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

ノンペナルティxG:除外される要素と残される要素

Opta Tips Team による 更新済み

ノンペナルティxG(期待得点)が何を除外しているか、なぜセットプレーは依然として含まれるのか、そして説明と予測とを混同することなくチャンス創出を比較する方法について学びましょう。

Generated illustration of an attacker shooting in open play with a nearby defender and goalkeeper, with the OptaTips logo at lower right.

A team can finish with the higher expected-goals total because it received a penalty, even when its other chances were less threatening than its opponent’s. Non-penalty xG helps expose that distinction. It removes the expected-goal value of penalty attempts from the total, leaving the modelled quality of the other shots.

That makes it useful for reviewing an attack, but the label needs to be read carefully. Non-penalty does not mean open play only. Corners, free kicks and other non-penalty situations can still contribute. Nor does removing penalties reveal a team’s unquestionable underlying ability.

The sensible task is to separate the sources of chance creation, then decide which comparison answers your question. Keep the original total visible so that a useful adjustment does not become a selective account of the match.

Remove penalty attempts, whether they were scored or missed

Expected goals assigns a modelled probability to an attempt. To construct a non-penalty total, exclude the penalty attempts themselves rather than subtracting the number of penalties scored.

If a fictional penalty has an assigned value of 0.76, a saved attempt still contributes 0.76 to an all-shot xG total under that convention. It contributes nothing to the non-penalty subset. Scoring or missing changes the actual goals, not whether the attempt was a penalty.

Sportmonks’ explanation of non-penalty xG describes the exclusion and distinguishes it from an open-play filter. Provider definitions should govern the calculation rather than an assumption based on the abbreviation.

Wyscout’s xG glossary, for example, documents a fixed penalty value of 0.76 in its model. That does not make 0.76 a universal constant. When working with a different provider or model version, use its own recorded values.

A penalty shootout also needs separate treatment from attempts during the defined match period. Do not append shootout attempts to a 90-minute chance total simply because both involve kicks from the spot.

A small example changes the apparent ranking

Imagine these two teams in a hypothetical completed match. The figures are invented and use the same model, period and penalty convention.

MeasureTeam ATeam B
Total xG2.101.80
Penalty-attempt xG0.760.00
Non-penalty xG1.341.80

Team A leads the all-shot total, but Team B leads the non-penalty comparison. Both statements are accurate. They answer different questions about the same collection of events.

A review saying only that A created more xG would omit an important feature of the advantage. A review saying A’s penalty did not matter would also be incomplete: it was an actual scoring opportunity in the match.

Record the source of the difference. If the purpose is to investigate repeated non-penalty chance creation, B’s figure deserves attention. If the purpose is to explain all scoring opportunities in that fixture, retain the penalty as well.

Neither comparison says which side deserved a compulsory score. The guide to losing despite higher xG explains why chance values and actual results should be read separately.

Separate open play from non-penalty set pieces

Suppose A’s remaining 1.34 consists of 0.96 from open play and 0.38 from non-penalty set pieces. Removing the penalty leaves both components. Describing the whole 1.34 as open-play xG would mislabel the evidence.

That matters when evaluating an apparent improvement. An attack might produce more threatening corners while creating little from settled open play. Another might enter the penalty area more effectively without receiving dangerous free kicks. Those are different mechanisms to investigate.

Use the provider’s phase classifications where they exist. A shot after a cleared corner can raise a boundary question: is it still part of the set piece or a new phase? A review should follow the documented event classification rather than invent a private definition midway through a comparison.

If the source supplies only total and non-penalty xG, say that an open-play breakdown is unavailable. The absence of a column does not justify treating the remaining number as whichever category would support the argument.

This is also a reason to preserve shot-level records when you have lawful access to them. They can help identify what changed, while the headline total alone cannot show the route by which chances were produced.

Read the number of attempts beside the total

Two teams can accumulate similar non-penalty xG through different distributions of shots. More attempts do not automatically mean better average opportunities, and a higher average can be driven by one unusual chance.

