Settings

Language, shortcuts & timezone

Language

Timezone

Time zone changes apply on this device. A new device may start at GMT+8; check match times before reading.

本文へスキップ
紹介

設定

設定をカスタマイズする

期待アシスト:アシスト数を超えたチャンス創出を読み取る

Opta Tips Team による 更新済み

シュートベースのxA(予期されたアシスト)を理解し、チャンス創出とフィニッシュを分けて考え、選手を比較する前にボリューム、出場時間、およびデータ提供元の定義を確認してください。

Hypothetical shot-based example adding 0.10, 0.20 and 0.30 to total 0.60 expected assists; OptaTips logo at lower right.

Expected assists estimate the value of the chances a player creates under a particular data model. They help separate the opportunity supplied to a team-mate from whether that team-mate actually scores. Before comparing players, however, check what the provider means by xA: a familiar abbreviation does not establish a common model.

A player can finish a match without an assist despite setting up strong chances. Another can receive an assist after a difficult shot is converted. Neither observation alone establishes which player created more useful opportunities, and xA does not turn either performance into a guarantee about the next game.

Identify whether the model starts with a shot or a pass

Wyscout’s glossary defines a pass’s expected-assist value through the xG of the shot it leads to. Under that shot-based convention, the creator receives the modelled value attached to the resulting attempt.

That is the convention used for the numerical examples here. Other products may evaluate passes differently, including approaches that assign probabilities directly to passes. Do not combine their totals without documentation showing compatible definitions.

For a concrete contrast, Stats Perform’s explanation describes an expected-assist value for every completed pass in its detailed data. That population differs from passes restricted to recorded shot assists. The distinction is a reason to check definitions, not to declare one provider’s total wrong because it differs from another’s.

Record the provider, model description, event coverage and period before calculating anything. Check which actions qualify as shot assists, how set pieces are treated, and whether historical values are revised when the underlying model changes.

The definition also tells you what is outside the column. A pass that opens a defence but does not lead to a recorded shot need not receive shot-based xA. That absence does not prove the pass was tactically worthless. It means the selected metric observes a narrower event chain.

Follow three invented chances through the calculation

Suppose a player supplies three qualifying passes leading to shots valued at 0.10,0.20 and 0.30 xG. Under the stated shot-based convention, the player’s total is 0.60 xA. These are hypothetical values, not results from an actual fixture or OptaTips record.

If none of the shots is scored, the player can finish with 0 actual assists and 0.60 xA. If two are converted, the actual assist count may be two, subject to the provider’s assist rules, while the supplied chances retain the same combined modelled value.

The calculation is not predicting that this particular set of three shots must produce 0.60 goals. It aggregates estimated chance values. Actual outcomes are discrete and can differ considerably over a small sample.

Nor should the 0.60 total be read automatically as a 60% probability that the player records at least one assist. That question requires a different calculation and appropriate assumptions about events and assist credit.

Even if the three hypothetical shot outcomes were independent, the probability that at least one is converted would be 1−0.90×0.80×0.70, or 49.6%, rather than 60%. Independence may be inappropriate for a real sequence; the arithmetic only demonstrates why a sum is not the same as a probability of at least one success.

Separate chance supply from finishing

Actual assists depend on an event being converted and credited. A creator’s xA and assist count therefore answer different questions. The first concerns the modelled opportunity supplied under the chosen convention; the second concerns recorded outcomes.

For an invented comparison, Player A and Player B each generate 0.60 xA from compatible events. A’s team-mates score twice, while B’s score none. Their assist totals differ, but those totals do not show that A supplied a greater modelled amount of chance value.

Do not then reverse the mistake by declaring the two performances identical in every respect. They may involve different pass types, pressure, locations and choices. Equal summary totals can conceal different processes.

The expected-goals review guide explains the distinction between chance quality and the realised score. Reading creator statistics requires the same discipline about what a model estimates and what an outcome records.

