What Is the Ultimate Oscillator?

Michael Whitman · Trading since 2017, full-time since 2020 · Reviewed 2026-09-14

The ultimate oscillator exists because single-period oscillators give false signals when the period does not match the market. Its solution is to use three periods at once — which solves one problem and introduces another.

How it works

Three momentum readings, combined. A short lookback, a medium one and a long one are each computed, then weighted together into one value between zero and one hundred.

The short period gets the largest weight, conventionally four times the long one, on the reasoning that recent action matters most while the longer windows supply context.

The problem it addresses is real. A single-period oscillator reads overbought whenever its window happens to be short relative to the move underway, producing signals that are artefacts of the setting rather than statements about the market.

But the weights are asserted, not derived. Nobody measured that four-to-two-to-one is correct. It is a sensible-sounding ratio that became convention through use.

Three dials instead of one

Every parameter is a place to fit the past. One lookback gives one dial to tune; three lookbacks plus three weights gives six. The indicator that was built to reduce arbitrariness contains substantially more of it.

And the averaging costs responsiveness. Blending a long window into every reading means the line moves more slowly than the short window alone would — so the false signals are reduced and the genuine ones arrive later.

That trade is unavoidable and it is not unique to this tool. Every smoothing buys calm with lag.

A worked example

Take this site’s shared series. Direction runs average 2.01 bars with a longest of 11. Median bar range is 0.493, ninetieth percentile 1.101.

The default periods are 7, 14 and 28. Against two-bar average runs, the 7-period component is responding to roughly three runs, and the 28-period component to about fourteen.

So the long window is measuring something with no persistence behind it. Fourteen direction runs inside one lookback means that component is close to a constant, contributing context that is mostly noise averaged flat.

Which explains the behaviour people notice: the indicator rarely reaches its extremes. Blending a near-constant into every reading pulls the line toward the middle, so the signals it was built to give become rare.

The overbought trap is unchanged

Reaching 70 means price has risen strongly. That is a description of strength, not a forecast of reversal, and it is exactly the same misreading that afflicts every bounded oscillator.

Strong markets stay overbought. On this series 85% of the 39 twenty-bar breakouts continued in the breakout direction — so the condition that looks like exhaustion is, measurably, more often continuation.

Combining three periods does not change this. It changes how often the reading is reached, not what the reading means.

The original data

On this site’s shared series: direction runs average 2.01 bars with a longest of 11. Breakouts continued in 85% of 39 twenty-bar events. A round trip costs 0.0098, about 2% of the median bar range of 0.493.

That 85% is the number that should govern how any overbought reading is used. An indicator saying “this has gone far” is describing the precise condition that, on measured data here, continued more often than it reversed.

Where it came from

Larry Williams published it in 1985, explicitly as a response to the single-period oscillator problem. The stated case was that a trader watching one lookback is hostage to whether that lookback matches the market’s current rhythm, and that no single period matches for long.

The diagnosis holds up. Anyone who has watched an oscillator sit at an extreme through an entire move has met the problem it describes.

The remedy is where reasonable people disagree. Averaging three periods does reduce the dependence on any one of them, and it also guarantees that every reading is partly computed from a window that does not fit. The single-period version is wrong sometimes; the three-period version is slightly wrong always.

Which of those is preferable depends on what you do with it. As a context reading, always-slightly-wrong is fine. As an entry trigger, the lag that averaging introduces has to be paid for by a wider stop, and that cost is rarely counted when the indicator is chosen.

When it fails

The characteristic failure is fading an extreme reading. The line reaches 70, the label says overbought, and a short is taken against a market that has demonstrated strength. On this site’s series the measured continuation rate after a breakout is 85%, so the trade is being placed against roughly five-to-one odds — and the indicator gave no warning because reporting strength is all it ever did. The word “overbought” carries an implication the arithmetic does not support.

A second failure is optimising all three periods on history, which is six degrees of freedom fitted to one sample.

A third is expecting it to be faster than a single oscillator. It is slower, by construction.

A fourth is using it on an instrument with no persistence, where the long component averages to a constant and contributes nothing.

And a fifth is treating a divergence as a signal on its own, which is one observation about two lines and no information about size or timing.

One practical note on reading it. Because the blended construction pulls the line toward the middle, the conventional 30 and 70 levels are reached less often than on a single-period oscillator. Traders who move the thresholds inward to get more signals have undone the filtering the indicator was chosen for — which is a common adjustment and worth recognising as a reversal of the original decision.

Related

RSI covers the single-period oscillator this was designed to improve on. Moving average covers the smoothing every component depends on. And technical analysis covers the wider tradition.

This page is educational, not financial advice. Test every idea on your own charts before risking money.