RSI vs Stochastic RSI
The relative strength index compares recent gains with recent losses over a chosen length. Stochastic RSI takes that output and applies the stochastic calculation to it, so it is a transformation of a transformation and reaches its extremes far more often than the original does.
One of these is made out of the other. That fact settles most of the questions people ask about the pair, and it is rarely the first thing said about them.
What each one is
The relative strength index compares recent gains with recent losses over a chosen length, on a fixed nought-to-one-hundred scale. The relative strength index covers it.
Stochastic RSI applies the stochastic calculation to that output. Instead of asking where the close sits in the recent price range, it asks where the strength index sits in its own recent range. Stochastic RSI covers the construction.
So the second is a function of the first. Nothing in it came from anywhere the first one had not already looked.
Where they differ
How often each reaches an extreme. The base indicator visits its extremes occasionally. The stacked version is at one a great deal of the time, because it is measuring position within a narrow range.
How much lag is carried. Each smoothing adds some. Stacking two means the second one is reacting to a series that was already late.
How many parameters there are. One length on the base tool; a length plus the stochastic’s lookback and smoothings on the stacked one — several more opportunities to fit the past.
What a signal means. On the base indicator an extreme is unusual. On the stacked one it is the normal state, which means acting on every one produces a very large number of trades.
Where they agree
They are the same measurement. The stacked version cannot disagree with its own input in any way that reflects the market rather than the second transformation.
Both are bounded and both pin in a trend. A sustained move holds either at an extreme, and reading that as a reversal signal is the standard error with both.
Both fail in a range. On this site’s shared series direction runs average 2.01 bars with a longest of 11, and short runs cross any threshold repeatedly on either.
And both cost a round trip per signal acted on — about 2% of the median bar range of 0.493 here — which the version producing many more signals pays many more times.
Which one to use
Run the relative strength index. One transformation of price, one parameter, and an extreme reading that is rare enough to be worth noticing when it happens.
Run stochastic RSI when you specifically want more signals and have counted them. It turns earlier, including on moves that go nowhere, which is the trade every amplification makes.
Run stochastic RSI when your method needs frequent triggers and something else is deciding direction. As a timing tool inside a larger framework it is doing a job it can actually do.
And never run both. That is one indicator on the chart twice, with the copy shouting, and the two agree by construction rather than by observation.
Why stacking transformations costs you
Because each one adds lag and none adds data. Two smoothings applied in sequence produce a reading that is later than either alone while still describing the same closes.
And because amplification makes noise look like signal. A small movement in the base indicator becomes a full swing in the stacked one, which reads as a decisive event and is not.
What to check before using the stacked version
Count the extremes over a month. If the reading is at a limit most days, the threshold is not marking anything unusual and every rule built on it will fire constantly.
Decide which tool sets direction. The stacked version is a timing tool at best, and using it for both direction and timing means one amplified series is making every decision.
Fix all the parameters at once. There are several, they interact, and adjusting them one at a time after losing runs is the fastest route to a rule fitted entirely to the past.
And compare the trade counts, not the charts. On this site’s shared series direction runs average 2.01 bars, so the amplified version will produce far more triggers than the market produces turns — and that gap is the entire cost of the swap.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two
directly — the pair is constructed from two subjects the corpus covers separately. Separately,
stochastic RSI appears in 24 titles at a median of 28,266 across 21 channels, and the relative strength
index in 820 at a median of 5,021. The counts come from site/corpus_count.py.
24 videos on the stacked version at a median of 28,266 against 820 on the base at 5,021. A thirtieth of the coverage and more than five times the audience per video — one of the largest interest-per-upload gaps measured on this site, on a subject that is a derivative of the one it dwarfs.
The answer to the question on that chart is that the base reading is the honest one. The stacked version is at a limit because a small move in a narrow range fills its scale — so the dramatic reading is describing something the other one correctly calls unremarkable.
When it fails
The failure is trading every extreme on the stacked version, and the trade count makes it expensive before the direction is even wrong. The reading sits at a limit most of the time, so a rule of entering on extremes fires constantly. Many of those entries are inside ranges, some are against trends, and each pays a round trip. The win rate can even look acceptable while the account shrinks, because the costs are charged on a number of trades the method was never sized for.
The second failure is running both together. One is made from the other.
A third is reading an extreme as a reversal. It means recent bars were one-sided.
A fourth is tuning the parameters after losses. They interact and it is fitting.
A fifth is expecting it to lead price. It lags more than its own input.
And a sixth is treating the amplification as sensitivity. It amplifies noise identically.
Related
The relative strength index covers the base calculation. Stochastic RSI covers the stacked version. And stochastic covers the transformation being applied on top.
If you have both of these on a chart, you have one indicator twice — and the second copy has been amplified until it spends most of its life at an extreme. That is not more information. It is the same information shouted.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.