1 + RS).
Our own series of 15 closes, in dollars, written as the continuous economic path:
100, 102, 101, 103, 105, 104, 106, 108, 107, 109, 111, 110, 112, 114, 113.
The 14 changes are +2, −1, +2, +2, −1, +2, +2, −1, +2, +2, −1, +2, +2, −1. Nine gains total 18 and five losses total 5, so average gain is 1.286 and average loss is 0.357. RS is 3.6, and RSI = 100 − 100 / 4.6 = 78.3. That's our baseline. Every other figure below is the same series with one event applied.
What a stock split does to your values
A split is a pure divide. According to the SEC's Investor.gov glossary, a split changes the number of shares outstanding and the price per share in inverse proportion, without changing the company's total value. On a 4-for-1 split, one $100 share becomes four shares at $25.
The vendor then has a choice about history. A split-adjusted feed divides every pre-split open, high, low and close by four, and multiplies every pre-split volume by four, so the chart reads continuously. A raw feed leaves the old prices alone, and the chart shows a drop from $109 to $27.75 overnight. That's a 75% one-bar fall in a company that lost no value.
Run our RSI over both. Take the 15-close series above and suppose the split happens after the tenth close. The adjusted feed shows the closes unchanged in shape: every pre-split value is divided by four, so the first ten closes read 25.00, 25.50, 25.25, 25.75, 26.25, 26.00, 26.50, 27.00, 26.75, 27.25, and the post-split closes continue at 27.75, 27.50, 28.00, 28.50, 28.25. RSI over that series is 78.3.
The raw feed keeps the first ten closes at their old values (100 to 109) and then drops to 27.75 and onward. The 75% fall lands inside the window as one enormous loss, and the average loss swamps everything else. RSI on the raw series is 13.3.
Same company, same days, same formula. One chart says the stock is overbought at 78. The other says it's deeply oversold at 13. If a rule says "buy when RSI drops below 30," the raw feed fires it on split day, and every volatility and momentum rule on that chart fires with it.
Notice the other half of this. If the whole 14-bar window sits before the split, or entirely after it, adjusted and raw RSI agree. RSI depends on ratios of gains to losses, so dividing everything by four changes nothing. The distortion only appears for windows that straddle the split date. That's why people can use unadjusted data for years and never notice, then get a wildly wrong reading in the fortnight after an event.
The volume side
Volume moves the other way. A 4-for-1 split multiplies the share count by four, so a vendor adjusting for it multiplies old volume by four. A raw feed doesn't. Across a split boundary, on-balance volume, accumulation/distribution and any volume-weighted average mix two unit scales. Pre-split bars count shares that were four times as expensive, and a single post-split day can look like a volume explosion just because each share is smaller.
Because OBV is built by adding the day's volume on up days and subtracting it on down days, a raw feed gives the first post-split day four times the weight of a normal day without any change in actual interest. The indicator isn't wrong about its inputs. It's right about numbers that don't mean the same thing.
What a dividend adjustment does to your values
A dividend adjustment is a subtraction, not a divide. When a stock goes ex-dividend, its price typically opens lower by about the amount of the payout, because buyers on or after that date don't receive it. According to Investor.gov's definition of the ex-dividend date, you have to own the shares before that date to get the dividend.
Adjusted-close feeds smooth that gap away by shifting the entire pre-ex-date series down by the cash amount (some vendors do it proportionally, and that's worth checking in their notes). The chart then looks continuous, and the gap disappears from the history.
Use the same series again. Say a $3 dividend goes ex after the tenth close. In a raw feed, the post-ex closes sit $3 lower than the continuous path, so the move from close 10 to close 11 is a $1 drop instead of the $2 rise. In an adjusted feed, the first ten closes are each cut by $3 and the series is a uniform shift of the original, which has an RSI of 78.3. The raw feed's RSI is 72.7.
That's a 5.6-point gap from one $3 dividend on a $110 stock, about 2.7%. Two things follow. First, a subtraction doesn't preserve ratios the way a divide does, so anything built on percentage change (rate-of-change, percent-B, Bollinger band width) shifts for every adjusted date, even on windows far from the event. Second, an old level moves. A support line at $100 on the raw chart is $97 on the adjusted chart.
That second point matters for any stop or limit. A trader's stop is filled at the price that actually traded that day. If a backtest places a stop at an adjusted level, it's testing a price that never printed. State which series a stop was set against, because "the 200-day low" is two different numbers depending on the feed.
Which indicators move and which don't
Roughly, they sort into three groups:
- Scale-invariant, so split-safe inside a window: RSI, stochastic oscillator, percent-B. Divide everything and the reading is unchanged, as long as the window doesn't cross the event.
