Order Flow Trading: What It Is, How Delta Works, and a Week of Real BTC Trades
Order flow trading reads the volume traded at the bid against the volume traded at the offer, rather than just the total. The running difference is called delta, and a cumulative delta that disagrees with price is the reading most people are after.
Covered on this page: TradingView.
Order flow is the record of every trade sorted by which side crossed the spread: a buyer who paid the offer, or a seller who hit the bid. Order flow trading reads that split, called delta, instead of total volume alone, to judge who was pushing price at each moment.
Most guides describe the idea and stop. This page builds real order flow from every BTC/USD trade Kraken printed in the week of 25 to 31 October 2026, 911,965 of them, and checks what it explained and what it did not.
How it works
Every trade has two sides, and only one of them chose to cross the spread. A limit order rests in the order book and waits. A market order crosses the bid-ask spread and takes it. The resting side is the maker; the crossing side is the taker, sometimes called the aggressor.
Order flow counts the takers. Size bought at the offer is buying flow. Size sold at the bid is selling flow. Three numbers follow from that:
- Delta is taker buying minus taker selling in one bar.
- Cumulative delta is the running total of delta, drawn as a line under price.
- A footprint splits one bar by price level, with selling and buying side by side at each level (our footprint chart page covers the display).
The exchange has to tell you the side. Kraken’s public trade feed marks each trade “b” or “s” for whether the taker was the buyer or the seller, which is what makes the numbers on this page real order flow rather than an estimate. Many stock charts only show total volume per bar.
A phrase that means something else. In brokerage, payment for order flow is a broker selling your orders to a wholesaler, covered on our payment for order flow page. Order flow trading has nothing to do with it.
The shortcut most charts use, and how often it is wrong
Without trade-by-trade data, charting tools estimate delta from the candle: a bar that closed up counts all its volume as buying, a bar that closed down counts it all as selling.
On the week of real data, that shortcut got the sign wrong in 758 of 2,001 five-minute bars (37.88%). In those bars price closed one way while takers, on balance, traded the other way. Those are exactly the bars order flow exists to show, and the shortcut labels every one of them backward. Six bars closed flat and are left out, as are the 9 five-minute windows with no trades.
A worked example
The bar from 15:20 to 15:25 UTC on 26 October 2026, the week’s largest disagreement between delta and candle. BTC/USD opened at $80,483.90, traded between $80,463.30 and $80,701.70, and closed at $80,535.70: up $51.80. A candle chart calls it a buying bar.
The 1,904 trades inside it say otherwise:
- Takers bought 32.95 BTC (769 trades) and sold 233.58 BTC (1,135 trades).
- Delta = 32.95 - 233.58 = -200.63 BTC.
Here is the footprint in $25 levels, BTC sold at the bid on the left and bought at the offer on the right:
| Price level | Sold by takers (BTC) | Bought by takers (BTC) |
|---|---|---|
| $80,700 to $80,724.9 | 0.00 | 0.02 |
| $80,675 to $80,699.9 | 0.21 | 0.87 |
| $80,650 to $80,674.9 | 0.01 | 4.02 |
| $80,625 to $80,649.9 | 0.00 | 17.96 |
| $80,600 to $80,624.9 | 0.01 | 4.84 |
| $80,575 to $80,599.9 | 0.04 | 0.07 |
| $80,550 to $80,574.9 | 6.01 | 3.16 |
| $80,525 to $80,549.9 | 59.49 | 0.33 |
| $80,500 to $80,524.9 | 100.22 | 1.68 |
| $80,475 to $80,499.9 | 65.09 | 0.00 |
| $80,450 to $80,474.9 | 2.51 | 0.00 |
Reading it: 224.80 BTC was sold into bids between $80,475 and $80,549.90, and price did not fall through that band. Resting buyers took all of it, which is what traders call absorption. A smaller burst of buying, 26.83 BTC between $80,600 and $80,674.90, printed the high.
What came next: the following bar closed at $80,431.10, below this bar’s low. One bar settles nothing in either direction. It shows how delta and the candle can disagree, not what to do about it, and the next section counts what happened across the whole week.
The original data
The week ran from 00:00 UTC on 25 October to 00:00 UTC on 1 November 2026: 911,965 trades after removing 912 rows that appeared twice where download pages overlapped, 15,924.16 BTC in all, grouped into 2,007 five-minute bars. Takers bought 7,595.23 BTC and sold 8,328.92 BTC, so delta for the week was -733.69 BTC, while price fell from $80,772.00 to $77,437.00 (-4.13%).
Over whole days, flow and price agreed every time. On all 7 days the sign of the day’s delta matched the direction of the day’s price change. The heaviest selling day, 29 October, had a delta of -501.67 BTC and a 4.29% fall. But 26 October shows the other side: a delta of -408.27 BTC, and price only 0.06% lower by midnight. Selling that large barely moved price, the same absorption story as the worked example, at the scale of a day.
