Bruin Sports Analytics
Project Restart: Winners and Losers of the EPL’s Return
By: Ethan Allavarpu
After a hiatus of around three months, the English Premier League resumed its season in mid-June in an all-out sprint to the finish line. Most of the league’s fans knew which team would hoist the trophy around a month later on the last weekend of July: the Liverpool Reds. Nonetheless, there was still a cornucopia of reasons for the season to play to its conclusion: would Liverpool continue on its historic march, shattering previous records? Which teams would play in Europe next season? Which unfortunate three clubs would be relegated to the Championship? How would home-pitch advantage manifest, if at all?
By comparing team performance before and after the league suspension, I delved into these questions. Although most teams noticed little-to-no difference between the two periods--statistically speaking, since it probably threw individual players into disarray--there were a few aspects of the data that stood out, trending toward clear winners and losers.
A few areas of focus include basic statistical analysis of changes in standings in the league as well as points before and after the restart. I also looked at how teams statistically performed against themselves, through two related metrics: average points earned per match in each of the periods as well as the use of a chi-square test for equality of proportions to see if any teams performed differently after the restart. When performing this chi-square test, I encountered a chi-square value of 36.735 and a corresponding p-value of 0.008563, indicating that we reject the null hypothesis at the 0.05 significance level that the points distribution for the twenty teams did not remain the same. Note: The “cutoff” which I utilized for the chi-square contributions was between Manchester United and Tottenham Hotspur, as there is a difference greater than one between the two subsequent contributions. In addition to these chi-square contributions, I looked at individual tests for difference in proportions, as mentioned below.
Because of the low p-value, I found it appropriate to perform follow-up analysis as to which teams performed at a different capacity concerning the proportion of total points possible the team earned. When looking at all twenty teams, to control the experimentwise error rate, I decided to use Bonferroni’s correction, changing my significance level (alpha) from the standard 0.05 to 0.05/20, or 0.0025. Now, even with this shift in evaluation criterion, I still noted instances of relatively low p-values that could suggest a difference in performance. Utilizing this data as a collective whole, I determined which individual teams won--and lost--Project Restart.
In the other section of this project, I turned my attention from individual teams to the belief of home-pitch advantage. During normal times, the atmosphere at a soccer stadium is absolutely electric, with fans often providing the enthusiasm and energy to help a team gain points previously thought unattainable while simultaneously causing the opposition to lose the concentration and focus so essential for victory in competitive sports. Unfortunately, this situation does not qualify for "normal times." Pitches are empty and, unlike on TV where spectators hear crowd noise for a sense of "normality," the in-stadium experience is eerily quiet, with the athletes only hearing their calls for the ball and the sounds of the ball hitting the woodwork reverberating in the stands. The momentum and energy that a fanbase musters no longer exist. Home-pitch advantage has seemingly been reduced to the grass itself, nothing more. Therefore, I analyzed the performances by home versus visiting teams in the two segments of the season to see if home-pitch advantage as we know it exists in any significant manner in empty stadiums.
Before Suspension: 45 points (5th)
Project Restart: 21 points (finished 3rd with 66)
Chi-Square Contribution: 2.8642
Percentage of Points Earned Change: +26.1 Percentage Points
Manchester United presented itself as one of the clearest, if not the clearest, winners of Project Restart. The team kicked its performance into a higher gear with the resumption, determined not to settle for a Europa League campaign again. United ended up gathering 21 points during project restart--second only to cross-town rivals Manchester City--and catapulted itself from 5th place to a third-place finish and a Champions League berth for next season. Before the suspension, Manchester United averaged around 1.55 points per game (52%); once coming back, though, the team earned an average of 2.33 points in each match (78%). This 26% skyrocket presented a p-value of 0.03 when testing for a difference in the proportion of points earned; while it is still higher than the set alpha of 0.0025 because of the twenty simultaneous confidence intervals, it is still rather low, indicating a marked difference--and in United’s case, improvement--after Project Restart began. Manchester United provided us with a perfect example of how to use the time off to prepare and improve its play, proving once again that, even in times of chaos and uncertainty, the Red Devils still possess their upper-class quality that has been present for the past couple of decades.
Before Suspension: 34 points (14th)
Project Restart: 18 points (finished 11th with 52)
Chi-Square Contribution: 4.4715
Percentage of Points Earned Change: +27.5 Percentage Points
For a team that finished middle-of-the-pack and was not in any legitimate danger of relegation even before the shutdown, Southampton played like a team with everything to lose and everything to gain.
The team had just 34 points through 29 games heading into the break, an average of 1.17 points per game (~39% of possible points) to 18 points in the final nine games (2.00 points per game; 67%). Southampton’s 27.5 percentage point improvement (p-value of 0.022) came almost out of the blue, considering that it had no reason to give such an effort. Compare this performance with the similar situation of Crystal Palace, discussed below, and we see a stark difference in mentality when there is nothing on the line. Southampton played remarkably well; had the team performed like this throughout the season, it would have certainly been in contention for European competition for the 2020/2021 season. For now, though, Southampton fans can be satisfied with the team’s recent play and hope that this vein of success will continue into the start of next year’s Premier League campaign.
