Stochastic Oscillator — momentum within the range
The answer
The Stochastic Oscillator asks 1 question: where did today's close land inside the high-to-low range of the last 14 days? If it closed at the very top of that range, the reading is 100. At the very bottom, it is 0. It measures position inside a recent range, which is why it works when there is a range and produces constant false signals when there is not.
Why this costs you money
Stochastic generates more signals than almost any other common indicator, and this is the source of the damage.
The classic rule is that %K crossing above %D below 20 is a buy and %K crossing below %D above 80 is a sell. On a normal daily chart with default settings, that produces something like 15 to 25 signals a year on a single stock. In a trending market, most of them point against the trend.
Here is what happens to a trader using it in an uptrend. Price rises. Stochastic goes above 80. Price pauses for 2 days. Stochastic crosses down. The trader sells or shorts. Price resumes rising. Stochastic goes back above 80. Two days later it crosses down again. The trader, who is now slightly annoyed, takes the signal again.
In a 6-month uptrend this can happen 8 or 10 times. Each loss is small. The total is not, and the trader has spent 6 months fighting the best-behaved chart on their screen.
There is a second, subtler cost. Stochastic saturates. In a strong uptrend it pins near 100 and stays there. When it is pinned, small price wobbles produce large indicator moves, because the range in the denominator has become small relative to the daily noise. The indicator becomes hypersensitive exactly when price is behaving most consistently.
Stochastic is a range tool. Used in a trend it does not merely fail. It fails frequently, in the same direction, all the way through the move.
How it works
The calculation is easier than most and it is worth knowing exactly, because the formula tells you where it breaks.
Over the last 14 days, find the highest high and the lowest low. Those 2 numbers define the recent range.
Then:
%K = (today's close − lowest low) ÷ (highest high − lowest low) × 100
That is it. If today's close equals the highest high of 14 days, %K is 100. If it equals the lowest low, %K is 0. If it sits exactly halfway, %K is 50.
%D is a 3-day moving average of %K. It is smoother and slower. The crossover between the 2 is what most people trade.
Fast, slow and full
| Version | %K | %D | Behaviour |
|---|---|---|---|
| Fast | Raw %K | 3-period average of %K | Very noisy, rarely used alone |
| Slow (14, 3, 3) | 3-period average of raw %K | 3-period average of that | The common default |
| Full | User sets all 3 numbers | User sets smoothing | Whatever you choose |
The "slow" version simply applies the smoothing one extra time. When a platform shows Stochastic (14, 3, 3) it usually means slow Stochastic, and that is what most charts display by default.`
How it differs from RSI
This matters, because people put both on a chart and believe they are getting 2 opinions.
RSI compares average gain to average loss. Stochastic compares the close to the extremes of the range. Those are genuinely different calculations.
The practical difference is this. A stock that grinds slowly upward every day will show a high RSI, because gains dominate losses. A stock that jumps up once and then sits still near the top of its range will show a high Stochastic and a much less extreme RSI.
But in most real market conditions the 2 move together, because a fast rise produces both dominant gains and a close near the recent high. Treating them as independent confirmation is the error described in the first article of this cluster.
What it tells you, and what it does not
Stochastic tells you the position of the close inside a recent range. That is a precise and honest description and it is the only thing it knows.
Four limits.
It does not know whether the range is real. The 14-day high and low always exist. On a trending chart they are just the most recent 14 days of a move that is going somewhere, and "close near the top of the last 14 days" is what an uptrend looks like every single day.
It saturates. Once %K is above 95, further strength cannot be expressed. The indicator has run out of room, and it will now respond violently to tiny pullbacks. A trader reads that violence as a momentum shift. It is a scaling problem.
It has more signals than information. More crossovers does not mean more opportunities. It means a shorter memory. Every crossover looks the same on the chart and they are not the same.
Its levels are conventions. 80 and 20 came from the tool's popularisers, not from a property of markets. On some instruments a level of 90 and 10 filters out most of the noise. Changing them changes the signal count, which proves they were choices.
The clean way to state the failure: Stochastic assumes the recent high and low are a boundary. In a trend they are not a boundary, they are a waypoint.
