What Is Order Flow Trading?
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.
The most technical-sounding thing on this site, and the one where the honest caveat has to come first rather than last.
How it works
Ordinary volume tells you how much traded. It does not tell you which side was willing to pay up to get filled.
Order flow splits it. Every trade happens between someone resting an order and someone crossing the spread to hit it. Volume that lifts the offer is counted as buying; volume that hits the bid is counted as selling. Delta is the difference.
The caveat, up front
The charts on this page are not real order flow. They use candle-direction delta — a bar’s whole volume counted as buying if it closed up and as selling if it closed down.
Real order flow splits each individual trade. That needs tick data with bid and offer, a feed that carries it, and a platform that renders it. This site does not have one, so the approximation is labelled rather than dressed up.
The distinction matters practically. A bar that closed up after heavy selling absorbed at a level is exactly the case order flow exists to catch, and candle-direction delta records it as pure buying. The approximation is blind to the most interesting thing the real tool sees.
Cumulative delta
Per-bar delta is noisy. The running total is what people actually read.
It gives you a second line with the same shape-language as price — highs, lows, and a slope — which is what makes the next section possible.
The reading people are after
Price made a higher high — 100.77 then 101.67. Cumulative delta made a lower one — 0.4 then −0.8.
The story attached to that: the second push cost less buying than the first, so the move up was weaker than it looked.
The honest status of the story: it is a plausible interpretation of two lines disagreeing, and it is the same kind of claim as relative strength index (RSI) divergence. Divergence is a description of the past that has to resolve before you know whether it meant anything.
Absorption
This is the idea worth keeping, and you can see it without any special data.
A lot of volume and very little movement means someone was taking the other side. Buying that would normally move price ten ticks moved it two, because there was enough resting supply to absorb it.
Volume high, range small. That is two numbers on any chart, and it is most of what order flow is being used to detect.
What a footprint chart is
The display most people mean when they say “an order flow chart”, and worth defining because the name gives nothing away.
A footprint chart draws each candle as a grid rather than a body. Every price level the bar traded at gets a row, and each row shows two numbers: volume that traded at the bid and volume that traded at the offer.
So one candle becomes a small table. Instead of “this bar closed up on heavy volume” you get “most of the volume happened in the top third, and it was two to one in favour of buyers there.”
That is a genuinely different object from anything else on this site, and it is also why the data requirement is not negotiable — a footprint built from candle direction would have one row and would be a candle.
The practical note: it is dense, it is designed for one instrument watched closely, and it does not scan. Nobody reviews four hundred symbols in footprint.
A worked example
Ask the question first, with ordinary volume. Price made a new high — was volume above or below average on the push?
Look for absorption at the level. Heavy bar, small range, near an old high: that is somewhere liquidity is resting.
Then, if you have the data, check the flow. Cumulative delta agreeing with the new high supports it; disagreeing is a reason for caution, not a reason to be short.
And the invalidation is still a price, from the chart, not from the panel. Nothing in the flow gives you a level.
The original data
Across our study of 24,971 trading videos, 139 cover order flow. The median one gets 20,510 views, 68% never pass 50,000, and the median length is 19.2 minutes.
19.2 minutes is among the longest medians measured here — against 8.5 for average true range (ATR) and 12.4 for scalping, and behind only market structure at 19.5 among topics with a real field.
And 20,510 is among the highest medians here despite the smallest practical audience: this is the one topic on the site that most viewers cannot act on without paying for data.
The corpus carries description text for 52 of those 139, and across those 52, two mention invalidation, failure, or what a bad read looks like.
When it fails
Flow and price disagree, and price wins
Price rose 0.90 across this chart with cumulative delta finishing negative.
Delta is not a leading indicator of price — it is a second measurement of the same period, and markets rise on net selling more often than the divergence story admits.
The approximation hides the interesting case
Covered above and worth repeating: candle-direction delta cannot see absorption, which is the main thing real order flow is used for. If you are running the approximation, you are running the version without the feature.
It is the shortest timeframe there is
Order flow is trade-by-trade data, which puts it at the fastest end of the scalping page’s cost table — where a round trip is 4.1% of a typical bar. The data is expensive and the style it suits is the one where costs bite hardest.
You read the divergence after it resolved
Divergences that preceded a turn are obvious afterwards. Divergences that preceded a continuation look identical at the time, and there are more of the second kind.
Related
Volume is the undivided number, and most of the question order flow asks can be asked with it.
Volume profile is the other way to slice the same data — by price rather than by side.
And liquidity is what absorption is made of: the resting orders that take the other side of a push.
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 - when price makes a new high, was there actually more buying, or did it drift up on nothing? You can ask that question with ordinary volume and get most of the value.
— Michael Whitman, from this video
This page is educational, not financial advice. Test every idea on your own charts before risking money.