- Level-dependent, so dividend-sensitive: moving-average crossovers against a fixed price, support and resistance levels, any hard-coded price threshold.
- Unit-dependent, so split-sensitive on volume: OBV, accumulation/distribution, VWAP across the boundary.
Our RSI glossary entry covers the formula if you want to rebuild the example yourself with a different series.
What a trading halt does to your values
A halt is a different kind of break. Nothing is rewritten. Bars simply don't exist. Per FINRA's explanation of trading halts, a stock can be halted for news pending, for regulatory reasons, or when volatility triggers a pause, and during the halt no trades print.
An indicator with a 14-period setting counts 14 bars, not 14 days of calendar time. After a three-day halt, the window that used to span 14 trading days now spans 17 days of elapsed time, with the halt gap sitting inside it as a single oversized step. The moving average is now averaging prices from a longer stretch than it did last month. The reading is a different measurement wearing the same label.
The usual symptom is a one-bar jump. The stock reopens at a price far from the last print, and that one change enters every window that includes it. For ATR-style volatility measures, that single gap can inflate the reading for the next 14 bars. For a moving-average crossover, it can trigger a signal on a move nobody could trade, since there was no market during the move itself.
Intraday charts show the same effect at a smaller scale: a five-minute halt leaves a hole of several bars, and the indicator closes the hole without comment. This is part of why tick data and bar data aren't interchangeable for testing. A bar series can hide a gap that tick data would show as missing time.
Delisting and ticker reuse
Symbols get reassigned. When a company is delisted or renamed, its ticker can later belong to a different business, and a long history pulled by ticker alone is sometimes two companies' prices stitched together. We won't put a number on how often that happens, because we haven't measured it.
What you can do is spot it. Scan the series for a single-bar gap with no corresponding corporate action in that company's filings. A move of 40% overnight, no split, no dividend, no halt on record: that's the signature. Indicators don't flag it. They treat it as a real move and carry it through every window for the next 14, 20 or 200 bars.
The two-minute check you can run on your own chart
You don't need to take any of this on trust. Here's how to settle it for your own symbol.
- Pick a stock that had a known split in the last three years.
- On your chart, read the close for the trading day before the split date.
- Find the same figure in the company's own filing or investor-relations notice, which gives the pre-split closing price and the split ratio.
- Compare. If the chart's figure matches the filing, the chart is raw. If it's the filing's figure divided by the split ratio, it's adjusted.
The same logic works for dividends. Take the close before an ex-dividend date and compare it with what the exchange reported that day. A gap equal to the cash amount, or close to it, means your chart is adjusted.
Split ratios and ex-dividend dates should come from the issuer's filings or the exchange's corporate-action notices, not from another chart. A chart can only tell you what the vendor decided, and the vendor is the thing you're trying to check.
Rules, not advice
None of this tells you what to trade. It's a set of data-hygiene rules, and Quantzee is analytical software, not advice. Here's the rule set we'd write into any stock test before looking at a single result:
- Decide adjusted or raw before the backtest starts, and write the choice down next to the results.
- Never mix the two inside one test. A strategy that reads signals from an adjusted series but sizes stops from a raw one is comparing two different currencies.
- Re-run any result that spans a split or ex-dividend date under both settings. If the conclusion changes, the result was about the data, not the strategy.
- Flag every gap larger than a threshold you set in advance, and match each against a filing before trusting it.
- Check for halts in the symbol's history, and note which indicator windows contain one.
For the mechanics of running a test without fooling yourself, our walkthrough on how to backtest a TradingView indicator covers the setup, and the backtesting glossary entry defines the terms. Paper trade first, always. A strategy that only works on one feed setting hasn't earned real capital.
What this means for the lists you've already read
When someone shows you a ranked table of stock technical indicators, ask what series it ran on. A win-rate table built on an adjusted index is a fine piece of work, and it answers a question about the adjusted index. It doesn't tell you what the same rule would have done on a single stock with a raw feed across a split. Both kinds of test have a place. They aren't the same test.
The same goes when you search for the best stock technical indicators and land on a list with eight chart images. The images were taken on some feed, on some date, with some adjustment setting, and the page rarely says which. We're not criticising those pages for it. It's just the missing line that decides whether the numbers can be reproduced.
Ours is a premium toolkit, and part of what it does is keep the data conventions explicit so a result can be reproduced. But the check above works with any charting platform, free or paid. Run it once on a symbol you trade. If you find your chart is adjusted and your stop levels were set on raw prints, you've learned something useful in two minutes.
Frequently Asked Questions
Reference reading: Investor.gov on stock splits, Investor.gov on ex-dividend dates and FINRA on trading halts.
