Ultimate Oscillator: three timeframes and Williams’ rules

The Ultimate Oscillator combines three measures of buying pressure, over 7, 14 and 28 bars, into one line between 0 and 100. Larry Williams designed it in 1985 to fix a weakness he saw in single-period oscillators: a short look-back reacts to noise, a long one reacts too late, and either one alone produces false divergences. Blending three timeframes, with the shortest weighted most, was his compromise.

Williams also published unusually specific rules for trading it, built around divergence rather than simple overbought and oversold levels. This guide works through the full calculation on Intel's chart, then follows those rules through Intel's remarkable 2026, when the stock more than quadrupled from its low and then gave back about a third of its value.

Buying pressure: the core idea

For each bar, Williams asked how much of the bar's true range the buyers captured.

• The true low is the lower of today's low and yesterday's close. Using the previous close accounts for gaps.
• Buying pressure is the distance from the true low up to the close.
• True range is the distance from the true low up to the true high (the higher of today's high and yesterday's close).

A close at the top of the true range means BP equals TR: buyers took everything. A close at the true low means BP is zero. Summing BP and TR over a window and dividing gives the share of the range buyers captured over that window, a number between 0 and 1.

The formula

The weights 4, 2 and 1 make the shortest window count four times as much as the longest. That keeps the line responsive while the longer windows dampen it.

A worked example on Intel

On April 16, 2026, Intel (INTC) traded between $64.27 and $68.61 and closed at $68.50, up from $64.94 the day before. Its true low that day was $64.27 and its buying pressure $4.23 of a $4.34 true range: buyers captured nearly all of it. Over the three windows ending that day, from Finabase's bars:

UO = 100 × (4 × 0.8321 + 2 × 0.7686 + 0.6192) ÷ 7 = 100 × 5.4848 ÷ 7 = 78.35

That matches Finabase's chart and was Intel's highest reading of the year. Over the previous week buyers had captured 83% of every day's range.

Williams' rules

Williams did not intend the oscillator to be traded on simple 70/30 crossings. His rules were:

Buy when there is a bullish divergence (price makes a lower low while the oscillator makes a higher low), the first oscillator low was below 30, and the oscillator then rises above the high it made between the two lows.

Sell when there is a bearish divergence (price makes a higher high while the oscillator makes a lower high), the first oscillator high was above 70, and the oscillator then falls below the low it made between the two highs.

He also used exits: close a long when the oscillator goes above 70, or when it rises above 50 and then falls below 45.

The Ultimate Oscillator on a real chart: Intel 2025–2026

Intel's year was extraordinary: from a low of $32.89 on November 21, 2025 to a high of $142.35 on June 30, 2026, more than four times the low, before falling back to $92.07 by October 29. The oscillator stayed mostly in the middle: above 70 on only 20 sessions, below 30 on 3.

The sell rule, step by step.

The divergence was stark: Intel had doubled between the two peaks, yet the second oscillator peak was 17 points lower. The confirmation came on July 7, after the stock had already fallen from its high. By July 31 Intel was at $90.20, 18% below the signal price. The rule did not catch the top; it confirmed that the advance had broken down.

An oversold reading without a rule. On December 22, 2025 the oscillator hit its low for the year, 25.98, with Intel at $36.37. That was not a Williams buy signal on its own; it needs a later higher low to form a divergence. Twenty sessions later Intel was at $54.32, nearly 50% higher, which shows both how strong the move was and why Williams insisted on structure: the oscillator stayed far from 30 for most of the rally, so a trader waiting for a textbook divergence would have had to find the trend some other way.

Pitfalls and limitations

Rare signals. Williams' full rules need an extreme reading, a divergence and a confirming break. On many stocks that happens a few times a year at most.

Late confirmation. As the July example shows, the confirming break comes after price has already turned.

Fixed weights. The 4-2-1 weighting is a design choice, not a law. Changing the periods changes the character of the line substantially.

Comparison table

See Williams %R: formula, −20/−80 levels and real signals for Williams' simpler oscillator, Stochastic oscillator (slow %K and %D): how to read it and Relative Strength Index (RSI): how to read and use it.

How Finabase calculates and scores the Ultimate Oscillator

Finabase computes it with TA-Lib's ULTOSC function on each bar's high, low and close. A new ULTOSC study ("Ultimate Oscillator") defaults to periods of 7, 14 and 28 (labelled Period 1, 2 and 3), all adjustable, and is drawn in its own panel.

ULTOSC (7, 14, 28) appears in the technical gauge on the symbol page with its latest value and carries no Buy or Sell rule, so it always counts as neutral. That is consistent with Williams' design: his signals depend on divergence and structure, which a single reading cannot capture.

Frequently asked questions

Why is it called the Ultimate Oscillator?

Williams named it for combining short, medium and long timeframes into one line, which he argued avoided the false divergences of single-period oscillators.

What are the best Ultimate Oscillator settings?

Williams' 7, 14 and 28 are the standard and Finabase's defaults. Keeping the 1:2:4 ratio when you change them preserves the design.

Is above 70 a sell signal?

Not by Williams' rules. A reading above 70 is the first condition for a bearish divergence; the sell signal comes only after a lower oscillator high and a break below the intervening low. He did use a move above 70 as an exit from long positions.

Summary

The Ultimate Oscillator measures how much of the true range buyers captured over three windows and blends them with 4-2-1 weights. It is best read with Williams' own rules, which demand an extreme, a divergence and a confirming break. On Intel's 2026 chart those rules caught the breakdown after a fourfold run, confirming it at $110.39 on the way to $90, while the year's oversold extreme preceded a 50% rally that no rule caught.