Why Football Heatmaps Change When the Grid Changes: A RubiScore Guide to Spatial Binning

Spatial binning groups football actions into areas of the pitch before counting or summarising them. It turns many individual coordinates into a readable map, but the boundaries influence the picture. For readers using RubiScore alongside visual analysis, that makes the grid part of the explanation, rather than a neutral background.

Match context at https://rubiscore.com establishes which teams, players and fixtures are under discussion. A heatmap adds a different layer: a set of decisions about how those actions are grouped. Two maps can use exactly the same events and still emphasise different patterns.

This RubiScore guide focuses on that transformation. The question is not whether the underlying passes or touches happened. It is whether the visual conclusion survives a reasonable change in the areas used to summarise them.

A useful distinction for a RubiScore reader is between a football boundary and an analytical boundary. The penalty area is marked on the pitch. A rectangular heatmap cell may exist only because the chart designer selected a convenient grid. Both can organise data, but they do not carry the same meaning.

What happens when an event becomes part of a cell?

Imagine placing every recorded pass origin on a pitch diagram. Each event has a location. A binned map replaces those individual locations with a summary for each area: perhaps the number of passes starting there, the proportion completed or the total value assigned to them.

The operation deliberately removes detail. Actions near opposite edges of one cell contribute to the same summary, while actions almost touching across a boundary contribute to different cells. That loss is useful when it makes a crowded picture readable. It becomes a problem when readers forget it happened.

The mplsoccer documentation provides a concrete implementation example: its heatmap workflow calculates a statistic over selected pitch bins before drawing the coloured cells. The grid and statistic are specified choices. This is a demonstration of how such charts can be constructed, not a description of any platform's internal visualisation system. mplsoccer: heatmap example.

Why do larger cells look more stable?

A larger cell pools more territory. With the same event sample, it will generally collect more actions than a smaller cell covering part of that territory. Local fluctuations can disappear inside the pooled total.

Suppose an illustrative winger receives in two narrow strips: near the touchline during build-up and farther inside near the attacking area. A coarse grid may combine those positions into one broad band. The map then suggests a general wide role while concealing the movement between distinct receiving locations.

That does not make the coarse map wrong. It answers a broader question. It might be suitable for comparing involvement on the left and right sides, while being unsuitable for explaining the exact route taken around a defender.

The reverse problem appears with very small cells. A few actions can create isolated patches that look tactically significant. The picture has more detail, but the evidence supporting each patch is thinner. Visual precision and analytical reliability are different properties.

Can moving a boundary change the apparent hotspot?

Yes. Cell size is only one decision. The placement of the boundaries also affects which events are pooled together. A cluster of nearby actions can be divided by one grid and contained within a single cell by another.

Consider a hypothetical set of receptions around a half-space. If the grid boundary cuts through the cluster, two adjacent cells may each look moderately active. Shift the boundary and the same receptions may create one prominent hotspot. No player has changed position; only the grouping has changed.

This is especially relevant when a claim depends on a single coloured square. The analyst should inspect the underlying locations or repeat the chart with another reasonable partition. A pattern that remains visible under both treatments has a stronger descriptive basis.

A pattern that disappears is not necessarily meaningless. It may be genuinely local. But the conclusion should then describe that local evidence, rather than treating one arbitrary cell as a stable tactical zone.

Why must the statistic be named as well as the grid?

Identical cells can display very different quantities. A count map rewards volume. A completion-rate map describes successful actions relative to attempts. A value map aggregates a model's evaluation of actions. Their colours answer different questions.

A busy passing area may have a high count because the team circulates the ball there repeatedly. A nearby area with fewer passes may contain the more difficult attempts. Comparing the colour alone cannot establish which area contributed more to progression.

A rate also needs its denominator. A cell containing very few attempts can show a perfect completion rate without providing much evidence about repeatable performance. A count beside the rate, or a rule for withholding sparse cells, can make that limitation visible.

Keep these elements together when reading a map:

• The events included in the analysis.
• The coordinate system and attacking direction.
• The cell boundaries and size.
• The statistic calculated within each cell.
• The colour scale and treatment of empty areas.
• The match range and amount of playing time.

Does a percentage map solve the comparison problem?

A percentage can align one aspect of the comparison while changing the question. Expressing each cell as a share of a player's actions describes distribution. It does not describe how many actions the player supplied.

Two players can have similar percentage maps despite very different involvement totals. Conversely, a player can increase activity in every area while showing a smaller percentage in one area because another area grew faster. The percentage describes the composition of the whole.

When using RubiScore to establish the players and matches behind a comparison, keep volume and distribution in separate sentences. One sentence can describe where actions were concentrated; another can describe how often the player participated.

Area size matters too. Comparing raw counts across unequal zones partly compares the amount of pitch each zone contains. If the purpose is to compare spatial density, the area denominator needs attention. If the zones represent distinct tactical functions, a different interpretation may be appropriate.

What other visual choices can imitate a tactical change?

Colour scales are a common source of confusion. If each chart chooses its own maximum, the darkest cell in a quiet performance can look as intense as the darkest cell in a busy one. A shared scale helps comparison, although it can make subtle patterns less visible.

Smoothing introduces another transformation. It spreads local information across nearby locations to create a continuous-looking surface. That can help readers see broad concentration, but it should not be confused with a literal record of every point occupied by a player.

Event maps also differ from tracking maps. A map of touches records where touches occurred. It does not directly record all the positions occupied between touches. A player who holds an important off-ball position can be almost absent from that event picture.

These distinctions should be resolved before discussing formation. Otherwise, a chart showing the geography of recorded involvement can be mistaken for a complete account of the team's shape.

How can readers run a simple sensitivity check?

Begin with the football claim, such as whether a player receives mainly outside the opposing midfield. Keep the event sample fixed. Then examine a coarser grid, a finer grid and the individual locations if they are available.

Next, ask whether the conclusion depends on one cell edge. If it does, describe the actual locations more carefully or use a zone whose football meaning matches the question. Do not keep changing boundaries until the preferred conclusion appears.

Finally, repeat the comparison using the same settings for every player or match. Differences in chart construction should not be allowed to masquerade as differences in behaviour. Any necessary exception should be explained beside the comparison.

RubiScore can supply the broader match context for this reading exercise. The grid check belongs to the analyst's treatment of the visual evidence. It is a practical way to test a description, not a demand that every chart use one universal layout.

What conclusion is safe to carry away?

A heatmap is a summary with a spatial design. Larger cells trade local detail for broader grouping; smaller cells expose detail while relying on fewer observations. Moving a boundary can change the prominence of a pattern without changing a single event.

The best grid is therefore the one suited to a clearly stated question and supported by enough relevant observations. A map of broad territorial involvement and a map of receiving pockets need not use the same partition.

Before turning a coloured patch into a tactical claim, check what was counted, how it was grouped and whether the pattern survives another reasonable view. The football is in the actions. The grid determines which relationships the picture makes easiest to see.