Stochastic RSI explained: the settings, %K, %D and how to use it
Stochastic RSI is an indicator of an indicator: it takes RSI and asks where today's reading sits inside the range RSI itself has covered lately. This guide is what that number means, what its four lengths change, and the comparisons you can build a SixOne8 condition from.
What Stochastic RSI measures
It is built in two steps. First, RSI is worked out as usual: recent gains against recent losses, on a 0 to 100 scale. Then the Stochastic calculation is applied to that RSI line rather than to price — find the highest and lowest RSI over the last N bars and place the latest RSI between them. A reading of 100 means RSI is at the top of its own recent range; 0 means it is at the bottom; 50 means it is in the middle.
The result is far more restless than RSI. RSI can drift between 45 and 60 for weeks; Stochastic RSI stretches that small range to fill the whole scale, so it visits 0 and 100 often and can sit at either for several bars in a row. Readings above 80 are conventionally called "overbought" and below 20 "oversold" — names for regions of the scale, not statements about what price does next. When every RSI value in the window is identical, the app reads it as 50, because the reading is then at the top and the bottom at once and neither extreme would be true.
Stochastic RSI settings
Four lengths, all counted in bars of the timeframe the condition reads.
- RSI length. The period of the RSI underneath. Fourteen by default.
- Stochastic length. How many RSI values the high-low range covers. Also fourteen by default. The two are routinely set apart, which is why the app names both when it writes the condition out.
- %K smoothing. An average applied to the raw reading to make %K. Three by default, the figure most published setups assume. One means no smoothing.
- %D smoothing. The average of %K that makes the signal line. Three by default. It is shown only for the two signal-line comparisons, the only ones that read it, and it must be at least two — at one, %D is %K and the two lines could never cross.
How to use Stochastic RSI in a strategy
These are the comparisons SixOne8 offers for Stochastic RSI, in the builder's own words. Every one reads %K except the last.
- Greater than / less than a level. True for every bar %K is on that side of the level. A state, so an alert fires when the strategy first becomes satisfied and not again while it stays there. The level defaults to 50.
- Crosses above / crosses below a level. True only on the bar where %K moves through the level — the usual choice for the condition that should trigger the strategy.
- Rising / falling. %K is higher, or lower, than it was the bar before.
- Crosses above / below its signal line. %K moves through %D. The two sit close together on this indicator, so this fires on most turns.
Stochastic RSI %K(14, 14, 3) on 1h · crosses above · 20
Read it as a sentence: on the 1-hour chart, %K of a fourteen-bar stochastic over a fourteen-period RSI, smoothed over three, has just moved up through 20. That is what the scanner evaluates at each hourly close, and what the alert will say if it is the condition that completed the strategy.
Stochastic RSI, RSI or the Stochastic?
All three are momentum readings on a 0 to 100 scale, and the difference is how easily each reaches its extremes. RSI is the steadiest: it reaches 70 or 30 only on a sustained run of closes one way. The Stochastic reaches its ends whenever the close lands at the edge of the recent price range. Stochastic RSI reaches them whenever RSI is at the edge of its own range, however small that range is — so a level crossing on it happens many times more often than the same crossing on RSI. That makes it quick to react and quick to repeat. A condition on it usually needs a slower partner that decides whether the moment counts: price above a long moving average, or a trend reading from the ADX.
Does Stochastic RSI work on your market?
Run an entry check. It will tell you whether, on the market you care about, the moments this condition fired moved any differently from a random moment over the history tested. A condition that fires as often as this one gives the check plenty of moments to count, which makes its answer worth having. That answer is a fact about the past on that market, and it is the only kind of fact a test here produces.