Indicator library · Momentum
Price Momentum Oscillator (PMO)
A rate of change put through two rounds of smoothing, with a signal line on top. Because its input is a percentage rather than a price, PMO values can be compared across a list of instruments, which is what it was built for.
The calculation
Carl Swenlin developed the Price Momentum Oscillator at DecisionPoint. It is built in four steps, and the third is the one that trips people up.
- Take the one-period rate of change of the close: ((close ÷ previous close) − 1) × 100. The input is a percentage, which is what later makes values comparable between instruments.
- Smooth it over 35 periods using Swenlin's custom smoothing: an exponential average whose multiplier is 2 ÷ n, not the conventional 2 ÷ (n + 1).
- Multiply by 10. A cosmetic step with no analytical content: it puts the result on a scale that reads comfortably on a chart. It is also why PMO values look nothing like the rate-of-change values they came from.
- Smooth again over 20 periods with the same multiplier. The signal line is then a 10-period smoothing of the finished PMO.
Two rounds of smoothing are the point of the indicator, not an implementation detail. A single-period rate of change is almost pure noise; smoothing it twice produces a line slow enough to have a readable slope, at the cost of turning later than the price does. Every property below follows from that trade.
Reading it
The zero line
Above zero, the smoothed rate of change is positive: recent percentage gains outweigh recent losses. Crossings are the slowest and most conservative signal the indicator offers, and they arrive well after the turn. Their value is as a filter, a rule that only takes long setups while PMO is above zero will miss the start of every advance and avoid most of the counter-trend attempts.
Signal-line crossings
PMO crossing its own 10-period smoothing is the faster reading, and correspondingly noisier. In a trending instrument these crossings arrive in clusters around consolidations, which is where the indicator produces most of its false starts.
Rank, not level
The use that justifies the construction: because the input is a percentage, the PMO of twenty different ETFs can be sorted into a single list. That answers "where is momentum concentrated right now" in a way MACD cannot, since MACD values carry the price level of each instrument. Sorting is a comparison of like with like; reading a PMO level as overbought is not.
Where it came from, and what problem it solved
Carl Swenlin built the Price Momentum Oscillator at DecisionPoint, a service whose whole output was ranked lists, sectors, industry groups, exchange-traded funds, sorted by momentum. That context explains the design entirely. A ranking tool has one hard requirement: its values must be comparable between the things being ranked. MACD fails that test because it carries each instrument's price level; a raw rate of change passes it but is too noisy to sort usefully, since the ranking would reshuffle daily on nothing.
PMO is the resolution: take the scale-free measure, then smooth it until the ordering is stable enough to act on. Every other property of the indicator (the lag, the unbounded scale, the absence of meaningful thresholds) falls out of that single purpose. It was never designed to tell you when to buy one instrument, which is how it is most often used and where it disappoints.
What each smoothing pass does
The two rounds are not redundancy. A one-period rate of change is close to pure noise: on any given session it is a single day's percentage move, and the series jumps between positive and negative constantly with no persistent shape.
The first pass over 35 periods turns that into something with a trend in it, but a trend with corners, because a single large session still pushes the average sharply and the line inherits the kink. The second pass over 20 periods removes the corners, which is what leaves a line whose slope can be read. That is the property PMO exists to provide: not the level, and not the crossing, but a direction-of-travel that does not reverse on one session's noise.
Two consequences follow, and neither is a defect to tune away. The line turns late at bottoms, because two averages must both roll over before it can. And it turns early at tops, because a decelerating advance shows up in the rate of change while price is still making progress. Asymmetric behaviour at the two ends is what double smoothing does.
