Oscillators explained — overbought, oversold and momentum
The answer
An oscillator converts recent price movement into a number that moves between fixed limits, usually 0 and 100. It measures speed, not value — how fast price has moved recently compared with its own recent history. A high reading means the move has been fast. It does not mean the move is finished, and in a strong trend it can stay high for months.
Why this costs you money
The sentence "RSI is above 70, the stock is overbought" is followed, in a predictable number of cases, by somebody selling a stock that then rises another 40%. Or worse, by somebody selling it short.
Here is the mechanism, and it is worth understanding rather than memorising.
An oscillator compares recent moves to slightly older moves. When a stock starts a powerful trend, the recent moves are large and the older ones were small, so the oscillator goes to an extreme. That is correct. It is measuring the fact that something new and strong has started.
Now the trend continues. The recent moves stay large. The older moves in the comparison window are also now large, because they were part of the same trend. So the oscillator stays high. It does not fall back. It sits near 80 for weeks while price doubles.
The trader who sold at the first extreme reading has now watched the whole move from outside. Worse, they have watched it while holding a belief — "this is overbought" — that got stronger the longer they were wrong, because the reading kept going higher.
The most expensive fact in this cluster: an extreme oscillator reading in a strong trend is a sign of trend strength, not of exhaustion. RSI below 30 in a downtrend means the downtrend is powerful. It is not an invitation to buy. It is the opposite.
How it works
Every oscillator does the same 3 things.
1. It measures a recent change. Usually over 14 periods, sometimes 20, 5 or
- The period is the comparison window.
2. It scales that change against something. RSI scales average gains against average losses. Stochastic scales today's close against the highest high and lowest low of the window. CCI scales the distance from a mean by the average deviation.
3. It expresses the result as a bounded number. Bounded means it has a floor and a ceiling and cannot go outside them. RSI and Stochastic run 0 to 100. Some oscillators, like CCI and MACD, are unbounded — they can keep rising with no limit.
That distinction matters more than most people realise.
| Bounded | Unbounded | |
|---|---|---|
| Examples | RSI, Stochastic, Williams %R | MACD, CCI, rate of change |
| At an extreme | Cannot go further, so it flattens | Keeps going, so it shows more |
| Fails by | Sitting pinned at 80 for weeks, telling you nothing new | Having no fixed level that means "extreme" |
A bounded oscillator loses information exactly when the move gets biggest, because it runs out of scale. An unbounded one keeps the information but gives you no reference point. Neither solves the other's problem.
What "overbought" actually means
The word is misleading and it is the source of most of the damage.
"Overbought" does not mean too many people have bought. It does not mean the price is above fair value. It has nothing to do with value at all.
Overbought means: price has risen quickly relative to its own recent range. That is the whole definition. A stock can be overbought at 100 and overbought again at 180 and overbought again at 300, because each time the comparison is against the immediately preceding period, not against any anchor.
If somebody had named it "recently fast" instead of "overbought", a great deal of money would have been saved.
What it tells you, and what it does not
An oscillator tells you the speed of the recent move and whether that speed is increasing or decreasing. That is genuinely useful, and it is information your eye cannot easily extract from a candle chart.
It does not tell you 3 things, and each omission has cost somebody money.
It does not tell you direction. A reading of 75 tells you price has risen fast. It says nothing about whether the next move is up or down.
It does not tell you the market condition. The same calculation runs in a trend and in a range and produces the same numbers. The numbers mean opposite things in the 2 conditions, and the oscillator has no way of telling you which one you are in.
It does not have a natural level. 70 and 30 are Wilder's choices from 1978. They were not derived from any statistical property of markets. A stock in a strong uptrend may spend most of its time above 50 and rarely visit 30 at all, which means the classic oversold line will almost never trigger on it.
Here is the honest statement of when an oscillator works:
An oscillator has a chance of being useful when price is contained — when there is a level above which it has repeatedly failed and a level below which it has repeatedly held. In that condition, "fast move up" and "near the top of the range" arrive together and the extreme reading has real meaning.
When price is not contained, the oscillator is measuring speed inside a move that has no ceiling, and the extreme reading means the opposite of what the textbook says.
The decision rule
Establish the market condition before you read any oscillator, and change the meaning of the reading based on the answer.
