Indicators, and what the readings mean

Stochastics: where price closed in its range, and why it pins in a trend

The stochastic oscillator answers one question and answers it exactly: where in its recent range did price close? Everything people believe about it beyond that comes from somewhere other than the formula.

About 10 minutes to read. Free, no account.

The number is a position inside a range

Take the last N bars, 14 by convention. Find the highest high and the lowest low in that window. Then ask where the latest close sits between them, as a percentage. That is %K, sometimes called raw or fast %K.

%K  =  100 times (close - lowest low) / (highest high - lowest low)

over the last 14 bars
  highest high   120
  lowest low     100
  latest close   115

%K  =  100 times (115 - 100) / (120 - 100)
    =  100 times 15 / 20
    =  75

So %K 75 means the close landed three quarters of the way up the last fourteen bars’ range. Not that price rose 75% of anything, not that the move is three quarters finished. A position in a range, nothing else.

At 100 the close equalled the window’s highest high. At 0 it equalled the lowest low. At 50 it sat exactly in the middle of the range, which can happen in a market that has gone nowhere and in one that has travelled violently in both directions.

%K and %D, fast and slow

%K computed bar by bar is jumpy, because the window’s high or low can change from one bar to the next and move the denominator. So it is smoothed, usually twice, and the names are worth getting straight.

raw %K     the range position above
slow %K    a 3-period average of raw %K
%D         a 3-period average of whichever %K is plotted

the usual "14, 3, 3" means  N = 14, then two 3-period smoothings

the last three %K values   75, 81, 90
sum                        246
%D  =  246 / 3             =  82

%D is therefore an average of a range position - one step further from the market than %K and one step later. The "fast" and "slow" stochastic differ only in how many of these smoothings are applied. Nothing new is measured by either.

Why it saturates in a trend

The conventional lines are 80 and 20, with the same unhelpful labels RSI gets. The structural reason they do not work as triggers is the one the RSI lesson sets out: a bounded indicator must saturate when an unbounded price keeps going. %K cannot exceed 100. Price can rise forever. In a sustained advance the two facts have to be reconciled somewhere, and the reconciliation is %K sitting near 100 for as long as the advance lasts.

Stochastics has a second mechanism on top of RSI’s, and it is worse. The window’s highest high is in the denominator, and in an uptrend the highest high is the current bar. So the ceiling is lifted by the same bar that lifts the close. Watch:

bar 1   window high 120   window low 100   close 120
        %K  =  100 times (120 - 100) / (120 - 100)  =  100

bar 2   window high 130   window low 100   close 130
        %K  =  100 times (130 - 100) / (130 - 100)  =  100

close rose from 120 to 130, which is 8.3%
%K stayed at 100

The reading is pinned and the trend is doing exactly what trends do. A reading of 100 does not mean the move is finished; it means every bar in the recent window closed at the top of the window, which is the definition of the move being strong. The mirror case pins at 0 through a decline while every "oversold" reading marks another step down.

The denominator makes it blind to size

Because the range appears on the bottom of the fraction, it cancels out. %K is scale-free: it has no idea whether the range it measured was enormous or trivial.

over the last 14 bars
  highest high   100.50
  lowest low     100.00
  latest close   100.50

%K  =  100 times 0.50 / 0.50  =  100

the entire 14-bar range was 0.50 on a price of 100  =  0.5%

A maximum reading produced by a market that did nothing for fourteen bars. This is the same blindness RSI has and the reason an oscillator belongs next to a measure of size rather than instead of one - see ATR. In a dead-flat market the oscillator will swing from one extreme to the other on noise, repeatedly, and look tremendously informative while doing it.

The crossover, and what it is worth where

%K crossing %D means the current range position has moved away from its own recent average. In a range-bound market that happens constantly in both directions and carries no information. In a trend it can usefully mark a pullback having finished pulling back - and it fires late, because %D is an average of three bars of a smoothed quantity.

• A cross near an extreme is the weakest version. At 95 the reading has almost no room above it, so a cross down is close to arithmetically inevitable and happens repeatedly inside strong advances.
• A cross in the middle of the band says little. Half the range is the default place for a close to be.
• The indicator cannot tell you which market you are in. Range or trend is a structural question, and nothing in a range position answers it.

Shortening N makes everything above happen more often; lengthening it makes extremes rarer. Both are choices about sensitivity, neither adds information, and picking the pair that would have worked on the chart in front of you is curve fitting however good the result looks.

What stochastics cannot do

• It cannot say a move is over. High readings are what strength looks like, for two separate structural reasons - the bound, and the window’s own high in the denominator.
• It cannot say how big anything was. The range cancels, so a half-point range and a fifty-point range produce identical numbers.
• It cannot give you a level. The high and low defining the window slide every bar.
• It cannot be read without knowing the regime. The same cross is reasonable context in a trend and pure noise in a range, and the reading is identical in both.

The honest use is as a question: this close is near the top of its recent range - is that because the range is being broken, or because the range is holding? The oscillator cannot answer that. Structure, the size of the bars and the behaviour of pullbacks can. Nothing in this lesson is an instruction to trade; see the risk disclaimer.

The terms, defined

%K

Where the latest close sits between the highest high and the lowest low of the last N bars, as a percentage of that range.

%D

A short moving average of %K, conventionally 3 periods. An average of a range position.

Fast and slow stochastic

The same calculation with fewer or more smoothing passes applied. Slow is steadier and later; neither measures anything new.

Saturation

A bounded reading sitting at the top or bottom of its scale while price continues. Forced by the bound, not evidence of exhaustion.

Scale-free

Having no reference to size, because the quantity divides by its own range. Convenient, and the reason an oscillator cannot tell a big move from a small one.

Range position

What this indicator measures, and the only thing it measures: a close expressed as a fraction of a recent high-to-low span.

What this means in practice

• Read the number as "the close landed this far up the last N bars’ range", and refuse to read it as a percentage of a move.
• Never sell on 80 or buy on 20. Both are where the readings belong in a trend, and the window’s own high is in the denominator making it worse.
• Put a size measure beside it. The oscillator cannot tell a half-point range from a fifty-point one.
• Discount crosses that happen near an extreme - there was almost no room for them not to happen.
• Establish whether you are in a range or a trend from structure first. The same reading means opposite things in each.
• Fix N and the smoothing before you look at the chart, and keep them.

Nothing in this lesson is financial advice or an instruction to trade. See the risk disclaimer. We never recommend a leverage setting - that decision is yours.