Before Suspension: 41 points (8th)
Project Restart: 18 points (finished 6th with 59)
Chi-Square Contribution: 1.7007
Percentage of Points Earned Change: +19.5 Percentage Points
I included Tottenham in the winners’ column because it performed relatively well at the end of the season. The team’s performance was markedly higher over the final stretch, earning around 0.6 more points per game than before the shutdown (from 1.4 (47% of points possible) to 2.0 points per game (67%), an improvement of twenty percentage points). In turn, this translates to approximately a six-point increase in their final standing scores. The p-value of this shift is 0.12, which is a bit high, but still low enough to suggest a potentially influencing factor: the return of a healthy Harry Kane to the starting lineup. Now, while the supposed six-point increase because of the shutdown may not seem monumental, it ended up being just enough for the team to qualify for Europa League qualifying play on goal differential (three goals better than seventh-place Wolverhampton), taking the last spot in European football offered to Premier League teams (Liverpool, Manchester City, Manchester United, and Chelsea all qualified for the Champions League; Leicester City qualified for the Europa League). Therefore, even though it may not be statistically significant, the movement of the team in the standings--from eighth to sixth--earned Tottenham a spot in the winners’ column, even though it was not at the same level as Manchester United or Southampton.
Before Suspension: 82 points (1st)
Project Restart: 17 points (finished 1st with 99)
Chi-Square Contribution: 3.4086
Percentage of Points Earned Change: -31.3 Percentage Points
Imagine a season in which your team earns 99 points in the Premier League--the second highest total of all-time--goes undefeated at home--with 18 victories in 19 home games--wins the Premier League by 18 points (the equivalent of six victories), and allows the fewest goals in the league, only to end the season with a whimper. Liverpool was certainly a loser of Project Restart despite the incredible success the team saw throughout the season.
Liverpool began the season fast out of the gate, racking up victories in an unstoppable manner. In March, though, the team hit a brick wall: COVID-19. For three months, Liverpool had time to rest and relax, but that also meant time for other teams to rest and recuperate while simultaneously depriving Liverpool of any momentum it had gathered during the season. Upon returning to play, Liverpool appeared sluggish, dropping their points-per-game average by a whopping 0.94 points per game (31 percentage points): 2.83 (94% of possible points) to a mere 1.89 (63%). This shift in performance seems highly unlikely by chance alone; the p-value of near-zero confirms that the Liverpool team that returned from the break was not the same one that left for it three months earlier.
Normally, a rate of 1.89 points per game is still extraordinary; in fact, it would rank sixth on a list of points scored by a team since the restart. However, Liverpool was destined for historic accomplishments, and the suspension of the season derailed the promise of an Earth-shattering campaign.
Liverpool had a fantastic season, one that we should not easily forget. However, it will be overshadowed by events that occurred during the same time--namely the coronavirus--and ultimately seen as a season of "what if"s. What if the season had progressed normally? What if Liverpool had gone on at its breakneck pace to end not with 99 points, but rather 107, or 108? What if the team had demolished the records that seemed unbreakable?
Unfortunately, those "what if"s will continue to remain hypothetical. Liverpool had a great season--an unforgettable one--before the suspension, but Project Restart got the better of them. Their season post-resumption was good, but nowhere near the same caliber of intensity and quality, leaving Liverpool fans a tinge dissatisfied and the team a rather clear loser of Project Restart even though it won the season.
Before Suspension: 53 points (3rd)
Project Restart: 9 points (finished 5th with 62)
Chi-Square Contribution: 3.8595
Percentage of Points Earned Change: -27.6 Percentage Points
Leicester City was playing at a rather exceptional level this season compared to the mid-table finishes of the past few seasons. Most people remember the 2015-2016 campaign in which Leicester stunned the soccer world by winning the Premier League after escaping relegation by the skin of its teeth only one season earlier. Since then, the team had become "mediocre," with no genuine threat of relegation, but also limited chances at European football.
This season, though, Leicester was powering through most opponents until COVID-19 struck. The Leicester team that left the pitch was not the same one that returned, as the team scored a pitiful 9 points in the final gameweeks, falling out of Champions League qualification. Furthermore, this drop of 27.6 percentage points in possible points was the fourth-largest contributor to the chi-square statistic and had a p-value of only 0.022; one that, though not statistically significant, is still of interest. The shift from elite play to mediocre performances will weigh heavily on the team’s players moving forward into next season, but perhaps they can utilize this setback as fuel to intensify their focus.
Ultimately, Leicester’s performance in the first segment of the season was enough to earn European play (via the Europa League), but the team should be utterly disappointed in their Project Restart performance. For Leicester City, this is an issue that the club should resolve during the quick turnaround before the start of the next season.
