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

シュート後のxG:サッカー試合レビューに追加されるもの

Opta Tips Team による 更新済み

ポストショットxGが通常のxGとどのように異なるか、ゴールキーパーやフィニッシュに関するどのような洞察を提供できるか、およびモデル定義や小規模なサンプルが分析にどのような制限をもたらすかを学びましょう。

Goalkeeper gloves and a notebook on a bench beside a tablet displaying a goal diagram at an empty football training ground.

A team can create good chances and still give the goalkeeper little to save. It can also score from a difficult chance because the shot is placed exceptionally well. Ordinary expected goals and post-shot xG help separate those two parts of a match.

Pre-shot xG assesses the chance at the moment of the attempt. A post-shot model adds information about the attempt that was actually delivered. Reading both can improve a match review, provided the numbers come from compatible definitions. Neither figure is a prediction of the next scoreline.

The practical question is simple: was the main difference in the chances created, the execution of the shots, or what happened when those shots reached the goalkeeper? This guide explains how to investigate that question without turning one performance into a claim about permanent ability.

Identify the model before comparing the numbers

Post-shot expected goals is a family of measures, not one universal statistic. A provider may label its output PSxG, xGOT or expected goals conceded. Check its glossary before combining that output with another source.

In Stats Perform’s explanation of xGOT, the model considers the underlying chance and where an on-target attempt is heading within the goal. A well-placed shot can therefore receive a different assessment from a poorly placed attempt from the same position.

Wyscout’s glossary distinguishes a pre-shot model covering all attempts from a post-shot model using on-target attempts, with blocked and wide attempts assigned zero post-shot value. Its documentation also describes provider-specific inputs. Similar labels do not establish identical models.

Write down the provider, competition coverage, date range and treatment of penalties, own goals and blocked attempts. When a site does not explain those details, preserve the uncertainty. A familiar abbreviation cannot supply missing documentation.

The existing guide to a team losing despite higher xG concerns the interpretation of chance quality. The extra question here is what the team did with those chances after the ball was struck.

Follow one small set of shots through both models

The following four attempts are invented teaching examples. They are not OptaTips output or estimates of a named player’s ability. The post-shot convention assigns zero to the attempt that misses the target.

AttemptPre-shot xGPost-shot xGOutcome
Close chance, weak placement0.350.08Saved
More difficult chance, strong placement0.100.40Goal
Low-quality chance, on target0.050.20Saved
Good chance, missed target0.300.00Wide
Total0.800.68One goal

The team accumulated 0.80 pre-shot xG. That total includes the good opportunity sent wide. Its 0.68 post-shot total describes a different information set: the on-target execution, with the missed attempt contributing zero under this convention.

It would be misleading to say the goalkeeper faced 0.80 of on-target danger. It would also be misleading to erase the missed chance from an assessment of how the attack created opportunities. Each total answers a different question.

The first attempt is another useful check. A high-quality chance need not become a difficult save if the execution is poor. Conversely, the second attempt becomes more threatening once its placement is known. These assessments describe probabilities, so the goal from the second attempt is still only one possible outcome.

Read goals prevented as a defined comparison

A common shot-stopping comparison subtracts goals conceded from the post-shot expectation for the same set of attempts. In the invented example, that gives:

Goals prevented = 0.68 − 1 = −0.32

The negative sign says the goalkeeper conceded more than this model expected for this particular sample. It does not establish that the keeper should have saved the goal, nor that the other two saves were easy. A probabilistic model does not assign a compulsory result to an individual attempt.

Before using the calculation, align the numerator and denominator. If the expectation excludes penalties and own goals, subtract goals conceded under the same exclusions. If one keeper played only part of a match, do not attach the entire team’s shot total to that player.

Check the sign convention on the page too. Some displays present expectation minus goals; another may display goals minus expectation. A positive number can mean opposite things under those definitions. Copying the label and formula is safer than inferring the meaning from a green arrow.

This is also a narrow measure of goalkeeping. Distribution, dealing with crosses, positioning before a shot and decisions outside the penalty area need other evidence. A shot-stopping total should not become an overall goalkeeper rating by default.

