On single match expected goal totals

It’s been a heady week in analytics-land with expected goals hitting the big time. On Friday, they appeared in the Times courtesy of Rory Smith, Sunday saw them crop up on bastion of proper football men, Sunday Supplement, before again featuring via the Times’ Game Podcast. Jonathan Wilson then highlighted them in the Guardian on Tuesday before dumping them in a river and sorting out an alibi.

The analytics community promptly engaged in much navel-gazing and tedious argument to celebrate.

Expected goals

The majority of work on the utility of expected goals as a metric has focused on the medium-to-long term; see work by Michael Caley detailing his model here for example (see his Twitter timeline for examples of his single match expected goal maps). Work on expected goals over single matches has been sparser, aside from those highlighting the importance of accounting for the differing outcomes when there are significant differences in the quality of chances in a given match; see these excellent articles by Danny Page and Mark Taylor.

As far as expected goals over a single match are concerned, I think there are two overarching questions:

  1. Do expected goal totals reflect performances in a given match?
  2. Do the values reflect the number of goals a team should have scored/conceded?

There are no doubt further questions that we could add to the list but I think these relate most to how these numbers are often used. Indeed, Wilson’s piece in particular covered these aspects including the following statement:

According to the Dutch website 11tegen11, Chelsea should have won 2.22-0.77 on expected goals.

There are lots of reason why ‘should’ is problematic in that article but ignoring the probabilistic nature and uncertainties surrounding these expected goal estimates, let’s look at how well expected goals matches up over various numbers of shots.

You’ve gotta pick yourself up by the bootstraps

Below are various figures exploring how well expected goals matches up with actual goals. These are based on an expected goal model that I’ve been working on, the details of which aren’t too relevant here (I’ve tested this on various models with different levels of complexity and the results are pretty consistent). The figures plot the differences between the total number of goals and expected goals when looking at certain numbers of shots. These residuals are calculated via bootstrap resampling, which works by randomly extracting groups of shots from the data-set and calculating actual and expected goal totals and then seeing how large the difference is.

The top plot is for 500 shot samples, which equates to the number of shots that a decent shots team might take over a Premier League season. The residuals show a very narrow distribution, which closely resembles a Gaussian or normal distribution, with the centre of the peak being very close to zero i.e. goal and expected goal values are on average very similar over these shot sample sizes. There is a slight tendency for expected goals to under-predict goals here, although the difference is quite minor over these samples (2.6 goals over 500 shots). The take home from this plot is that we would anticipate expected and actual goals for an average team being approximately equivalent over such a sample (with some level of randomness and bias in the mix).

The middle plot is for samples of 50 shots, which would equate to around 3-6 matches at the team level. The distribution is quite similar to the one for 500 shots but the width is quite a lot wider; we would therefore expect random variation to play a larger role over this sample than the 500 shot sample, which would manifest itself in teams or players over or under-performing their expected goal numbers. The other factor at play will be aspects not accounted for by the model, which may be more important over smaller samples but even out more over larger ones.

One of these things is not like the others

The bottom plot is for samples of 13 shots, which equates to the approximate average number of shots by a team in an individual match. This is where expected goals starts having major issues; the distributions are very wide and it also has multiple local maximums. What that means is that over a single match, expected goal totals can be out by a very large amount (routinely exceeding more than one goal) and that the total estimates are pretty poor over these small samples.

Such large residuals aren’t entirely unexpected but the multiple peaks make reporting a ‘best’ estimate extremely troublesome.

I tested these results using some other publicly available expected goal estimates (kudos to American Soccer Analysis and Paul Riley for publishing their numbers) and found very similar results. I also did a similar exercise using whole match totals rather than individual shots and found similar.

I also checked that this wasn’t a result of differing scorelines when each shot was taken (game state as the analytics community calls it) by only looking at shots when teams were level – the results were the same, so I don’t think you can put this down to differences in game state. I suspect this is just a consequence of elements of football that aren’t accounted for by the model, which are numerous; such things appear to even out over larger samples (over 20 shots, the distributions look more like the 50 and 500 shot samples). As a result, teams/matches where the number of shots is larger will have more reliable estimates (so take figures involving Manchester United with a chip-shop load of salt).

Essentially, expected goal estimates are quite messy over single matches and I would be very wary of saying that a team should have scored or conceded a certain number of goals.

Busted?

So, is that it for expected goals over a single match? While I think there are a lot of issues based on the results above, it can still illuminate upon the balance of play in a given match. If you’ve made it this far then I’m assuming you agree that metrics and observations that go beyond the final scoreline are potentially useful.

In the figure below, I’ve averaged actual goal difference from individual matches into expected goal ‘buckets’. I excluded data beyond +/- two expected goals as the sample size was quite small, although the general trends continues. Averaging like this hides a lot of details (as partially illustrated above) but I think it broadly demonstrates how the two match up.