The decision rule
Take Stochastic crossover signals only when price is contained, and use the long moving average as the gate.
If the 200-day moving average is flat and price has turned near the same high and the same low more than twice, take crossovers at both extremes.
If the 200-day moving average is sloping, take crossovers only in the direction of the slope. In an uptrend, that means only the oversold crossovers, and only after a pullback. Ignore every overbought crossover completely — those are the ones that will cost you the trend.
Unless %K has been pinned above 90 or below 10 for more than 5 sessions. That is saturation. The next crossover is a scaling artefact and it carries no information about the trend.
Try this now
This takes 5 minutes and it demonstrates that the setting is a choice, not a fact.
- Open a 1 year daily chart of a stock from your own watchlist. Add Stochastic (14, 3, 3).
- Count every %K/%D crossover that happened above 80 or below 20 in the last 6 months. Write the number down.
- Now change the period from 14 to 5. Count again over the same 6 months. Write the number down.
- Change it from 5 to 21. Count again. Write the number down.
- Now put RSI (14) in a panel underneath the Stochastic. Look at the last 3 months. Count how many times one of the 2 entered its extreme zone without the other doing the same within 2 trading days.
What you should see. In steps 2 to 4, the count changes a great deal. A 5-period Stochastic will typically produce 2 or 3 times as many extreme crossovers as a 21-period one on the same chart. Nothing about the stock changed between step 2 and step 4. Only your setting changed. That is the proof that the number of signals an indicator gives you is something you chose.
In step 5 the count of disagreements will be low. Most of the time RSI and Stochastic reach their extremes together, because a fast move produces both a dominant run of gains and a close near the recent high. They are 2 arithmetics applied to 1 phenomenon. Putting both on a chart does not double your information.
Write down the 3 signal counts. The next time somebody shows you a strategy with a specific Stochastic setting, you will know that the setting is doing at least as much work as the market.
Three real cases
1. George Lane and Investment Educators, 1950s (United States) — an attribution that is less certain than it looks George C. Lane is widely credited with the Stochastic Oscillator and he certainly taught and popularised it for decades. The origin is murkier than most sources admit. Lane worked at Investment Educators in Chicago, and the tool appears to have emerged from work by a group there, with Ralph Dystant sometimes named as the actual originator. Lane also said in later interviews that the indicator "does not follow price, it does not follow volume, it follows the speed or momentum of price". The point of including this as a case is that a tool used by millions has an origin story nobody can fully verify, and the confidence with which it is repeated is not matched by the evidence.
2. The NIFTY 50, April 2020 to October 2021 (India) — saturated for months After the March 2020 low the Indian index rose for roughly 18 months with only shallow interruptions. Through long stretches of that advance, daily Stochastic sat above 80. Each brief pause produced a %K/%D cross down through 80. Each of those was, by the textbook rule, a sell signal. Almost none of them were followed by a decline of any size. The indicator was doing its job correctly — the close was near the top of the recent range, over and over — and the textbook interpretation of that fact was wrong for 18 months.
3. The S&P 500, March 2020 (United States) — a range tool in a crash During the fastest part of the COVID decline in March 2020, the S&P 500 fell roughly a third in about a month. The Stochastic denominator — the 14-day high minus the 14-day low — became enormous. When the denominator is huge, %K becomes insensitive: even a large daily bounce barely moves the reading off the floor. So the indicator sat near 0 through the decline and gave almost no usable crossover at the actual low. The formula did not break. It behaved exactly as division by a large number requires, and that behaviour made it useless in the single most important month of the decade.
The question that resolves it
A novice sees %K crossing below %D above 80 and asks: is this a sell signal?
An expert asks: is the 14-day high a boundary or a waypoint?
In a range, the recent high is a boundary. Sellers have appeared there before and the close reaching it means something. In a trend, the recent high is simply where price has got to so far, and the close being near it is the trend functioning normally. The crossover is identical in both cases. What it means is not.
What would make this wrong
The claim is that Stochastic measures position inside a recent range, works when that range is a real boundary, and produces frequent losing signals when it is not.
