How to Read the Taker Buy/Sell Ratio
Every exchange tells you how much of today's volume was buyers hitting the offer. Here is what that number is made of, what it has been worth, and why the tape now agrees with the candle almost every day.
By Pavel Penev, MScFounder, TradeWize · 10+ years trading the marketsThe short answer
The taker buy/sell ratio is the volume that bought by crossing the spread divided by the volume that sold by crossing it. A reading of 1.20 means aggressive buyers traded 20% more than aggressive sellers. It measures who wanted it badly enough to pay up. Six years of bitcoin, ether and solana say the number is almost always within a few percent of 1, that it tracks the day's candle so closely it adds nothing to it, and that it tells you nothing about tomorrow. The one thing it used to catch — a day when price went one way and the aggressors went the other — has nearly stopped happening.
Every trade has two sides, and only one of them chose the moment. Somebody left a limit order resting on the book and waited. Somebody else looked at the price, decided it was good enough, and crossed the spread to take it. The second person is the taker. They paid the spread for the privilege of trading now.
Exchanges tag every trade with which side was the taker. Add up a day's taker-buy volume, divide by its taker-sell volume, and you have the ratio. It shows up on every perpetual futures data page as "Taker Buy/Sell Volume", and the reading is always the same: buyers are aggressive, so they're in control, so follow them. Or the opposite, if you're a contrarian. So this is the number.
What the number actually counts
Volume tells you how much traded. The taker ratio tells you which side was impatient. If 100 contracts traded in an hour and 60 of them were a buyer lifting the offer, the taker-buy volume is 60, the taker-sell volume is 40, and the ratio is 1.50.
It is a measure of size, not of people. The long/short ratio from last week counts accounts, so a $50 trader and a $5 million trader each count once. Here a $5 million market order counts a hundred thousand times more than a $50 one. That is the whole difference between the two gauges, and it is why they behave nothing alike.
The long/short ratio
How many accounts lean each way.
- Counts heads, whatever their size.
- Sits far above 1 as a resting state — 1.5 to 2.5.
- Swings against each big candle.
The taker buy/sell ratio
How much volume crossed the spread each way.
- Counts size, whoever traded it.
- Sits within a few percent of 1 almost every day.
- Moves with each big candle — it is the candle.
Finding 1: it lives in a narrow band, a hair under 1
Start with what an ordinary day looks like. Bitcoin's median daily ratio over six and a half years is 0.99. Ether's is 0.98, solana's 0.97. The middle 90% of bitcoin's days sit between 0.90 and 1.09. The most lopsided day in the whole sample was 1.30.
| Contract | Median ratio | 5th percentile | 95th percentile | Days above 1 |
|---|---|---|---|---|
| bitcoin | 0.99 | 0.90 | 1.09 | 41.4% |
| ether | 0.98 | 0.89 | 1.08 | 35.0% |
| solana | 0.97 | 0.87 | 1.05 | 23.4% |
| bitcoin, spot | 0.97 | 0.83 | 1.11 | 31.0% |
Taker-buy volume divided by taker-sell volume, per UTC day. Bitcoin and ether from 2020-01-01, solana from 2020-09-14, to 2026-09-11.
Two things to notice. First, the band is tight. A reading of 1.10 is not "buyers are in control" — it is a top-5% day on bitcoin, and it happens a couple of times a month. Second, the number sits slightly under 1. Bitcoin's takers were net buyers on only 41% of days, solana's on 23%. On a perpetual, the impatient side is a little more often the seller.
The shaded band holds the middle 90% of days. The dashed line is 1.0 — buyers and sellers crossing the spread in equal size.
Why it leans under 1
Somebody has to be on the passive side of every trade, and on a perpetual that is mostly market makers and larger accounts working limit orders. The crowd, which uses market orders, is a net buyer by headcount — last week's number — but it sells in bigger clips than it buys: a stop-loss is a market sell, a liquidation is a market sell, and profit-taking after a run is usually a market sell. So by size the aggressive flow tilts a few percent to the sell side, and the resting bids absorb it. That tilt is who is in the room, not a view.
Finding 2: it is the candle
Here is the mechanical fact that explains most of what the ratio does. Take every day bitcoin rose 5% or more and look at the ratio on that same day. The median is 1.06, and it sits above 1 nine times in ten. Now take every day it fell 5% or worse: the median is 0.93, above 1 one time in twenty.
