What do stochastics measure that RSI doesn't?
The stochastic oscillator, popularised by George Lane in the mid-twentieth century, asks a question RSI never asks: where does today's close sit inside the recent trading range?
The formula
%K = 100 × (Close − Low₁₄) ÷ (High₁₄ − Low₁₄)
where High₁₄ and Low₁₄ are the highest high and lowest low of the last 14 sessions. Then:
%D = a 3-period simple moving average of %K
The "slow" version smooths once more: slow %K is a 3-period average of fast %K, and slow %D averages that again. All the smoothing does is trade responsiveness for calm — the same dial as every other indicator.
A worked example
Over the last 14 sessions the high was 120 and the low was 100. Today closes at 116:
%K = 100 × (116 − 100) ÷ (120 − 100) = 100 × 16 ÷ 20 = 80
Today's close sits 80% of the way up the recent range. If tomorrow closes at 104, %K falls to 20 — a fifth of the way up, with the range unchanged.
How that differs from RSI
RSI weighs the size and direction of daily changes; %K only cares where the close landed between two extremes and is blind to the path taken to get there. Two instruments with identical %K can have quite different RSI values.
In practice, though, both are functions of the same fourteen closes and they usually move together closely — the redundancy lesson applies with full force. Running both is defensible only if you can state two different questions.
The limitation nobody teaches
Look at the denominator. When High₁₄ − Low₁₄ is small — a quiet, narrow fortnight — a trivial move sends %K from 10 to 90. The oscillator is loudest exactly when the least is happening. In a wide, volatile range the same absolute move barely registers.
So a %K of 95 in a dead-flat month and a %K of 95 after a violent rally are not comparable readings, even though the platform prints the same number. Checking the range width before reading the oscillator takes three seconds and prevents a whole category of mistake.
Here is the flat month. The last bar moves fifty-five cents and carries the reading almost the full width of the scale:
Divergence, described honestly
Price makes a higher high while %K makes a lower one. It is a real, observable pattern and traders watch it. It also resolves against the observer routinely, especially in strong trends where an oscillator pinned near its ceiling has nowhere to go but sideways. Treat it as a reason to look harder at price and volume — not as a conclusion, and never as an instruction.
In the data
A stochastic takes three settings, not one: the window for the high and low (usually 14), the smoothing that turns fast %K into slow %K (3), and the average that makes %D (3). Here is Apple's published stochastic at those standard 14/3/3 settings, three sessions to 25 September 2026:
The %K a platform shows you is usually the slow one, the three-session average of the formula above, not the formula itself. A single "14" typed into a settings box does not say which of the three windows it set, so a stochastic quoted without all three is not reproducible.
Try it now
- Compute %K on both schematics above. One window is twenty points wide and one is seventy cents wide; the readings come out on the same 0–100 scale and mean nothing like the same thing. Write down the denominator beside each answer — that habit is the whole lesson.
- Now verify it against the published figure. Take Apple on 25 September 2026, the table below: over the fourteen sessions ending that day the highest high was 345.34 and the lowest low 309.90, both findable on the month of candles with a Level. Reproduce the bar by hand, then find your figure in the table in the section above. It will not match at first, for the reason named there: the published line is the slow %K, so average your figure with the same calculation for 23 and 24 September, whose closes were 337.02 and 335.92 inside the same high and low.
- Write one sentence contrasting what %K and RSI each told you about that bar. If the two sentences are identical, you have one indicator with two names on your screen.