Guides · ORDER FLOW
One candle, two tapes: spot and the perpetual disagree about who was buying
The same fifteen minutes of Bitcoin read +61% delta on spot and +22% on the perpetual. On NEAR, the sign flipped. A volume figure without a market attached is not a figure.
A volume figure without a market attached is not a figure. Neither is a delta, a point of control, or a high and a low. Spot and the perpetual are two different books with two different sets of participants, and they print two different tapes for the same minute of the same asset.
This is easy to agree with and easy to forget, because the price lines look identical. So here is one candle, read twice.
The same fifteen minutes, read on both books
Binance, 20 September 2026, 05:00 to 05:15 UTC. Both reads cover the whole quarter hour, and both separate buying from selling by the venue's own aggressor tag rather than inferring it from where the bar closed.
| BTCUSDT, 05:00–05:15 UTC | Spot | Perpetual |
|---|---|---|
| Volume | 299.76 BTC | 1,201.56 BTC |
| Aggressive buying | 240.78 BTC | 735.70 BTC |
| Aggressive selling | 58.98 BTC | 465.87 BTC |
| Delta | +181.81 BTC | +269.83 BTC |
| Delta as a share of volume | +60.7% | +22.5% |
| Trades | 7,974 | 8,811 |
| Range | 80,408.70 – 80,668.15 | 80,384.20 – 80,618.00 |
The perpetual traded four times the volume in roughly the same number of trades, so the average clip was about three and a half times larger. That much is unsurprising.
The second-to-last row is the one worth sitting with. On spot, buyers crossed the spread for 61% of everything that traded: a tape that reads as one-sided demand. On the perpetual, the same fifteen minutes reads as a market with two active sides and a modest tilt — 22.5%. Both numbers are correct. They describe different crowds.
The last row matters for anyone drawing a level. The two books did not agree on the extremes of the move: spot printed a high 50 dollars above the perpetual's, and a low 24 dollars above it. A wick you marked on one book is not there on the other.
When the sign itself changes
Bitcoin is the forgiving case. Take the same candle on a mid-cap, NEARUSDT:
| NEARUSDT, 05:00–05:15 UTC | Spot | Perpetual |
|---|---|---|
| Volume | 418,883 NEAR | 451,626 NEAR |
| Delta | +8,473 NEAR | −13,826 NEAR |
| Delta as a share of volume | +2.0% | −3.1% |
| Point of control | 3.481 | 3.482 |
| Delta at the point of control | +1,961 | −24,760 |
Same asset, same minutes, opposite conclusions. On spot, aggressive buyers were slightly ahead. On the perpetual, sellers were. At the price where most of the volume changed hands, the two books disagree by a factor of twelve and a sign.
If you are trading the perpetual and reading the spot tape, you are not looking at a slightly different version of your market. You are looking at a different answer.
Why the two books diverge
Arbitrage keeps the two prices close — the basis at the time of this candle was −0.038%, about thirty dollars on eighty thousand — but nothing keeps the flow similar, because the reasons for trading are not the same.
The perpetual carries leverage, and leverage carries forced flow: liquidations, margin top-ups, stops that must be taken regardless of what anyone thinks of the price. It also carries hedges. A market maker filling spot demand often offsets it on the perpetual, so one trade in the world becomes aggressive buying on one book and aggressive selling on the other.
It is also where the positioning data lives. At the time of this candle, open interest on the Bitcoin perpetual stood at 8.79 billion dollars, funding at 0.01% per interval — about 10.95% annualised — with the account long/short ratio at 0.92 while the top-trader ratio was 2.14. None of those numbers exist on spot. There is no funding on a spot book, no open interest, no liquidation to cascade.
Spot, meanwhile, is where an allocation actually changes hands. It is slower, thinner outside of peak hours, and much more sensitive to a single large participant — which is exactly why a 61% delta share on spot should be read as "someone was buying", not as "the market was buying".
What this breaks in practice
Comparing figures across books. A volume from one market against a liquidation map from the other is two unrelated numbers in one sentence. It looks like analysis and carries no information.
Levels. Highs, lows and points of control differ. On this candle, by tens of dollars on Bitcoin and by a whole price bucket on NEAR. If your entry sits inside that gap, you will be filled on one book and not on the other.
Delta divergence. The classic reading — price makes a high, delta does not — is only a signal within one book. Half of the divergences you will find by mixing books are an artefact of the mixing.
Anything an AI assistant hands you. A model reading a tape has no way of noticing that it is on the wrong book unless the answer says which book it read. Ask for the market to be named, every time.
The rule
Read the tape of the market you are trading. If you trade the perpetual, read the perpetual: its liquidity is what fills you and its participants are who you are trading against. If you hold spot, the perpetual is still worth watching, but as context about leverage, not as your tape.
And when you quote a number — in a note, in a group chat, to a model — say which market it came from. A delta of +182 BTC and a delta of +270 BTC described the same fifteen minutes above. Without the label, neither figure is worth repeating.
Every figure here was read with the Basis order flow tools on Binance spot and Binance USDⓈ-M perpetuals, over the candle stated, with full coverage of the window. In the Basis terminal the market is the toggle beside the symbol; over the MCP connection, every data tool takes a market parameter for one call, so an agent can read the perpetual while the chart stays on spot.
4 min read · Filed to Guides · Nothing here is investment advice.