Across 1,800 four-hour samples, the median taker buy/sell ratio was 0.958.
Ten major perpetuals, 30 days, 180 four-hour candles per asset: taker buy volume divided by taker sell volume. The median for none of the ten assets was above 1. The highest was $BTC 0.979, and the lowest was AVAX at 0.924.
P5 to P95 ranged from 0.767 to 1.191, a spread of 0.42.
Only 28.0% of the time did the ratio actually fall between 0.95 and 1.05.
A lot of people think 1.0 is the dividing line between buy and sell balance. In fact, over the long run, all ten coins leaned toward taker selling. The ratio was above 1.0 only 37.1% of the time.
Not far off. Consistent direction.
The difference is how tightly distributed the values are. BNB stayed in the middle band only 22.8% of the time; for SOL and LINK, it was 32.8%. Same metric, different resolution.
So, does 1.15 count as strong buying?
Divide each asset's ratios into ten bins. The top 10% occurred 165 times; over the following 24 hours, prices rose 44.2%, with a median decline of 0.49%.
The full-sample baseline over the same period was a 45.6% rise rate and a 0.26% decline.
Entering when buying pressure was at its strongest had a win rate 1.4 percentage points lower than picking a random time. That's the story.
Chasing buying often comes in on the last leg. Price has already made three moves by the time taker buying piles up in the 90th percentile. Money entering then is taking someone else's bags. It sounds plausible, but all I have is correlation.
I can't establish causation.
The bottom 10% occurred 184 times. Prices rose 51.6% afterward, 6 percentage points above the baseline. It's precisely when taker selling was fiercest that prices were more likely to rise next.
This metric doesn't indicate direction at all. It describes trade flow, not positioning.
A high number of taker buy orders could mean people are chasing a rally, or shorts are covering; a high number of taker sell orders could mean a sell-off, or longs are stopping out. The same bar supports either story.
You also have to look at the sample size. The ten assets rise and fall together, so 184 samples really aren't 184 independent events. In practice, they cover only about 30 stretches of market action over those 30 days. There are enough data points, but not enough independent information.
So what this chart can tell you is where things are.
Location, not direction.
A ratio that spends only 28% of the time in the middle band is skewed the other 70% of the time. If you wait for it to return to 1.0 before acting, you're waiting for an uncommon case. On the other hand, seeing 0.9 and shouting “short,” or seeing 1.1 and shouting “long,” is probably just reading noise.
What I can't confirm is whether this distribution holds in a different market. Perpetual market-maker orders, hedging flows, and liquidation engines can all leave traces on the taker-sell side. That might explain why the median is below 1 for all ten coins—or it might just be a feature of these 30 days. The data can't tell those apart.
The structure worth remembering is this: a ratio treated as a directional indicator has a median that's nowhere near 1.0.
Any approach that treats a single ratio as a buy or sell signal will eventually pay tuition at the tails of the distribution.
Because location is always easier to measure than direction.
$BTC is now 80,618, $ETH 2,350; both are still hovering near their 30-day lows.
Data source: Binance Futures API https://fapi.binance.com/futures/data/takerlongshortRatio, four-hour intervals, 180 candles per asset; the unfinished candle was excluded.
Ten major perpetuals, 30 days, 180 four-hour candles per asset: taker buy volume divided by taker sell volume. The median for none of the ten assets was above 1. The highest was $BTC 0.979, and the lowest was AVAX at 0.924.
P5 to P95 ranged from 0.767 to 1.191, a spread of 0.42.
Only 28.0% of the time did the ratio actually fall between 0.95 and 1.05.
A lot of people think 1.0 is the dividing line between buy and sell balance. In fact, over the long run, all ten coins leaned toward taker selling. The ratio was above 1.0 only 37.1% of the time.
Not far off. Consistent direction.
The difference is how tightly distributed the values are. BNB stayed in the middle band only 22.8% of the time; for SOL and LINK, it was 32.8%. Same metric, different resolution.
So, does 1.15 count as strong buying?
Divide each asset's ratios into ten bins. The top 10% occurred 165 times; over the following 24 hours, prices rose 44.2%, with a median decline of 0.49%.
The full-sample baseline over the same period was a 45.6% rise rate and a 0.26% decline.
Entering when buying pressure was at its strongest had a win rate 1.4 percentage points lower than picking a random time. That's the story.
Chasing buying often comes in on the last leg. Price has already made three moves by the time taker buying piles up in the 90th percentile. Money entering then is taking someone else's bags. It sounds plausible, but all I have is correlation.
I can't establish causation.
The bottom 10% occurred 184 times. Prices rose 51.6% afterward, 6 percentage points above the baseline. It's precisely when taker selling was fiercest that prices were more likely to rise next.
This metric doesn't indicate direction at all. It describes trade flow, not positioning.
A high number of taker buy orders could mean people are chasing a rally, or shorts are covering; a high number of taker sell orders could mean a sell-off, or longs are stopping out. The same bar supports either story.
You also have to look at the sample size. The ten assets rise and fall together, so 184 samples really aren't 184 independent events. In practice, they cover only about 30 stretches of market action over those 30 days. There are enough data points, but not enough independent information.
So what this chart can tell you is where things are.
Location, not direction.
A ratio that spends only 28% of the time in the middle band is skewed the other 70% of the time. If you wait for it to return to 1.0 before acting, you're waiting for an uncommon case. On the other hand, seeing 0.9 and shouting “short,” or seeing 1.1 and shouting “long,” is probably just reading noise.
What I can't confirm is whether this distribution holds in a different market. Perpetual market-maker orders, hedging flows, and liquidation engines can all leave traces on the taker-sell side. That might explain why the median is below 1 for all ten coins—or it might just be a feature of these 30 days. The data can't tell those apart.
The structure worth remembering is this: a ratio treated as a directional indicator has a median that's nowhere near 1.0.
Any approach that treats a single ratio as a buy or sell signal will eventually pay tuition at the tails of the distribution.
Because location is always easier to measure than direction.
$BTC is now 80,618, $ETH 2,350; both are still hovering near their 30-day lows.
Data source: Binance Futures API https://fapi.binance.com/futures/data/takerlongshortRatio, four-hour intervals, 180 candles per asset; the unfinished candle was excluded.