Keep opportunity volume beside the total

A 0.60 total from three qualifying passes differs from 0.60 concentrated in one large chance. The aggregate is the same; the distribution is not. That distinction can matter when investigating whether a player repeatedly reaches useful situations or benefits from one unusual moment.

Show the number of qualifying events beside the total where the provider makes it available. Under the stated convention, xA divided by shot assists gives average modelled shot value per qualifying creation event. It is not average value per every pass attempted.

A player who supplies three shot assists worth 0.60 has an average of 0.20 per qualifying event. Dividing 0.60 by 50 total passes instead gives 0.012 per attempted pass, a different denominator answering a different question.

Neither calculation should silently inherit the other’s label. If an article claims that a player creates high-quality chances repeatedly, inspect both the event count and the values instead of presenting the total as complete proof.

The xG-per-shot explanation develops a comparable distinction between volume, averages and concentration. The unit of observation must remain visible.

Playing time changes the comparison

Season totals favour players who have had more minutes in which to accumulate events. A per 90 rate can help compare exposure, but it does not establish that every minute offered the same opportunities.

For example, an invented 2.40 xA in 900 minutes is 0.24 per 90. Another player with 1.20 in 300 minutes has 0.36 per 90. The second has the higher rate but the smaller underlying sample and lower total.

Report all three fields rather than selecting whichever supports the desired ranking. Substitute appearances against tired opponents, roles after a dismissal and set-piece responsibilities may differ from regular starting minutes.

Do not treat the per 90 calculation as a literal forecast for the next 90 minutes. It standardises an observed rate; a prediction requires information about likely participation, role, team-mates and the opponent.

When aggregating across matches, calculate from compatible totals and minutes if the question concerns the overall rate. An unweighted average of match rates can give a short cameo the same influence as a full game.

Check role and phase before assigning credit

A corner taker can accumulate chances through deliveries that another player never attempts. A central midfielder may make earlier passes that enable an attack without supplying the final ball. A winger may draw defenders and create space without receiving xA credit for that movement.

Those differences make role context essential. Separate open-play and set-piece events when that division serves the question and the data support it. Do not describe a combined total as open-play production simply because the player often operates in open play.

Keep actual assist definitions separate as well. Deflections, rebounds and other interventions can be credited differently across data products. A comparison of actual assists and expected assists is only as coherent as the event definitions underneath it.

Video or event sequences can show what the summary leaves out. Did a pass arrive cleanly into a strong shooting position, or did a receiver create most of the opportunity through a later individual action? The selected model’s attribution convention should be explained rather than confused with complete causal credit.

Use uncertainty honestly in short samples

One match can be dominated by a single high-value chance. A few more or fewer finishing outcomes can also produce a large gap between assists and xA. That gap is descriptive evidence to investigate, not proof that a player is owed future assists.

Extend the sample only while preserving compatible coverage and a relevant role. Combining several seasons across different positions might create a larger number without answering the question about the player’s current task.

A provider change is another boundary. Keep the earlier series separate until you know the models and event definitions are comparable. A shared unit label does not supply a valid conversion between systems.

Make a claim another reader can reproduce

A useful creator review names the provider and period, states the xA convention, and reports minutes, qualifying events, total xA and actual assists. It then explains any important role or phase differences.

For the opening illustration, a defensible summary is that three qualifying creations produced 0.60 shot-based xA regardless of whether the attempts were converted. It should not claim a guaranteed assist, a 60% scoring-related probability or complete measurement of all creative work.

That narrower statement is valuable because a reader can check it. It leaves room for the rest of the football rather than asking one column to explain every action, every outcome and the next match.

For adults aged 18 and over. All numerical examples are hypothetical. References explain provider definitions, not a data-supply relationship with OptaTips or a promise of future performance.

読者による議論

コメント 0

まだコメントはありません。

コメントを残す

Your email address will not be published. Required fields are marked *