Consider a hypothetical four-match sample with total xG values of 2.10, 1.40, 2.00 and 1.20. It contains three penalty attempts, each assigned 0.76 solely for this example. The calculation is:

Total non-penalty xG = 6.70 − (3 × 0.76) = 4.42.

The all-shot average is 1.675 per match, while the non-penalty average is 1.105. That difference deserves explanation before describing the team as consistently generating around 1.7 of non-penalty threat.

If those four matches contained 41 non-penalty attempts, the non-penalty xG per attempt would be 4.42 ÷ 41, approximately 0.108. Divide by the matching shot subset. Using all attempts in the denominator after removing penalties from the numerator would produce a different measure.

Keep the individual match values too. Four matches are a small teaching sample. A mean cannot establish that the same pattern will recur against a different opponent or with a different line-up.

Compare goals and expectation on the same subset

When reviewing non-penalty finishing, compare non-penalty goals with non-penalty xG. A player’s penalty goals should not be used to make a non-penalty finishing comparison look stronger.

Imagine a fictional player scoring twice in the example match, once from a penalty. The non-penalty goal count is one. Against 1.34 non-penalty xG, the descriptive difference is 1 − 1.34 = −0.34.

That is an accounting comparison for the defined sample, not proof of poor technique. An assessment of execution may also use post-shot information and observation, with the source and exclusions kept consistent. Our post-shot xG guide explains the additional question those models address.

Apply the same discipline defensively. A non-penalty xG-against figure excludes penalty attempts faced under the source’s convention. It cannot by itself explain the whole goals-conceded total if that total includes penalties or other differently classified events.

Do not silently switch between player and team records. Team totals can include attempts by several players, while a player comparison also depends on minutes, role and access to chances.

Removing penalties does not remove match context

A penalty often arrives within a match whose state has already affected both sides. A red card, an early lead or a tactical change can influence the remaining attempts. Subtracting one shot value does not rewind those events.

For example, a team may receive a penalty and then protect its lead. Its later shot pattern reflects that situation. The non-penalty figure is still a useful description, but it is not an experiment showing how the entire match would have unfolded without the penalty.

Opponent strength and venue also remain relevant. Compare samples with their dates and fixtures visible, and inspect whether a change in the opposition explains some of the apparent movement. The guide to short home-and-away form samples shows why an average needs its match context.

A disciplined review can therefore keep several observations together: the all-shot total, its penalty component, the non-penalty shot distribution, and the situations in which those attempts occurred. No single adjustment should replace the whole account.

Use historical values without pretending they are a forecast

Non-penalty xG describes modelled chance quality from attempts that happened. A future match has not yet supplied those attempts. Converting a recent average into tomorrow’s expected goals requires a forecasting method, not a change in the column heading.

Use information that existed before the forecast timestamp, preserve the opponent and personnel context, and evaluate the method on later unseen matches. A model can use past non-penalty data without treating it as a complete forecast on its own.

Watch for provider changes within a series. If a model is updated, a revised historical total may not be directly comparable with an earlier screenshot. Keep the source version or release information where it is available.

The backtest checklist explains how to inspect those data and time boundaries. Neither a clean historical calculation nor a convincing explanation establishes a guaranteed result or a profitable price.

Make the adjustment visible in the final review

A useful note identifies the provider, period and sample, then shows the total and the excluded penalty component. It states whether the remaining figure includes non-penalty set pieces and whether the comparison concerns teams, players or individual matches.

For the invented example, a defensible description is: “A generated 2.10 total xG, including a 0.76 penalty attempt. Its 1.34 non-penalty total included 0.38 from set pieces, so it should not be described as 1.34 of open-play xG.”

That wording lets another reader reproduce the adjustment and understand its limits. It is more informative than replacing one impressive headline number with another.

For adults aged 18 and over. The numerical examples are hypothetical, not OptaTips performance records. Provider references explain their own metrics and do not establish a data-supply relationship with OptaTips.

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 *