Breadth · Volume-weighted
TRIN (Arms Index), Breadth Against Volume
Richard Arms's index divides the advance/decline ratio by the up/down volume ratio, which puts breadth and volume into one number. It is also inverted, so the strongest sessions print the lowest readings, the source of most misreadings.
The calculation
Four numbers, published daily per exchange: advancing issues, declining issues, the volume traded in advancing issues and the volume traded in declining ones.
TRIN = (advances ÷ declines) ÷ (up volume ÷ down volume)
Everything cancels except the relationship between the two ratios, which is why the units of the inputs are irrelevant (issues in hundreds, volume in millions, it makes no difference), and why the neutral reading is one rather than zero. At exactly one, each side is attracting volume in proportion to the number of issues it contains.
The inversion follows from the volume ratio being the denominator, and it is the single thing to keep hold of: a strong session prints a low number. If advancing issues are pulling more than their proportional share of the day’s volume, the denominator grows and the index falls. Below one, the advancing side carried the volume; above one, the declining side did.
Why the extremes sit at the bottom
Almost every measure of market internals is asymmetric, and TRIN is the clearest case. Its very high readings cluster at market lows, and it has no equivalent tail at highs.
The reason is in how selling happens. Forced liquidation (a margin call, a fund meeting redemptions, a mandate breached) sells without regard to price or merit, and it happens to many holdings at once. That concentrates a very large volume into declining issues over one or two sessions, which is precisely the arithmetic that produces a reading of two or three. Buying, by contrast, is deliberate and spread over weeks, and it rarely produces the mirror image.
Two consequences follow. The upper tail is where the indicator has something to say, and the lower tail, very low readings during an advance, is mostly unremarkable enthusiasm. And a top cannot be identified this way at all: distribution is quiet by nature, and the measure that describes it is narrowing breadth over months rather than a spike in a daily ratio.
Reading it against its own history
The numbers usually quoted (extreme above two, very rare above three, the opposite tail around 0.5) come from American exchange data over particular decades, and the exchange’s composition has changed since. Treating them as constants is the commonest error after the inversion.
The alternative takes an afternoon. Compute the daily series for the market you actually follow over a few years, and find its percentiles. The 95th and 99th percentile of your own data are meaningful thresholds in a way an inherited number is not, and the exercise usually reveals that the folklore figure sits somewhere unexpected, too strict on one market, too loose on another.
Keep the raw daily series and a ten-day average side by side rather than choosing one. The smoothed line answers whether volume has been favouring the declining side over a period, which the daily reading is far too noisy to say; the daily reading captures the one-session spike, which is the only part of this indicator with a reputation. Each destroys what the other is for.
Where it misleads
| Situation | What goes wrong |
|---|---|
| The inversion forgotten | A low number is a strong session. Reading TRIN like an oscillator that rises with strength inverts every conclusion. |
| Inherited thresholds | Two and three come from particular decades on particular exchanges. Recompute the percentiles from your own data. |
| Reading it at highs | There is no upper-tail equivalent at tops. Distribution is quiet, and a daily ratio cannot describe it. |
| Non-operating issues | Funds and preferreds move together on rate news, distorting the issue count and the volume count by different amounts. |
| Smoothed extremes | A moving average removes the one-day spike, which is the only reading here with a documented record. |
| Index rebalance days | Mechanical volume enters both volume terms and can produce a reading that describes fund maintenance rather than a market. |
What it does that nothing else here does
Every other breadth measure on this site counts issues and stops. TRIN is the only one that asks whether the volume matched the count, and it is worth keeping for that alone, a session in which 55 per cent of issues advanced while 75 per cent of the volume went into them is a genuinely different market from one where those numbers are reversed, and no advance/decline figure can distinguish the two.
What it gives up is the absolute level of either input. A ratio of ratios can print exactly one on a very busy session and on a very quiet one, and it will read as neutral both times. That is the argument for keeping the four raw counts in view rather than only the number derived from them, the same argument that runs through the whole of this reference, which is that a compression is only safe when you know what it discarded.
Frequently asked questions
How is TRIN calculated?
It is a ratio of two ratios: the number of advancing issues divided by the number of declining issues, divided in turn by the volume in advancing issues divided by the volume in declining issues. Written out, (advances ÷ declines) ÷ (up volume ÷ down volume). Everything cancels except the relationship between the two, which is why the units of the inputs never matter and why the result centres on one rather than on zero.
Why is it inverted?
Because the volume ratio sits in the denominator. When advancing issues are attracting more than their proportional share of the volume, the denominator grows faster than the numerator and the index falls, so a strong session prints a LOW number. A reading below one means the advancing side carried more volume than its issue count implied; above one means the declining side did. Every mistake with this indicator starts with forgetting that direction.
What is a neutral reading?
One, exactly, and it has a definition rather than a convention behind it: the two sides are attracting volume in proportion to how many issues they contain. Readings between roughly 0.8 and 1.2 are ordinary and carry almost no information. What the indicator is actually for is the tails, and specifically the upper tail, the sessions where declining issues absorbed far more volume than their number.
Why do its extremes cluster at lows rather than at highs?
Because selling is more synchronised than buying. Forced liquidation — margin calls, funds meeting redemptions — sells regardless of price or merit, concentrating enormous volume into declining issues over one or two sessions, which is exactly the arithmetic that produces a very high TRIN. Buying is deliberate, spread out, and rarely produces the mirror image. The result is that the most extreme readings in a whole decline usually arrive at its end, which is a real asymmetry rather than an artefact.
Is a high reading bullish or bearish?
On the day, it describes a bad session. Read as a condition rather than a description, a genuinely extreme reading has historically appeared closer to lows than to highs, so practitioners treat the upper tail as a contrarian observation. Both statements are true at once, and the difference between them is horizon. Neither makes it a signal: extreme readings have occurred well before the eventual low, and any threshold quoted for one market and decade should be recomputed for another.
Should it be smoothed?
Usually yes, for reading, and never for the extremes. The daily series is very noisy, and a ten-day moving average of it is a much more legible measure of whether volume has been favouring the declining side over a period. But smoothing removes precisely the one-day spike that carries most of the indicator’s value, so the two are best kept side by side rather than one replacing the other.
What thresholds are commonly quoted?
Readings above two are described as extreme and above three as very rare, with roughly 0.5 as the opposite tail. Those numbers come from American exchange data over particular decades and should be treated as folklore until checked. The defensible version is a percentile computed from the data you actually have: establish where the 95th and 99th percentiles of the daily series fall on your market, and use those rather than a number inherited from a book.
Does the exchange’s composition affect it?
Substantially, and in both terms of the ratio. A list containing several hundred closed-end funds and preferred issues will move a large number of issues together on interest-rate news while their volume stays small, which distorts the issue count and the volume count differently. A common-stock-only version of both inputs is preferable wherever the data allows, and it is one reason historical thresholds drawn from earlier decades describe a differently composed exchange.
How does it relate to up/down volume on its own?
TRIN normalises the volume split by the issue split, which is both its advantage and what it hides. Up and down volume read alone tell you where the day’s activity went; TRIN tells you whether that was more or less than the number of participating issues would suggest. A session where 60 per cent of volume went into advancing issues means something different when 80 per cent of issues advanced than when 55 per cent did, and TRIN is the number that distinguishes them, at the cost of concealing the absolute level of either.