How it compares with the alternatives
| Measure | Input | Bounded? | Comparable between instruments? |
|---|---|---|---|
| PMO | Rate of change, smoothed twice | No | Yes, the input is a percentage. This is the point of it. |
| MACD | Difference of two price averages | No | No, carries the instrument’s price level, and resets on a split. |
| Rate of change | Percentage change over n periods | No | Yes, but too noisy to rank on without smoothing. |
| RSI | Gains against losses, smoothed | Yes, 0–100 | Levels transfer; it is not a ranking measure despite the name. |
Read across the last column and the division of labour is obvious. Use a bounded measure when you want a level that means the same thing everywhere, and a percentage-based one when you want to sort a list. Reaching for PMO to answer "is this overbought" asks the one question its scale cannot answer.
A worked reading
Suppose you hold a list of twelve sector funds and sort them by PMO. The top three are above zero and rising, the bottom three are below zero and falling, and the middle six are clustered close together. Three statements follow.
- Supported: momentum is concentrated in the top three, their smoothed percentage gains have been accelerating relative to the rest. That is exactly what the indicator was built to say.
- Supported: the clustered middle six are not distinguishable by this measure. A ranking whose neighbours differ by less than the indicator's own noise is a false ordering, and treating positions four through nine as a sequence is reading precision that is not there.
- Not supported: that the top three are due to continue or the bottom three to reverse. PMO describes what has happened to the rate of change; whether it persists is a separate empirical question the indicator does not address.
The practical use is therefore narrower than it looks and more defensible: it tells you where to look first in a list too long to examine one by one, and it tells you when the list has no clear leaders. Both are useful. Neither is a signal.
Where it misleads
| Situation | What goes wrong |
|---|---|
| Universal thresholds | PMO is unbounded and unnormalised. Overbought and oversold levels quoted for one instrument mean nothing on another. |
| Wrong smoothing constant | A platform using 2 ÷ (n + 1) instead of 2 ÷ n draws a different line for the same stated settings. |
| Short history | Double smoothing needs several hundred bars to converge; a short series produces values that depend on where it starts. |
| Sharp reversals | The lag that makes the line readable also makes it late. In a V-shaped bottom the crossing arrives well into the recovery. |
| Illiquid instrument | A single-period rate of change on a stale or gapping quote injects percentage moves that were never traded. |
Volume gives the crossing a second reading
A zero-line crossing says the smoothed percentage change has turned positive. It says nothing about whether anyone was there. The same crossing on expanding volume, with the advance/decline ratio confirming that most issues took part, describes a broad turn; the same crossing on thinning volume in a narrowing market describes a drift. The oscillator cannot tell them apart, because the only thing it ever looks at is the closing price.
Frequently asked questions
Why does my PMO differ from another platform’s?
Two reasons, and the first is almost always the culprit. Swenlin defined the smoothing with a multiplier of 2 ÷ n, where conventional exponential averages use 2 ÷ (n + 1); a platform that quietly substitutes the standard EMA produces a visibly different line for the same stated periods. The second is seeding: the first value has to start somewhere, and a series beginning 40 bars ago will not match one beginning 400 bars ago until the double smoothing converges. Feed it several hundred bars before comparing.
What is a high or low PMO reading?
There is no answer that transfers between instruments, and this is the single most misused property of the indicator. PMO is not bounded and not normalised: a volatile small-cap can reach values several times anything a broad index ever prints. A reading of 4 means nothing on its own. It is only meaningful against that instrument’s own history. Any article quoting universal overbought and oversold levels for PMO has misunderstood it.
How is it different from MACD?
Both are momentum oscillators with a signal line, but they measure different quantities. MACD is the difference between two moving averages of price, expressed in the instrument’s own units, so a $500 stock produces larger MACD values than a $5 one. PMO is built from the rate of change — a percentage — so it is at least scale-free with respect to price level, which is why it is used to rank a universe of instruments against one another. Neither is normalised for volatility.
What is it actually used for?
Two things in practice. As a trend filter: PMO above zero and rising describes an instrument whose percentage gains have been accelerating, which is a different statement from "the price is up". And as a ranking tool: because the input is a percentage, PMO values can be sorted across a list of ETFs or sectors to see where momentum is concentrated, the use Swenlin built it for at DecisionPoint.