Median taker ratio on days price moved 5% or more. Green: up days. Red: down days. The number is on whichever side the candle is.
That is not a discovery. It is arithmetic. Price rises when buyers cross the spread faster than sellers do, so a green day and a ratio above 1 are the same event described twice. The rank correlation between the daily ratio and the daily return is +0.66 on bitcoin, +0.62 on ether, +0.48 on solana. Hour by hour it is +0.67. Whatever clock you put it on, the ratio and the candle are one line.
The perpetual is where this happens. On bitcoin, Binance's perpetual turns over about 7 times the dollar volume of its spot market, and the perpetual ratio tracks the day's move more tightly than the spot ratio does (+0.66 against +0.55). The two ratios agree with each other only loosely (+0.49), and neither one leads the other into the next day (−0.01 and +0.01). The spot tape is a smaller, noisier copy of the same candle.
The practical version
If bitcoin closed up 5% and the taker ratio reads 1.06, you have not learned that buyers are in control. You have learned that the candle was green. The ratio's only job is to tell you HOW the candle was made — with aggression or without — and finding 4 is about the rare day when those two disagree.
Finding 3: it does not predict, on any clock
So the ratio describes today. Does it say anything about tomorrow? Put the day's ratio against the next day's return: the rank correlation is +0.01 on bitcoin, −0.03 on ether, +0.03 on solana. Against the next week it is +0.04, +0.00, +0.02. Those are zeros.
Go faster and it gets no better. Bitcoin's hourly ratio has a correlation of −0.03 with the next hour's move, and its correlation with its own previous hour is +0.01. An hour of heavy aggressive buying says nothing about whether the next hour will have any. There is no session pattern to lean on either: the median ratio by hour of the UTC day runs from 0.98 to 1.00, which is to say flat.
The fairest test is the one from last week. Rank each day's ratio against that contract's own trailing year — so "extreme" means extreme for that contract, then, with no peeking — and score the bottom 10%, middle 80% and top 10% on what price did over the following week.
| Contract | Takers least buying (bottom 10%) | Middle 80% | Takers most buying (top 10%) |
|---|---|---|---|
| bitcoin | −0.1% · up 48% of the time | +0.6% · up 52% | +1.3% · up 54% |
| ether | +0.3% · up 55% of the time | +1.0% · up 51% | +0.5% · up 47% |
| solana | +1.4% · up 51% of the time | +0.8% · up 48% | +0.0% · up 46% |
Mean return over the following 7 days. Each day ranked against its own trailing 365 days. Gross, in-sample, one venue.
On bitcoin the heavy-buying days did a little better than the middle. On solana they did a little worse, and the heavy-selling days did a little better. On ether both extremes sat under the middle. When the same test points three different ways on three contracts, and every gap is about a percentage point, it isn't pointing. "Follow the aggressors" and "fade the aggressors" both came out as noise.
Learn it by doing
Reading about it is one thing — it clicks when you do it. Learn it hands-on with free, interactive lessons on TradeWize.
Try the free lesson →Finding 4: the one reading that paid has almost stopped happening
There is one reading of the ratio that order-flow traders have always cared about, and it isn't the level. It is the disagreement: a day when price fell hard but the takers were net BUYERS, or rose hard while the takers were net SELLERS. The candle went one way and the aggressors went the other, which means the passive side — the resting orders — did the moving.
Define it plainly. A disagreement day is a 3%-or-more move with the ratio on the wrong side of 1. Over the whole sample, a red disagreement day on bitcoin was followed by a mean +7.8% over the next week, up 82% of the time, against +1.5% after an ordinary red day. Ether: +3.3% against +1.4%. Solana: +10.5% against +1.3%. The green disagreement — a rally the takers sold into — also beat the ordinary green day on all three (+2.0% against +1.3% on bitcoin, +5.3% against +2.0% on solana).
| Contract | Red day, takers buying | Red day, takers selling | Green day, takers selling | Green day, takers buying |
|---|---|---|---|---|
| bitcoin | +7.8% (11 days) | +1.5% (241 days) | +2.0% (43 days) | +1.3% (262 days) |
| ether | +3.3% (18 days) | +1.4% (376 days) | +3.7% (139 days) | +0.7% (313 days) |
| solana | +10.5% (22 days) | +1.3% (487 days) | +5.3% (272 days) | +2.0% (275 days) |
Mean return over the following 7 days. "Takers buying" = ratio above 1 on the day; "takers selling" = below 1. Whole sample.