Inside a bar, delta describes the bar. In 1,243 of 2,001 bars (62.12%) price moved the same way as that bar’s delta, and the correlation between a bar’s delta and its return was 0.20. That is a description of the same five minutes, not a forecast.
Delta did not point to the next bar. Across 1,998 pairs of back-to-back bars, the next bar’s close moved in the direction of this bar’s delta 994 times, 49.75%. Taking only the 10% of bars with the biggest delta (199 bars, each 10.33 BTC or more either way) gave 100 of 199, 50.25%. Both are a coin flip, and the gap from 50% is well within chance.
One more count, from our study of 24,971 trading videos: 110 have order flow in the title, with a median length of 19.9 minutes, against 8.5 for average true range (ATR) and 19.6 for market structure.
Absorption, the reading worth keeping
Absorption is heavy flow one way that fails to move price. The worked example and 26 October are both cases: sellers kept hitting bids, and resting buyers kept filling them.
You can see the outline of it with ordinary volume: a bar with heavy volume and a small range. What only real order flow adds is the side, which tells you whether buyers or sellers were the ones spending effort for nothing. That is also why the candle shortcut is weakest exactly here: a bar that absorbed heavy selling and closed up is counted as buying.
Absorption is a description, not a level. It says someone was willing to take the other side at a price. How much more they are willing to take is never on the screen, and the order book page shows how quickly large resting orders get moved.
Cumulative delta and divergence
The reading most courses teach is divergence: price makes a higher high while cumulative delta makes a lower one, so the second push is said to have had less buying behind it.
This week shows the problem with treating it as a signal. Cumulative delta sat below zero from the first hour to the last, reaching -1,066.52 BTC at its low, through two up days (27 and 28 October) and a rise on the 31st. Cumulative delta did rise on those days, but from so far below zero that every rally carried a lower delta than the week’s start, which a divergence reading would flag each time. The next-bar test above found no lean in the direction delta pointed. It is the same kind of claim as relative strength index (RSI) divergence, and the delta divergence page covers the pattern in more detail.
When it fails
The data is not the whole market
These are trades on one exchange. Bitcoin also trades on dozens of others, and a stock trades on many venues plus off-exchange. Delta from one venue is a sample of the flow, not all of it.
The side label is a convention
Exchanges that publish the taker’s side settle it for you. Where they do not, tools infer the side by comparing each trade’s price with the last bid and offer, and trades inside the spread or in a fast market get guessed. The candle shortcut is cruder again, as the 37.88% above shows.
Big takers split their orders
A large buyer working an order with limit orders shows up as a maker, not a taker. The flow a large player most wants hidden is the flow delta counts least. Hidden and iceberg orders add to this.
It is costly and fast
Real order flow needs trade-level data and a platform that draws it, and the timeframe it suits is where the spread and fees take the largest share of each move, as the scalping page measures.
The word imbalance means two things
In order flow, an imbalance is one side outnumbering the other at a price on a footprint. On a price chart, an imbalance usually means a three-candle gap. Check which one a source means before you compare numbers.
Questions people ask
What is order flow in simple terms?
It is the list of trades sorted by which side paid to trade at once. If more size was bought at the offer than sold at the bid, flow was net buying, whatever the candle color says.
Does order flow trading actually work?
On this week of BTC/USD, delta described each bar and each day well, and pointed the next 5-minute bar’s way 49.75% of the time, a coin flip. That is one week on one exchange; it does not settle the question for every market, and it does show that delta alone did not lead price here.
Can you see order flow on TradingView?
TradingView’s own blog lists the volume footprint chart type, which shows delta under each bar, for Premium and higher plans (read 1 Nov 2026). It is built from the data feed TradingView has for each market, so check whether that feed carries the trade side before trusting the split.
What is cumulative delta?
The running total of delta from a chosen start. Rising means net taker buying since the start; falling means net taker selling.
What time frame is best for order flow?
It is built for the shortest charts, from seconds to minutes. Over whole days in this week it only ever agreed with price, so on longer charts it adds little that the close does not already show.
Related
Volume is the undivided number, and most of the question order flow asks can be asked of it first.
Volume profile slices the same trades by price instead of by side.
Liquidity is what absorption is made of: the resting orders that take the other side of a push. The tape is where each trade first appears.
I do not trade this, and I want to be straight about why: doing it properly needs a data feed and a platform I do not run, and doing it improperly is worse than not doing it. What I take from the idea is the question rather than the tool, which is whether a new high came with more buying or drifted up on nothing. Ordinary volume answers most of that.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.