Why smooth twice rather than once with a longer period?
Because the two operations do different things. A single long average of a one-period rate of change still passes through the sharp spikes that a single noisy session produces; it dampens them but preserves their shape. Smoothing the smoothed series removes the corners as well, which is what gives PMO a readable slope rather than a jagged line with a trend in it. The cost is a second helping of lag, and the whole character of the indicator is that trade taken twice.
Does the multiplication by 10 change anything?
Nothing analytical. It scales the output so the numbers read comfortably on a chart instead of clustering near zero with three decimal places. It does mean PMO values look nothing like the rate-of-change values they came from, which trips people up when they try to reconcile the two, and it is one more reason a level quoted for PMO cannot be interpreted without knowing whose implementation produced it.
What signal-line period should I use?
Ten is the published default and there is no derivation behind it. What changing it does is move the indicator along one axis: a shorter signal line crosses more often and earlier, a longer one less often and later. Since the crossings are already the noisiest reading PMO offers, shortening the signal line mostly buys more false starts. If you want earlier information from this indicator, the honest answer is that it does not have any, that is what the double smoothing traded away.
How do I know what is extreme for a particular instrument?
By looking at that instrument’s own distribution, which takes one pass over its history: plot the PMO, note the values it reached at the last several tops and bottoms, and use those. It is unglamorous and it is the only defensible method, because the indicator is unbounded and unnormalised. Any published table of overbought PMO levels is a table of one author’s watchlist.
Can PMO be used on an index rather than a stock?
Yes, and it behaves better there. An index is an average of many instruments, so its rate of change is already smoother than any single member’s, and PMO’s double smoothing has less noise to remove. The corollary is that index PMO values are systematically smaller in magnitude than single-stock values, so the two cannot be compared even though both are percentages, the diversification, not the momentum, accounts for the difference.
What does a divergence look like on PMO?
Price makes a higher high while PMO makes a lower one, meaning the second advance was made with slower smoothed percentage gains than the first. Because the line is heavily smoothed, PMO produces fewer apparent divergences than an unsmoothed measure, which makes the ones it does produce more interesting and correspondingly later. It remains a statement about the character of a move, never a timing signal.
Is it a leading or lagging indicator?
Lagging, and the marketing language around momentum indicators obscures this. Everything PMO knows comes from closing prices that have already happened, passed through two averages. It can turn before price does, and often does at tops, because a decelerating advance shows in the rate of change before it shows in the level, but that is not the same as leading. At bottoms it is reliably late, for the same reason.
How does it compare with RSI?
They answer different questions and only one of them transfers between instruments. RSI is bounded 0–100 and measures the balance of gains to losses, so 70 means something comparable everywhere. PMO is unbounded and measures the smoothed speed of percentage change, so its values rank instruments against each other but have no absolute interpretation. Using RSI for levels and PMO for ranking plays to what each was built for.
Does volume enter the calculation at all?
No. PMO reads the closing price and nothing else, which is worth stating plainly on a site about volume. Two sessions with identical closes produce identical contributions whether one traded ten times the other’s volume. That is the specific gap, and it is why a zero-line crossing is worth pairing with a participation measure rather than treated as a complete observation.
Should the periods be changed for a weekly chart?
The defaults were chosen on daily data, and applying them unchanged to weekly bars gives you 35 and 20 weeks of smoothing: well over a year of history in the second average, which is slower than almost any decision needs. Practitioners who use PMO weekly usually shorten it. The general rule holds: the period is a number of bars, not a duration, and moving to a different timeframe changes what that number means even though the setting looks the same.
Why do some platforms show a PMO histogram?
It plots PMO minus its signal line, exactly as the MACD histogram does. It contains no information the two lines do not, but it makes the distance between them and the rate at which that distance is changing directly visible, which is the earliest reading available and, being the earliest, the one that most often comes to nothing.