Actual goals compared to expected goals for single matches when binned into 0.5 xG buckets.

Actual goals compared to expected goals for single matches when binned into 0.5 xG buckets.

The figure also illustrates that ‘winning’ the expected goals (xG difference greater than 1) doesn’t always mean winning the actual goal battle, particularly for the away team. James Yorke found something similar when looking at shot numbers. Home teams ‘scoring’ with a 1-1.5 xG advantage outscore their opponents around 66% of the time based on my numbers but this drops to 53% for away teams; away teams have to earn more credit than home teams in order to translate their performance into points.

What these figures do suggest though is that expected goals are a useful indicator of quality over a single match i.e. they do reflect the balance of play in a match as measured by the volume and quality of chances. Due to the often random nature of football and the many flaws of these models, we wouldn’t expect a perfect match between actual and expected goals but these results suggest that incorporating these numbers with other observations from a match is potentially a useful endeavour.

Summary

Don’t say:

Team x should have scored y goals today.

Do say:

Team x’s expected goal numbers would typically have resulted in the following…here are some observations of why that may or may not be the case today.

Advertisement

Uncertain expectations

In this previous post, I describe a relatively simple version of an expected goals model that I’ve been developing recently. In this post, I want to examine the limitations and uncertainties relating to how well the model predicts goals.

Just to recap, I built the model using data from the Premier League from 2013/14 and 2014/15. For the analysis below, I’m just going to focus on non-penalty shots with the foot, so it includes both open-play and set piece shot situations. Mixing these will introduce some bias but we have to start somewhere. The data amounts to over 16,000 shots.

What follows is a long and technical post. You have been warned.

Putting the boot in

One thing to be aware of is how the model might differ if we used a different set of shots for input; ideally the answer we get shouldn’t change if we only used a subset of the data or if we resample the data. If the answer doesn’t change appreciably, then we can have more confidence that the results are robust.

Below, I’ve used a statistical technique known as ‘bootstrapping‘ to assess how robust the regression is for expected goals. Bootstrapping belongs to a class of statistical methods known as resampling. The method works by randomly extracting shots from the dataset and rerunning the regression many times (1000 times in the plot below). Using this, I can estimate a confidence interval for my expected goal model, which should provide a reasonable estimate of goal expectation for a given shot.

For example, the base model suggests that a shot from the penalty spot has an xG value of 0.19. The bootstrapping suggests that the 90% confidence interval gives an xG range from 0.17 to 0.22. What this means is that on 90% of occasions that Premier League footballers take a shot from the penalty spot, we would expect them to score somewhere between 17-22% of the time.

The plot below shows the goal expectation for a shot taken in the centre of the pitch at varying distances from the goal. Generally speaking, the confidence interval range is around ±1-2%. I also ran the regressions on subsets of the data and found that after around 5000 shots, the central estimate stabilised and the addition of further shots in the regression just narrows the confidence intervals. After about 10,000 shots, the results don’t change too much.

Test.

Expected goal curve for shots in the centre of the pitch at varying distances from the goal. Shots with the foot only. The red line is the median expectation, while the blue shaded region denotes the 90% confidence interval.

I can use the above information to construct a confidence interval for the expected goal totals for each team, which is what I have done below. Each point represents a team in each season and I’ve compared their expected goals vs their actual goals. The error bars show the range for the 90% confidence intervals.

Most teams line up with the one-to-one line within their respective confidence intervals when comparing with goals for and against. As I noted in the previous post, the overall tendency is for actual goals to exceed expected goals at the team level.

Expected goals vs actual goals for teams in the 2013/14 and 2014/15 Premier League. Dotted line is the 1:1 line, the solid line is the line of best fit and the error bars denote the 90% confidence intervals based on the xG curve above.

Expected goals vs actual goals for teams in the 2013/14 and 2014/15 Premier League. Dotted line is the 1:1 line, the solid line is the line of best fit and the error bars denote the 90% confidence intervals based on the xG curve above.

As an example of what the confidence intervals represent, in the 2013/14 season, Manchester City’s expected goal total was 59.8, with a confidence interval ranging from 52.2 to 67.7 expected goals. In reality, they scored 81 non-penalty goals with their feet, which falls outside of their confidence interval here. On the plot below, Manchester City are the red marker on the far right of the expected goals for vs actual goals for plot.

Embracing uncertainty

Another method of testing the model is to look at the model residuals, which are calculated by subtracting the outcome of a shot (either zero or one) from its expected goal value. If you were an omnipotent being who knew every aspect relating to the taking of a shot, you could theoretically predict the outcome of a shot (goal or no goal) perfectly (plus some allowance for random variation). The residuals of such a model would always be zero as the outcome minus the expectation of a goal would equal zero in all cases. In the real world though, we can’t know everything so this isn’t the case. However, we might expect that over a sufficiently large sample, the residual will be close to zero.