You would falsify it by running the counting exercise on 10 trending charts and finding that overbought crossovers were followed by declines more often than not. If your data says that, believe your data.
You would falsify the saturation claim by finding a strongly trending chart where Stochastic did not spend extended periods above 80 or below 20. The formula makes that very unlikely, and seeing it is more useful than being told.
The honest limits are 3.
First, this article treats the crossover rule as the standard use, and it is — but it is not the only one. Some traders use Stochastic only for divergence, or only to time entries in a direction already decided by another tool. Those uses are much less exposed to the failure described here.
Second, the counting exercise measures how often, not how much. A rule that works 3 times in 10 can still be profitable.
Third, none of this establishes that Stochastic has any predictive power at all in any condition. It establishes what the number measures and when the textbook reading of it is structurally wrong. Whether the range-bound use makes money is something your own count on your own instruments has to answer.
In India
Indian platforms ship Stochastic at 14, 3, 3 with lines at 80 and 20, and it is widely used by Indian intraday traders on 5-minute and 15-minute charts.
Three Indian conditions matter.
Price bands truncate the range in the denominator. A share outside the derivatives segment carries a daily band of 2%, 5%, 10% or 20%. When a stock locks at its upper band repeatedly, each day's high is the band price. The 14-day high therefore rises in fixed steps, and each close sits exactly at the top of the range. %K reads 100 every day. The indicator is reporting an exchange rule.
Gaps at the open are common and they are not tradable. Indian equities have no regular overnight session. A stock that reacts to overnight news gaps at 9:15. On an intraday Stochastic that gap immediately resets the range, and the first few readings of the day are dominated by a price nobody could have traded at. Intraday Stochastic signals in the first 15 minutes of an Indian session are mostly artefacts of the gap.
Thin volume makes the 14-day extremes unreliable. Below the most liquid few hundred NSE names, a single large order can set the 14-day high or the 14-day low on its own. That one print then defines the denominator for the next 14 sessions. Every Stochastic reading during that period is calibrated against a price 1 order created.
In the United States
The same defaults, and 2 differences that change what the range means.
Extended-hours trading does not enter the daily high and low. US shares trade before 9:30 and after 16:00 Eastern time. Most charting platforms use only regular-hours data for the daily candle. So the highest high and lowest low in the Stochastic denominator exclude any extended-hours move. On a results day, the real high of the day may have been printed at 16:30 and never appear in the calculation at all. The close, meanwhile, is the 16:00 print, from before the announcement.
No daily price bands means the range expands violently. A single 12% session enlarges the 14-day range immediately, and for the next 14 sessions the denominator is large. Large denominator means insensitive %K. So in the US, after any violent day, Stochastic goes quiet for a fortnight regardless of what price is doing. Traders read that quiet as calm. It is a division problem.
The US market also gives the cleanest environment for testing the range-bound claim, because there are thousands of liquid stocks with long histories and free data.
Where they differ, and what that tells you
The difference is in the denominator, and the denominator is the whole indicator.
Indian price bands cap the 14-day high and floor the 14-day low, so the range in the denominator is artificially small and the reading is artificially sensitive. A stock in a banded run will show %K pinned at 100.
American extended-hours trading and uncapped moves inflate the 14-day range, so the denominator is artificially large and the reading is artificially insensitive. A US stock after a big results day will show a Stochastic that barely moves.
What that tells you is which mistake to guard against in each market.
In India, guard against over-reading. A pinned Stochastic on a banded stock is not extraordinary momentum, and a crossover off it is not a reversal.
In the United States, guard against under-reading. A quiet Stochastic in the 2 weeks after a large move is not evidence that the move is over. The indicator is scaled to an event that already happened.
The general principle is worth carrying past this indicator. Any indicator with a range in its denominator will change character after a violent day, in every market, and the change is arithmetic rather than informational.
Carry this
- Stochastic answers 1 question: where did the close land inside the last 14 days' range?
- That question is meaningful when the range is a boundary and meaningless when it is a waypoint. The slope of the long average tells you which.
- Change the period from 14 to 5 and the signal count roughly triples. The number of signals you get is a choice you made.