Order flow and data quality · 4 / 5
Read diagonal footprint imbalances with a sample threshold
A footprint arranges classified volume by price inside a bar. Highlighted cells can make an imbalance look definitive, but the calculation depends on the pairing rule, price grouping, classification method and minimum amount of data.
Athenum8 minUpdated:
Understand which cells are compared
In a common diagonal convention, buyer-initiated volume at a price is compared with seller-initiated volume one price row below. The reverse comparison pairs seller volume with buyer volume one row above. This differs from subtracting the two values on the same row. Confirm the convention rather than inferring it from colours.
A threshold of 3× means the larger tested quantity must be at least three times the comparison quantity under that rule. It is not the same as a 300% increase, which would imply four times the starting value. Write the ratio directly when configuring or documenting a study to avoid ambiguous percentage language.
Check how volume was classified
A trade-level footprint can use exchange-reported taker sides. Other implementations estimate buy/sell volume from lower-timeframe price movement. TradingView documents such an intrabar direction-based method and changing historical granularity. An estimated footprint and a taker-classified footprint can legitimately disagree; neither should be described as the other.
The bucket size also changes which executions are grouped together. Coarser rows can hide local imbalances; finer rows can produce extreme ratios from tiny samples. Use a minimum combined volume or another explicitly defined sample filter, and test whether the observation survives reasonable grouping changes.
160 divided by 40 is a 4× diagonal imbalance
In the hypothetical table, compare buy volume of 160 at 100.50 with sell volume of 40 at 100.00. The ratio is 4× and the paired volume is 200 units. It passes an illustrative 3× threshold and a 50-unit minimum combined-volume filter.
The 60 buy units at 100.00 face zero sell units at 99.50 in the next diagonal comparison. Division by zero is undefined; do not silently turn it into the strongest possible signal. Choose and document a handling rule such as flagging it separately or requiring a nonzero comparison cell. That rule is part of the method you must evaluate.
| Price | Sell volume | Buy volume |
|---|---|---|
| 101.00 | 0 | 12 |
| 100.50 | 20 | 160 |
| 100.00 | 40 | 60 |
| 99.50 | 0 | 30 |
- Buy at 100.50
- 160 units
- Sell at 100.00
- 40 units
A large ratio can describe almost no trading
Four units against one also gives a 4× ratio, but only five units traded in the pair. Treating that as equivalent to 160 against 40 ignores scale. Consecutive highlighted rows are similarly not independent evidence when they come from the same short burst. Evaluate the sequence and subsequent price response, not the number of coloured cells.
Before acting
- Identify diagonal or same-row comparisons.
- Write the ratio threshold unambiguously.
- Set a minimum sample and explicit zero-denominator rule.
- Check trade-side versus intrabar-estimated classification.
- Save row size and historical granularity with the example.
Check your understanding
A diagonal pair contains 90 buy units and 20 sell units. Does it pass a 3× ratio and 150-unit combined-volume requirement?
Show the explained answer
The ratio is 90 ÷ 20 = 4.5×, so it passes the ratio test. Combined volume is only 110 units, so it fails the sample-size condition. Both conditions were part of the declared method; the bright ratio alone is insufficient.