In the figure below, I’ve again bootstrapped the data and looked at the model residuals as the number of shots increases. I’ve done this 10,000 times for each number of shots i.e. I extract a random sample from the data and then calculate the residual for that number of shots. The red line is the median residual (goals minus expected goals), while the blue shaded region corresponds to the standard error range (calculated as the 90% confidence interval). The residual is normalised to a per shot basis, so the overall uncertainty value is equal to this value multiplied by the number of shots taken.

BootStrap_xGdiff_col

Goals-Expected Goals versus number of shots calculated via bootstrapping. Inset focusses on the first 100 shots. The red line is the median, while the blue shaded region denotes the 90% confidence interval (standard error).

The inset shows how this evolves up to 100 shots and we see that over about 10 shots, the residual approaches zero but the standard errors are very large at this point. Consequently, our best estimate of expected goals is likely highly uncertain over such a small sample. For example, if we expected to score two goals from 20 shots, the standard error range would span 0.35 to 4.2 goals. To add a further complication, the residuals aren’t normally distributed at that point, which makes interpretations even more challenging.

Clearly there is both a significant amount of variation over such small samples, which could be a consequence of both random variation and factors not included in the model. This is an important point when assessing xG estimates for single matches; while the central estimate will likely have a very small residual, the uncertainty range is huge.

As the sample size increases, the uncertainty decreases. After 100 shots, which would equate to a high shot volume for a forward, the uncertainty in goal expectation would amount to approximately ±4 goals. After 400 shots, which is close to the average number of shots a team would take over a single season, the uncertainty would equate to approximately ±9 goals. For a 10% conversion rate, our expected goal value after 100 shots would be 10±4, while after 400 shots, our estimate would be 40±9 (note the percentage uncertainty decreases as the number of shots increases).

BootStrap_xGdiff_col_wTeams

Same as above but with individual teams overlaid.

Above is the same plot but with the residuals shown for each team over the past two seasons (or one season if they only played for a single season). The majority of teams fall within the uncertainty envelope but there are some notable deviations. At the bottom of the plot are Burnley and Norwich, who significantly under-performed their expected goal estimate (they were also both relegated). On the flip side, Manchester City have seemingly consistently outperformed the expected goal estimate. Part of this is a result of the simplicity of the model; if I include additional factors such as how the chance is created, the residuals are smaller.

How well does an xG model predict goals?

Broadly speaking, the central estimates of expected goals appear to be reasonably good; the residuals tend to zero quickly and even though there is some bias, the correlations and errors are encouraging. When the uncertainties in the model are propagated through to the team level, the confidence intervals are on average around ±15% for expected goals for and against.

When we examine the model errors in more detail, they tend to be larger (around ±25% at the team level over a single season). The upshot of all this is that there appears to be a large degree of uncertainty in expected goal values when considering sample sizes relevant at the team and player level. While the simplicity of the model used here may mean that the uncertainty values shown represent a worst-case scenario, it is still something that should be considered when analysts make statements and projections. Having said this, based on some initial tests, adding extra complexity doesn’t appear to reduce the residuals to any great degree.

Uncertainty estimates and confidence intervals aren’t sexy and having spent the last 1500ish words writing about them, I’m well aware they aren’t that accessible either. However, I do think they are useful and important in the real world.

Quantifying these uncertainties can help to provide more honest assessments and recommendations. For example, I would say it is more useful to say that my projections estimate that player X will score 0.6-1.4 goals per 90 minutes next season along with some central value, rather than going with a single value of 1 goal per 90 minutes. Furthermore, it is better to state such caveats in advance – if you just provided the central estimate and the player posted say 0.65 goals per 90 and you then bring up your model’s uncertainty range, you will just sound like you’re making excuses.

This also has implications regarding over and under performance by players and teams relative to expected goals. I frequently see statements about regression to the mean without considering model errors. As George Box wisely noted:

Statisticians, like artists, have the bad habit of falling in love with their models.

This isn’t to say that expected goal models aren’t useful, just that if you want to wade into the world of probability and modelling, you should also illustrate the limitations and uncertainties associated with the analysis.

Perhaps those using expected goal models are well aware of these issues but I don’t see much discussion of it in public. Analytics is increasingly finding a wider public audience, along with being used within clubs. That will often mean that those consuming the results will not be aware of these uncertainties unless you explain them. Speaking as a researcher who is interested in the communication of science, I can give many examples of where not discussing uncertainty upfront can backfire in the long run.

Isn’t uncertainty fun!

——————————————————————————————————————–

Thanks to several people who were kind enough to read an initial draft of this article and the proceeding method piece.