Look at the day counts before you get excited. Bitcoin has had 11 red disagreement days in six and a half years, against 241 ordinary ones. And they are not spread across the sample. They belong to 2020 and 2021.
Rank correlation between the daily taker ratio and the same day's return, per calendar year. Higher means the tape and the candle agree more often.
That chart is the real finding. In 2020 bitcoin's same-day correlation was +0.49; by 2023 it was +0.85. Solana's went from +0.28 in 2021 to +0.73 two years later. Count the disagreement days on bitcoin by year and you get 19, 21, then 3, 1, 2, 8, 0. In 2021, 15% of big days went against the flow. Since 2022 it has been 4%.
So the reading that paid — passive orders overpowering aggressive ones, the classic absorption print — now barely occurs on the daily clock. Bitcoin had 11 disagreement days in the whole second half of the sample, from 2023 on, and the green ones returned −0.1% over the next week. Solana still throws a few dozen a year, and in its second half the green disagreement made +0.7% against +1.8% for the ordinary green day. The edge is gone there too.
Why the tape and the candle coupled up
The data can't say why, but the shape of it points somewhere. A disagreement day needs a passive side deep enough to move price against the aggressors — big resting bids soaking up a sell-off, or big offers capping a rally. As the perpetual market matured, the passive side got thicker and more professional, and a large player who wants to move price now does it with the aggressive side, sliced up, rather than by sitting on the book. The result is a market where the candle is made by the takers almost every day, and the ratio has nothing left to add.
Finding 5: on the slow clock, the sign flips
One more clock. Instead of a day, take a rolling six months: the total taker-buy volume over 180 days divided by the total taker-sell volume, against what price did over those same 180 days. On the daily clock the two move together. On this one they move apart. The rank correlation is −0.36 on bitcoin, −0.45 on ether, −0.51 on solana. The six-month stretches where the takers were net buyers were the stretches where price fell.
| Year | Bitcoin median | Bitcoin return | Ether median | Ether return |
|---|---|---|---|---|
| 2020 | 0.99 | +302% | 0.97 | +464% |
| 2021 | 0.98 | +60% | 0.96 | +399% |
| 2022 | 1.00 | −64% | 1.00 | −68% |
| 2023 | 0.99 | +156% | 0.99 | +91% |
| 2024 | 0.99 | +121% | 0.98 | +46% |
| 2025 | 0.98 | −6% | 0.98 | −11% |
| 2026 (to date) | 0.99 | −12% | 1.00 | −15% |
Median daily taker ratio per calendar year, and the year's price change from its first open to its last close.
2022 is the only year in the sample where the bitcoin and ether medians sat above 1 — the takers were net buyers on more days than not — and bitcoin fell 64%, ether 68%. The big up years read the other way: 2021's ether median was 0.96, its lowest, with the takers net buyers on 18% of days, and it finished up 399%.
This is what a bear market looks like from the tape: people buying the dip with market orders, into resting sells that keep getting refilled lower. And what a bull looks like: people taking profit with market orders, into resting bids that keep getting refilled higher. On the day the aggressor moves the price. Over a year the aggressor is the one who's wrong.
A description, not a forecast
Before you build a strategy on it: the six-month ratio's correlation with the NEXT six months is +0.10 on bitcoin, −0.12 on ether and +0.20 on solana — no consistent sign — and the sample holds about 12 non-overlapping windows per contract. The slow clock tells you which kind of market you have been in. It does not tell you which kind you are about to be in.
How to read it yourself
- Read it as a description of the candle, not a signal on top of it. A ratio above 1 on a green day and below 1 on a red day is the normal case. It confirms what you can already see.
- Know the scale. The middle 90% of days sit roughly between 0.90 and 1.09 on bitcoin. A reading of 1.10 is a top-5% day; nothing on the daily clock ever reads 2.0.
- Don't use the level to forecast. Its correlation with the next day, the next week and the next hour rounds to zero, and the extreme buckets point different ways on different contracts.