If price is in a range — repeated failures near the same high, repeated holds near the same low, a flat longer moving average — then an extreme reading is a warning that the move is near the edge of the range, and it can be used.
If price is in a trend — a sloping longer moving average, higher highs and higher lows — then an extreme reading in the direction of the trend is confirmation of strength, and an extreme reading against the trend is the only one worth acting on.
Unless you have no way of establishing the condition. Then the correct use of the oscillator is to watch whether momentum is building or fading over weeks, and to take no signal from any single reading.
That last case is more common than it sounds. Many charts are genuinely ambiguous. The right answer there is not to guess. It is to use the oscillator as a description rather than a trigger.
Try this now
This takes 5 minutes and it will change what the word "overbought" means to you permanently.
- Open 2 daily charts side by side, both set to 1 year. From your own watchlist, pick 1 stock that has trended strongly in one direction, and 1 that has moved sideways between roughly the same 2 levels.
- Add RSI (14) to both.
- On the ranging stock, find every date where RSI went above 70. For each one, look at what price did over the next 10 trading days. Mark each one fell or did not fall.
- On the trending stock, do exactly the same. Count the RSI readings above 70, and mark each one fell or did not fall over the next 10 days.
- Write down 2 fractions. For example: ranging stock 5 out of 7 fell. Trending stock 2 out of 11 fell.
What you should see. On the ranging stock, a decent share of the overbought readings will be followed by a fall, because in a range price genuinely does turn near the top. On the trending stock, most of them will not, and you will find long stretches where RSI stayed above 70 for many consecutive days while price kept rising.
The number that will surprise you is the count itself. A strongly trending stock often produces 3 or 4 times as many overbought readings as a ranging one, because it keeps making fast moves. The tool fires most often precisely where its classic interpretation is wrong.
Keep both fractions written down. When somebody tells you an indicator is overbought, you will now ask which of the 2 charts they are looking at.
Three real cases
1. The S&P 500 in 2008 (United States) — oversold all the way down Through the second half of 2008 the S&P 500 fell steeply over several months. Momentum oscillators reached deeply oversold readings repeatedly during that decline, and the index kept falling, reaching its low in March 2009. A trader buying each oversold reading bought at progressively lower prices for around 6 months. Every one of those readings was arithmetically correct. Every one of them was a description of how fast price was falling, which in a crash is very fast, and none of them was a bottom.
2. The NIFTY 50 from April 2020 to October 2021 (India) — overbought for a year and a half After the March 2020 low, the NIFTY 50 rose for roughly 18 months with only shallow interruptions, roughly doubling from the low. During that period the index spent long stretches with RSI above 70 on the daily chart and on the weekly chart. Anybody who used "overbought" as a sell signal in mid-2020 gave up most of the largest and fastest Indian equity advance in recent memory. The oscillator was not broken. It was reporting a very fast, very persistent move, which is what it is built to report.
3. J. Welles Wilder Jr., 1978 (United States) — where 70 and 30 came from Wilder introduced RSI in New Concepts in Technical Trading Systems in 1978, along with ATR, ADX and Parabolic SAR. He set the default period at 14 and suggested 70 and 30 as the levels of interest. He was working on commodity futures with hand calculation, in a market with different volatility from a modern large-cap share. He did not claim 70 was a sell signal, and later writers including Andrew Cardwell developed the argument that in a strong uptrend RSI tends to hold above 40 and reach 80 or higher, so the levels themselves should shift with the trend. The numbers that a generation of traders treats as facts about markets were a 1978 author's starting suggestion.
The question that resolves it
A novice sees a reading of 78 and asks: is this stock overbought?
An expert asks: is price contained, or is it trending?
The same 78 means "near the top of a range, and the range has held 6 times" on one chart, and "this trend is strong and accelerating" on another. There is no reading of an oscillator that is meaningful before that question is answered, and the oscillator itself cannot answer it.
What would make this wrong
The claim is that oscillator extremes carry information in a range and mislead in a trend, and that "overbought" is a statement about speed rather than value.
You would falsify it by running the counting exercise above on 10 trending charts and finding that overbought readings were followed by falls more often than not. If that is what your own charts show, this article is wrong about your instrument and you should trust your count over this page.
