The existence of so-called clutch players has been long debated. There are claims that certain players can “up their game” and produce a goal when their side needs it most; be it a last minute winner or a vital equaliser in a cup final. But is this really the case? Read on to find out.
The kind of events mentioned above are relatively rare in a low scoring sport, such as soccer, and it isn’t surprising that they remain in the memory long after other matches where the game meanders to an inconclusive draw.
It is wise to treat memorable events as being rarely typical of the contribution expected, even from outstanding players - Brazil’s Roberto Carlos scored one of the game’s most astonishing free kicks against France, but he rarely repeated the feat with even as much as a goal when taking subsequent dead ball shots.
To ascribe “value” to a goal we need to be able to judge the probability of a successful overall outcome for the scoring side immediately before and immediately after the scoring event.
Data tends to look at important events in a sporting contest with a dispassionate eye, merely recording the frequency at which events occur and with some manipulation their likely effect on the outcome of a contest.
However, some goals are more important to the match outcome than others - something pointed out by Andrew Beasley in his article that asks; how important is the first goal in soccer?
The two situations mentioned above increase the value of a goal under the context in which they are scored. The sixth goal in a 6-0 rout has little effect on a match in comparison to a last minute winner.
How do you define a “clutch scorer”?
To ascribe “value” to a goal we need to be able to judge the probability of a successful overall outcome for the scoring side immediately before and immediately after the scoring event which can be done using the idea of a side’s expected points.
Pinnacle quote low margin match odds for a wide range of soccer matches - one example of this would be the latest round of Premier League fixtures. The closing odds for these matches represent Pinnacle’s best estimate of the likely outcome of such a match based on a variety of inputs and bettors’ opinion (Wisdom of the crowd).
Weaker teams are unlikely to run up many large margin wins, so on average a goal scored by a relegation threatened team results in a larger change in expected points.
Converting odds into probabilities is crucial in betting. If we use the odds for an Arsenal win (1.38*), Leicester win (8.94*) and a draw (5.17*) in their opening fixture, the probability of all three possible outcomes adds up to 1.03 - 0.725 for the Arsenal win, 0.112 for a Leicester win and 0.193 the draw.
The cumulative probability of every possible outcome in a contest in reality can only total 1.00; the extra 0.03 represents Pinnacle’s low margin. But to arrive at the likely “true” chance of any of the three outcomes occurring, we should reduce the implied probabilities of each outcome to satisfy this condition.
There are many appropriate ways to do this, but the simplest is to divide each implied probability by 1.03 to give 0.70, 0.11 and 0.19 (or 70%, 11% and 19%). These estimates of the implied true probabilities of each outcome can then be used to calculate the expected points for both sides at kick off.
There is a 0.70 probability that Arsenal will win the game and three points and a 0.19 probability that they will draw and receive one point. Overall, their expected points comes to (0.70*3) + (0.19 * 1) = 2.29 expected points. The figures for Leicester are (0.11*3) + (0.19*1) = 0.52
We can use a variety of methods to quantify and celebrate outstanding contributions from players and teams, the data is much less optimistic about individuals and sides continuing to star as the hero of the hour.
These calculations can be repeated as the game progresses and the odds and associated implied probabilities alter with goals scored and time elapsed. For example, if Leicester were to take the lead in the first few minutes, the odds will change in their favour.
In this scenario, Arsenal (strong pre game favourites) still have almost the entire game to retrieve the situation and the new implied probabilities could become 0.44 for the Arsenal win, 0.3 for Leicester and 0.26 the draw.
The expected points for Arsenal would have fallen to 1.58, while Leicester’s has risen to 1.16. The hypothetical Leicester goal is relatively valuable, as is any that breaks a stalemate, but if the game had remained scoreless and the Foxes had instead scored in the very last minute of the game the goal would have carried even more value.
A goalless game in the final minute is highly likely to conclude in a draw and Leicester’s expected points at that stage would be very close to one, but a goal in the later stages of a match would increase their expected points to very nearly three.
Finding clutch scorers in the Premier League
We’ve now got the basis for a method to measure the importance of a goal to the likely long term outcome of the remainder of the match, based on probabilistic estimations and expressed in the change in expected league points.
With this, we can attempt to see which players bolster their actual goal numbers by scoring in matches where the outcome is largely already determined and who contributes important scores. Of course, we still need to allow for other influences.
The combined goals scored by Ozil and Giroud, on average improved Arsenal’s expected points by 0.6 per goal compared to just 0.47 for Sanchez.
For example, weaker teams are unlikely to run up many large margin wins, so on average a goal scored by a relegation threatened team results in a larger change in expected points than a top team who consistently wins by three or four goals.
There’s also the question of opportunity, how many minutes each player was on the pitch and also during which stage of the match. Therefore, compiling a list of players who scored valuable goals at opportune moments becomes slightly subjective; albeit with a large data input.
Unsurprisingly, as Arsenal’s leading scorer in 2016/17 Alexis Sanchez’ goals amassed the largest overall expected points change for Arsenal. Each goal scored by Arsenal improved their expected points by an average of 0.55 expected points and nine of the Chilean’s goals exceeded this team average.
If there was a clutch scorer at Arsenal last season, the title perhaps lies between these two, particularly Giroud who played only 1200 minutes
The combined goals scored by Ozil and Giroud, on average improved Arsenal’s expected points by 0.6 per goal compared to just 0.47 for Sanchez. Five of Ozil’s eight goals and seven of Giroud’s twelve individually exceeded Arsenal’s team average of 0.55 expected points per goal.
If there was a clutch scorer at Arsenal last season, the title perhaps lies between these two, particularly Giroud who played only 1200 minutes rather than the also impressive, but bulk scoring of Sanchez.
Players from other Premier League teams whose less frequent goals regularly advanced the cause of their side over and above the average for their team included Willian at Chelsea, Zaha at Crystal Palace, Slimani at Leicester, Mata at Manchester United, Wijnaldum at Liverpool, Allen at Stoke, Redmond at Southampton, Ayew at West Ham and McAuley at WBA.
Clutch scoring or randomness?
Are the players mentioned above the ones to rely on at crucial stages in a match or was their 2016/17 record merely a consequence of the randomness at which goals arrive? Will the mantle of “clutch” scoring pass largely to other teammates next season?
Evidence from other related aspects of scoring strongly suggest that while the players mentioned are accomplished scorers, they have much less control about when they score.
Newcastle in 2011/12 and Reading last season showed signs of exhibiting team “clutch” ability, scoring when it most benefitted their final outcome and conceding in games that were already lost causes.
The scorers of vital goals are much more likely to have been plucked at random from the principal cast of available scorers on the day.
Newcastle could not repeat this ability in 2012/13 and doubts remain around Reading’s chances of reprising the feat of turning a +4 goal difference into a 3rd place finish.
At a player level hot streaks analogous to clutch scoring are also transient, if spectacular. Consider the Premier League’s strikers who have greatly out performed their expected goals pre-Christmas by scoring many more actual goals than would be likely from the chances they have been presented.
The post-Christmas record for these players is invariably much closer to the expected totals based on the chances they receive. This isn’t just the case for expected goal totals in soccer, regression towards less extreme returns is universal across all measured outputs in all sports.
So although we can use a variety of methods to quantify and celebrate outstanding contributions from players and teams, the data is much less optimistic about individuals and sides continuing to star as the hero of the hour.
Much more mundanely, the scorers of vital goals are much more likely to have been plucked at random from the principal cast of available scorers on the day.