- The reading worth a second look is the disagreement — a big candle with the takers on the other side. It used to precede a good week. It now happens a handful of times a year on bitcoin, so treat it as a curiosity to check, not a setup to wait for.
- On a six-month view, invert your instinct. Sustained net aggressive buying has been the signature of a falling market, not a rising one — because dip-buyers use market orders and profit-takers do too.
- Read it beside the gauges that measure something else. The long/short ratio counts people, open interest counts commitments, funding prices the lean. This one measures impatience, and impatience is mostly spent by the close.
What is the taker buy/sell ratio?
The volume that traded by crossing the spread to buy, divided by the volume that traded by crossing the spread to sell, over a period. The taker is the side that used a market order or an aggressive limit order and got filled immediately; the other side was a resting order. A ratio of 1.2 means aggressive buyers traded 20% more volume than aggressive sellers. Exchanges publish it because they tag every trade with which side was the taker.
What does a taker buy/sell ratio above 1 mean?
More volume was bought aggressively than sold aggressively. On its own that mostly tells you the candle was green: on bitcoin the ratio sat above 1 nine times in ten on a +5% day and one time in twenty on a −5% day. It says nothing reliable about the next candle — the correlation with the next day's move is +0.01.
Is the taker buy/sell ratio a leading indicator?
Not in six and a half years of Binance data. The ratio's correlation with the next day is +0.01 on bitcoin and −0.03 on ether; with the next hour it is −0.03. The top and bottom 10% of readings, ranked against each contract's own trailing year, were followed by weeks within about a percentage point of the middle, and in different directions on different contracts.
What is the difference between the taker ratio and the long/short ratio?
The long/short ratio counts accounts — how many are positioned long against how many short — so every trader counts once regardless of size. The taker ratio counts volume — how much was traded aggressively each way — so a large order counts for more than a small one. They behave very differently: the long/short ratio sits well above 1 and swings against each big candle; the taker ratio sits close to 1 and moves with it.
What is a taker buy/sell divergence?
A day when price moved one way while the aggressive volume went the other — a sharp fall with the ratio above 1, or a sharp rise with it below 1. It means resting orders, not market orders, moved the price. Historically such days were followed by a better-than-average week, but they have become rare: bitcoin had 19 of them in 2020 and 0 in 2026 so far.
Where can I find the taker buy/sell ratio?
Binance, Bybit and OKX each show it on their futures data pages, usually as 'Taker Buy/Sell Volume' with a month or so of history at five-minute to daily intervals. It can also be computed from any candle feed that reports taker-buy volume alongside total volume — Binance's public kline data does, which is where the full multi-year record used here comes from.
Method, and what this cannot tell you
Each UTC day's ratio is that day's taker-buy volume divided by its taker-sell volume, read straight from the exchange's daily candle; the hourly series is the same calculation on hourly candles. The ratio is compared with the same day's price change, with the next day's, week's and month's, and — ranked against its own trailing 365 days so that 'extreme' never uses the future — sorted into the bottom 10%, middle 80% and top 10% of readings and scored on the following week. A disagreement day is a day price moved 3% or more while the takers were net on the other side. The slow clock is the ratio over a rolling 180-day window against that window's return and the next window's.
- One venue. Binance classifies its own trades; the taker ratio on Bybit or OKX is a different set of aggressors and can disagree.
- Volume, not people. A single account hitting the offer all day and a thousand accounts doing it once each look identical here. The long/short ratio counts heads; this counts size.
- A day is a coarse clock. Most of the ratio's information is spent inside the candle it describes — the same-day correlation is the strongest number in the study, and the hourly one is just as strong. Anything the ratio knew, the candle has already priced.
- The disagreement days cluster in 2020-2021 and have nearly vanished since, so their payoff cannot be checked out of sample. The finding is that they used to pay and no longer occur, not that they still pay.
- The slow-clock result is a description of the years in the sample — six on bitcoin, one full bull-bear cycle — not a forecast. Its next-window correlation has no consistent sign.
- In-sample and gross. No fees, no funding, no slippage, and no attempt to trade any of it.
Learn to read the tape, not just the candle
Our crypto track builds the perpetual contract from scratch and then teaches the gauges around it — funding, open interest and the cascade — on real charts, one drill at a time. By the end you can tell a market that is being pushed from one that is being absorbed.