Order flow delta

2. How it is computed

Properly, it requires trade-by-trade data with the aggressor flagged: every print tagged as buy-initiated or sell-initiated. Futures exchanges provide this.

Most charting platforms do not. They estimate instead, usually one of two ways:

From the candle's shape. Volume is split between the two sides according to where the close sits inside the bar's range. A bar that closes near its high is counted as mostly buying.

From lower-timeframe bars. The bar is broken into smaller bars, and each one's volume is assigned to whichever direction it closed.

The second is better. Neither reads the order book.

3. What it rests on

That the side which initiates is the side with conviction, and that conviction precedes movement.

The first half is reasonable: crossing the spread costs money, so it signals urgency. The second half is the leap. Urgency is not the same as being right, and a market can absorb a great deal of urgent buying without moving at all.

4. Where it is wrong

The estimate is weakest exactly where the concept would be most useful.

At absorption — heavy selling into a level that does not fall — the real delta is strongly negative while price holds. A shape-based estimate reads those bars as buying, because they close high. It gets the sign backwards at the one moment the number would have told you something.

This is not hypothetical. On one matched window a candle-shape estimate read plus 12,110 while the real aggressor delta over the same bars was minus 20. That measurement is published on this site.

Beyond that: lower-timeframe history is served for a limited stretch only. Scroll back far enough and the estimate silently reads zero, which looks identical to a bar with no trading.

5. Is there a measurement

Closed, and it refuted the claim. The candle-shape estimate is not a usable substitute for aggressor data — not merely imprecise, but opposite in sign in the case measured.