Footprint: inside the candle

A candle shows four prices: open, high, low and close. How much was traded between them, and where, it does not say: a long green candle may have crossed the whole range on one large buy, or it may have sat near the low for half an hour and shot up only in the last minute.

A footprint (cluster chart) opens the candle up. The bar's range is cut into price levels with an equal step, and each level shows two numbers separately: market sells × market buys. On the left — how much was sold at market at that level (hitting the bid), on the right — how much was bought (lifting the ask). Each such “bar × level” cell is called a cluster.

Everything you already know from earlier sections is built from the cells. Both sides summed over all levels is the bar's volume. Total buys minus total sells is the bar's delta (section 7). Volume by level is a small volume profile of one bar (section 6), just split by side.

The footprint in EVA. The «Clusters (footprint)» layer in the «Layers» panel — “buys and sells inside each candle by price”. The cluster chart is a «Pro» mode tool (the switch at the top of the «Layers» panel). Chart view: over candles (candle bodies dimmed) or clusters only (candles hidden, each bar a column of cells). Style: “cells”, “histogram”, “cells + numbers”. Cell colouring is your choice: by delta (side colour, density — the level's skew), by volume (one colour, density — the level's turnover) or by imbalance. The bar's POC can be outlined with a sand-coloured frame.

Important caveats from the help. Values are in dollars and are counted across all venues, spot together with derivatives. Candles come from one exchange and are counted in coins — so footprint volume does not match the candle's volume bar, and the two cannot be reconciled directly. The level step is chosen for the coin and timeframe; you can make it finer or coarser in the settings. Recent hours are held in memory minute by minute; older data comes from files.

The candle's POC (point of control) is the level inside the bar where the most volume traded, buys and sells together. It is the same POC as the volume profile's from section 6, just for one bar. Do not confuse it with the close (the bar's last price), with the volume-weighted average price (VWAP is an average, not a level) or with the level of the largest delta (the POC is by volume, not by the difference between sides).

Where the POC sits inside the bar is a description in itself. POC at the low of a green candle: most trading happened at the bottom, and the price went up on little volume. POC at the high: the price got to the top and traded a lot there. POC in the middle: the bar spent most of its time mid-range, and the extremes were short probes.

Volume clusters are neighbouring cells where volume is noticeably higher than in the rest of the bar. These are the places where the bar “spent” its money: market and limit orders met there. The thin cells between them are prices crossed quickly because there were few counter-orders. With “by volume” colouring such concentrations show up as dense colour at once, without numbers.

Combinations worth being able to read.

• Large market sells at the bottom of the bar, a negative bar delta, yet the bar closed near its high. At the bottom market sells met limit buys — the absorption from section 8, only here you see it at a specific level — and then the price was lifted.
• A level where buys are three times the sells while the price stands still on it — that often looks like buying being absorbed by a limit seller, and only the footprint shows it.
• A wide bar with volume spread evenly — the price went through the book without a clear place of resistance.

Each such reading is a story about what happened inside the bar. None of them says what will happen in the next one.

The best bid sits below the best ask — for a liquid coin usually exactly one step (section 3). A market buy takes the ask — so it prints on the upper of the two levels; a market sell hits the bid — it prints on the lower one. So at the same instant buys and sells land on neighbouring levels of the footprint, not on one.

Hence the diagonal comparison rule: buys at a level are compared with sells one level below. If the buys are several times larger, the level has a buy imbalance: buyers were lifting asks far more actively than sellers were hitting bids one step lower. The mirror: sells at a level against buys one level above — a sell imbalance. Comparing the two numbers of one row (sells and buys at the same level) would mix two different moments of the book.

Let us count with the same illustrative numbers (sells × buys: 102 — 3 × 1, 101 — 12 × 40, 100 — 30 × 95, 99 — 24 × 110, 98 — 25 × 8). The threshold in the example is 3×.

• Buys at 100 vs sells at 99: 95 / 24 = 3.96 — a buy imbalance.
• Buys at 99 vs sells at 98: 110 / 25 = 4.40 — a buy imbalance.
• Buys at 101 vs sells at 100: 40 / 30 = 1.33 — no.
• Buys at 102 vs sells at 101: 1 / 12 = 0.08 — no.
• Sells at 98 vs buys at 99: 25 / 110 = 0.23 — no sell imbalance; sells at 101 vs buys at 102: 12 / 1 = 12 — formally a sell imbalance, but there is only 1 BTC of buys at 102.

The last item shows why a minimum level volume is needed: 12 against 1 is “twelve times”, but on tiny volumes, and against an empty level any number gives an infinite ratio. So levels with volume below the minimum are not counted as imbalances.

A stacked imbalance is several same-side imbalances on neighbouring levels in a row. One imbalance is common; a series means that over a whole stretch of price one side took almost everything resting in the book. In the example above the imbalances at 99 and 100 are two in a row; if a series requires three or more, there is none there.

In EVA this is a separate «Stacked imbalance» layer: it highlights the bar's levels where there are N or more buy or sell imbalances in a row. The ratio (“imbalance from K%”), the number of levels in a row and the minimum level volume are chosen by you in the layer settings, along with the side colours. Optionally the layer keeps a zone extending right until the price comes back into it (first leaving the zone, then touching it). The step used for counting is the data step, the same at any zoom; bars before the fine recording began are not counted, and the boundary is labelled. Under its settings the layer itself says: “This is a fact from the trade tape, not a signal”.

Why it is not a signal follows from the mechanics: one large market order walking several book levels produces imbalances on each of them by itself — that is, a series. A series describes flow, not the number of participants or their plans.

An honest count over a week. The same rule on all 15-minute BTC bars from 07.10 to 13.10.2026 inclusive: 664 bars with a known price. For comparison: of all bars 329 closed above the open and 335 below — nearly equal.

• Buy series only — 39 bars. The bar itself closed up in 37, down in 2. The next bar: up 25, down 14.
• Sell series only — 58 bars. The bar itself closed down in 53, up in 5. The next bar: down 30 (51.7%), up 28.
• No bar had both series at once; 567 bars had no series.

What follows. A series nearly always points the way its own bar closed (37 of 39 and 53 of 58): it is born while the price is being moved and describes that move. About the next bar it says little: after sell series — 30 to 28, a coin flip. After buy series — 25 to 14, but that is 39 cases in one week on one coin: such a skew easily arises by chance, and in another week it may vanish or flip. To speak of it seriously, it would need checking on other weeks and coins. We have not done that — so it cannot be relied on.

What the footprint does not show.

• Who traded and why. One order walking five levels and five different participants look the same.
• Opening or closing. These are still trades, not positions: a market buy can be a new long or a short being closed.
• Limit orders that were not filled. Pulled and resting orders do not get into the cells — you see them in the book and on the heatmap, not here.
• The venue. A cell adds up all venues, spot and derivatives together; there is no breakdown by exchange in it.
• The order within the bar. A cell is a sum over the whole bar: whether the volume at a level went through at the start or at the end cannot be seen from it.
• The future. Neither the POC, nor imbalances, nor series promise that the price will return to a level, bounce off it or pass straight through.