Turning a terrace argument into a question we can check

One complaint can contain several claims

A fan might say a player does little defending, holds the ball too long, fails to break lines, cannot shoot and gets pushed off it. This is a hypothetical example of a terrace argument. It is not a quotation attributed to an identifiable supporter or a verified assessment of a named player.

Those are five claims, and they need five separate definitions. A verdict on one should not silently become a verdict on the rest. The aim is to make the question precise enough that another reader can see what evidence would support it, and what evidence is still missing.

1. “Does little defending”

Start by naming the action: a tackle, an interception, a recovery or another defined defensive event. Then name a relevant comparison and time period. Playing time, role, opponent and how much defending the team had to do can affect the context.

A recorded count of actions does not describe every defensive contribution. Positioning, pressing instructions and covering a passing lane may not appear in the same data. A narrow answer about recorded actions must not become “the player does not work hard”.

2. “Holds the ball too long”

This needs evidence about possession duration or the relevant decisions. An event count alone may not tell us how long someone had the ball, which team-mates were available, or whether delaying was sensible.

A carry count is not a direct measurement of excessive delay. If the necessary timing or decision evidence is absent, the answer is unavailable. Replacing the missing measurement with an easier number would change the question without telling the reader.

3. “Fails to break lines”

First define what counts as breaking a line and what evidence identifies the defensive line at that moment. A forward pass, a progressive pass and a pass through an opposition line can overlap, but they are not interchangeable definitions.

If the available source does not support the definition, say so. A lack of recorded line-breaking passes cannot be treated as a verified zero when the source does not measure them. Nor can a collection of other passing statistics establish that a player is incapable of the action.

4. “Cannot shoot”

Separate shot frequency, accuracy and chance quality. They answer different questions. A low goal total by itself does not settle all three, and a chance of scoring in one match is not a general judgment about technical ability.

A checkable answer states the outcome being assessed, the opportunities included and the period covered. It also keeps observed results separate from a prediction about a future appearance. Neither should be rewritten as a claim about effort or character.

5. “Gets pushed off it”

Specify the relevant contest or loss of possession. Different sources can define duels and dispossessions differently. Without knowing the definition and whether the attempts are recorded completely, a success percentage can give a misleading impression.

A physical contest recorded in match data is not a measurement of strength, fitness or a medical condition. Even a supported answer should remain within the event definition used.

From a defined question to a REDS answer

A predictive answer needs a defined outcome to learn, inputs that would have been available before the match, and an evaluation that respects time order. A useful comparison with a simple alternative is part of that check. Where no usable trained predictor exists, REDS should say it cannot answer the item.

An observed fact can still be shown as a fact. It should not acquire a predictive verdict simply because a percentage can be calculated. This column explains the question-setting process; it does not build a new predictor or publish a fresh player assessment.

Read confidence with its date and scope

A confidence label belongs to a particular question and version. It does not mean “this percentage of the player’s ability”, and it cannot be carried unchanged into another role, season or definition. The applicable threshold and evaluation need evidence for that particular answer.

The previous column quoted dated percentages and a general 70% threshold without giving readers a verifiable record. Those numbers have been withdrawn. We do not assert a replacement threshold or a current verdict here. Open the relevant REDS answer to see what it can support now; an unavailable or uncertain item should remain visible as such.

Disagree with the claim, not the person

A useful challenge names the definition, the comparison, the period or the source that appears wrong. “The recovery count excludes this action” is a question we can investigate. A broad allegation about someone’s attitude cannot be established from a chart of events.

When sharing an answer, keep its question, date, conditions and uncertainty together. That gives the next reader a fair chance to understand what was actually assessed.