Separate volume from the difficulty of the attempts

Two goalkeepers can have the same total goals prevented while facing very different workloads. Here is another wholly hypothetical comparison, with consistent exclusions and one provider:

MetricKeeper AKeeper B
On-target attempts faced2570
Post-shot expectation7.014.0
Goals conceded512
Expectation minus goals+2.0+2.0

The shared +2.0 total does not mean the records are interchangeable. A faced fewer attempts and a higher average modelled difficulty per attempt: 7 ÷ 25 equals 0.28, compared with 14 ÷ 70 equals 0.20 for B.

Those averages still hide the distribution. Several difficult attempts and many routine ones can produce the same mean as a more even workload. Look at the underlying shots where they are available. Rebounds, crowded sightlines and unusual deflections can matter to the interpretation even when a summary table is tidy.

You may show totals alongside minutes and attempts, but label any rate precisely. “Per 90 minutes” and “per on-target shot” use different denominators. Neither creates a universal comparison when the competitions, providers or shot populations differ.

Do not treat the gap as a pure finishing score

Subtracting pre-shot xG from post-shot xG can be tempting. In the four-shot example, 0.68 minus 0.80 equals −0.12. It is a description of those model outputs, not a clean measurement of one player’s technical skill.

Start by checking which attempts enter each total. If the pre-shot column includes every attempt while the post-shot total gives off-target efforts zero, the difference reflects getting attempts on target as well as their placement. Comparing only the on-target subset answers a narrower question and must be labelled accordingly.

Defenders also influence what happens to a shot. A block can prevent an apparently promising attempt from reaching the goal at all. Match context, the available angle and pressure on the shooter complicate any claim that the final output belongs entirely to finishing ability.

Use restrained wording in a review: “The on-target execution was less threatening than the chance total suggests” is a bounded observation. “The striker cannot finish” requires much broader evidence and a method capable of supporting that conclusion.

A large pre-shot total and a small post-shot total need a match explanation

Imagine an attack producing 1.60 pre-shot xG but only 0.40 post-shot xG in a fictional match. The first follow-up is to inspect the attempt list. Did valuable chances miss the target? Were several efforts blocked? Did the on-target shots travel into relatively comfortable locations?

The answer may involve several mechanisms rather than one. Do not turn the gap into a story about goalkeeping before checking whether the keeper actually faced difficult attempts.

Now imagine an opponent creating 0.70 pre-shot xG and producing 1.20 post-shot xG. That is compatible with strong execution of a smaller collection of chances. It does not show that the opponent will reproduce that execution in its next fixture.

A useful review connects numbers with observable actions, then states what remains uncertain. Shot data can direct attention towards a phase of play; it should not replace the footage, personnel context or provider definitions. The before-and-after prediction worksheet helps preserve that distinction.

Keep post-match information out of a pre-match test

A match’s post-shot values do not exist before its attempts happen. They can explain that match retrospectively. Using them as if they were available before kick-off would give a forecasting test information from the future.

A pre-match analysis may use previous matches’ post-shot data, provided those matches and the relevant data releases occurred before the forecast timestamp. Preserve that cutoff. A corrected dataset published later should not silently replace the information available at the time.

Previous performances also need context. Changes in goalkeeper, opponents or the defensive structure can make a simple recent average less representative. A large goals-prevented total does not translate directly into a specific clean-sheet probability, and a team’s post-shot total is not automatically its expected goal total for the next match.

Any such forecast needs an explicit model and evaluation on information it could genuinely have known. Our backtest checklist explains how to look for those time and data boundaries.

Leave a review that another reader can check

For a practical match note, identify the provider and the included attempts first. Record pre-shot chance quality, post-shot execution and actual goals separately. Explain the strongest observable reason for the difference, then add the limitation that most affects the conclusion.

A defensible note might say: “In this hypothetical four-attempt sample, the attack created 0.80 xG and delivered 0.68 post-shot xG. One good chance missed the target. One goal was conceded from the three on-target attempts. The sample is too small for a claim about lasting finishing or goalkeeping ability.”

That note is less dramatic than a claim that one side deserved a particular score. It is also more useful. It identifies the part of the performance being assessed and preserves the evidence a future review would need.

The provider explanations cited here describe their own methods. Their inclusion does not establish that OptaTips uses those feeds or has a commercial relationship with those organisations. For adults aged 18 and over; statistical interpretation does not guarantee a match outcome or profit.

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 *