You would also falsify the speed claim by finding an oscillator whose reading depends on the absolute price level rather than on recent change. None of the standard ones do, and you can check this directly: multiply every price in a series by 10 and the RSI does not change.
The honest limits are 3.
First, "trend" and "range" are judgements, and the boundary between them is genuinely unclear much of the time. This article gives you a better question, not a mechanical answer.
Second, counting how often a signal was followed measures frequency, not profitability. A signal that works 3 times out of 10 can be excellent if the 3 wins are large. The counting exercise is a test of the textbook claim, not a full evaluation.
Third, the academic evidence on oscillator rules specifically is thin compared with the evidence on moving averages. Park and Irwin's 2007 survey covered a broad set of technical rules and flagged data-snooping problems throughout the literature. Nobody should take "oscillators work in ranges" as an established empirical fact. It is a structural argument about what the calculation does, and your own count is the evidence that matters.
In India
Indian charting platforms ship RSI at 14 and Stochastic at 14, 3, 3, with the same 70/30 and 80/20 lines as everywhere else. There is nothing Indian about the defaults, and 2 Indian conditions make them behave differently.
Price bands change the shape of the extremes. A share outside the derivatives segment carries a daily price band of 2%, 5%, 10% or 20%. When a stock locks at its upper band day after day, each daily gain is capped at the same number. An oscillator comparing recent gains to older gains sees a series of identical moves, which is unusual input. RSI can pin at a very high value and stay there mechanically, not because momentum is accelerating but because the exchange is issuing the same number every day.
Thin volume produces false extremes. Below the largest few hundred NSE companies, daily volume falls sharply. A single order can produce a 6% day in a stock that normally moves 1%. An oscillator will register that as a major momentum event. It was one buyer. Oscillator signals on illiquid Indian shares should be treated as unverified until you have looked at the volume on the day that produced them.
Indian markets run 9:15 to 15:30 with a pre-open call auction from 9:00 to 9:08. All the momentum an indicator measures happened inside a single continuous session, so the daily reading is a complete picture of the day. That is an advantage over the US, and it is worth knowing you have it.
In the United States
The same tools, the same defaults, and 2 structural differences that change what an extreme reading represents.
Extended-hours trading hides part of the move. US shares trade before 9:30 and after 16:00 Eastern time, and most charting platforms exclude that activity from the standard daily candle. A stock that reports results after the close can move 15% before the next session opens. The daily close-to-close change that RSI uses will record that 15% as a single day's move, so the oscillator will show a violent momentum event with no visible path. In India that path would have been visible inside the session.
Volatility is not capped. With no daily price bands, a US stock's largest daily moves enter the oscillator at full size. Oscillator extremes in the US are therefore reached faster and unwound faster, and thresholds copied from an American source will fire more often on a US chart than the same thresholds do in India.
The US also gives you something Indian markets do not: an enormous set of liquid stocks with long price histories, which makes the counting exercise in this article far easier to run at scale. If you want to test an oscillator claim on 100 charts rather than 2, US data is where you can do it.
Where they differ, and what that tells you
The difference is in how the extremes are produced.
In India, the biggest moves are cut off by price bands and spread across several sessions. In the United States, the biggest moves land whole, and part of them lands outside the trading session where nobody can see the path.
What that tells you is how to read a sudden extreme reading in each market.
An Indian stock showing an extreme oscillator reading after several circuit-locked days is not showing accelerating momentum. It is showing an exchange rule. Check whether the stock hit its band before you interpret the indicator.
A US stock showing an extreme reading the morning after results is not showing a market that traded through the move. It is showing a gap. The oscillator has recorded a change that no continuous auction produced, and the usual argument for why speed matters — that buyers kept paying up — does not apply.
The wider lesson is that an oscillator is only as honest as the price series it is fed. Both markets distort that series, in opposite ways, and neither distortion is visible in the indicator itself.
Carry this
- An oscillator measures speed, not value. "Overbought" means "recently fast", nothing more.
- In a range, an extreme reading is near the edge of the range and can be used. In a trend, it is a measure of strength and using it as a reversal signal is how people lose money in the best markets.
- Establish the condition first. A reading with no context is not information.