Before Suspension: 39 points (11th)
Project Restart: 4 points (finished 14th with 43)
Chi-Square Contribution: 5.8753
Percentage of Points Earned Change: -30.0 Percentage Points
Crystal Palace presents a story that is, in essence, the polar opposite of Southampton’s narrative. Neither team had much reason to strive to perform well in Project Restart--both teams were mid-table sides with little ability to reach European competition and enough of a margin of error to not have to worry about relegation. They even ended up trading positions in the standings: 11th and 14th. However, while Southampton rose and performed at an impressive level, Crystal Palace provided its fans with a performance in the restart that was the worst by almost any team in the Premier League (the one exception being Norwich City, discussed below).
When considering the differences between pre-COVID and post-restart performance, Palace’s chi-square contribution was 5.8753, the second-largest contributor. When assessing the squad’s disastrous performance, the drop of 30 percentage points in points earned had a p-value of 0.01, one of the lowest (and thus closest to significance) of all twenty teams--in the nine games since the restart, Crystal Palace only obtained 15% of all possible points, with a seven-game losing streak in the midst of that.
If anything, the end of the Premier League season was a blessing for Crystal Palace. If there were five or so more games in the season, this story could have become even bleaker for the Selhurst faithful. The lack of success in such a five-game stretch could have meant relegation for a team that was 12 points clear of the relegation zone before the shutdown. A final-day draw with Tottenham at home will, unfortunately, be a highlight for the Crystal Palace squad since returning mid-June. One potential source of consolation is that the short turnaround will provide Palace with the chance to quickly forget about the chaos of Project Restart and look forward to the upcoming season.
Before Suspension: 21 points (20th)
Project Restart: 0 points (finished 20th with 21)
Chi-Square Contribution: 6.7879
Percentage of Points Earned Change: -24.1 Percentage Points
Norwich City. I think that there is only one way for me to describe the team’s performance after the restart: an absolute train wreck. Heading into the suspension of the season, Norwich had played 29 games--the same as all but a few teams which had to reschedule matches--but had somehow only managed to earn a pitiful 21 points from those matches, losing 18 of those 29 games. Some people saw the COVID-19 suspension as a chance for Norwich to rebound and potentially claw back from the six-point hole out of relegation. Unfortunately, a three-month wait only delayed the inevitable, and Norwich returned to play with a pitiful display.
0. Not 0 wins--that would be shocking, but not completely unfathomable. No, the 0 refers to the number of points Norwich earned in its final 9 games of Premier League play. For a team that seemed to disappoint at every turn, Norwich found a way to outdo itself, truly crashing and burning before relegation to the Championship. Their percentage of points earned fell from 0.72 points per game (a sad 24% of possible points) to an unthinkable 0.00 points per game (obviously 0%). This change of 24 percentage points carried a p-value of 0.011, which although is higher than the Bonferroni-corrected alpha of 0.0025, is small enough to give second thoughts.
I see Norwich’s collapse as significant, simply because the team could not have done any worse. It scored a single goal in 9 games, allowing 23 goals for a difference of -22. For comparison, over the first 29 games, Norwich’s goal differential was -27. Now, this lack of production very well could stem from a lack of motivation and thus effort from Norwich, but as a team that should have given everything to stay in the Premier League, Norwich City simply did not perform, which makes the squad a complete and utter failure in Project Restart.
Home-pitch advantage should have gone into freefall in Project Restart. With no fans, no canned, simulated noise pumped into the stadiums, and no supposed swings the supporters could garner for their respective sides, almost every site should have been treated as a neutral site, at least concerning the match results. However, there was no evidence of a significant impact on the home side’s performance in relation to the points earned from matches. In fact, the number of points earned by the home side as a proportion of the total points awarded increased (though by a statistically insignificant margin), from 58% of points awarded to 58.2%, indicating that our significance test for a difference in the proportion of points for the home team was not necessary to fail to reject the null of equal proportions (the p-value was essentially 1 due to the extremely close nature of our two sample proportions).
Home-pitch advantage will obviously impact some players negatively. Those who admire the spotlight, rejoice in the sounds of the fans cheering on the team to victory, and thrive on the fame will likely have faced a slump in performance at home relative to before the restart. Other players, though, may see a surprising uptick in production. Some players prefer to keep the game simple, not worry about tens of thousands watching in the stands, and just play soccer in the same way they played throughout their childhood.
Maybe home-pitch advantage has something to do with the stadium and pitch itself, not just the fans. Maybe it’s the familiarity, the regularity of playing in a usual location. Maybe it’s the lack of travel to another location. Whatever home-pitch advantage is, it appears to have survived the quiet, supporter-less Project Restart.
Ultimately, Project Restart’s impact on the twenty clubs of the Premier League varied: some teams felt a minor impact, while others succeeded (such as Manchester United and Southampton) and even more completely faltered (Crystal Palace, Norwich City). If the teams play the start of the next season without fans in the stadium, it will be interesting to see whether these performances after the break were a result of the lack of spectators and this new dynamic of soccer or if it was simply the result of chaos and confusion. We will also see three new teams enter the fray--will they be able to handle the big (but empty) stage?
Home-pitch advantage, surprisingly, has not seemed to be as significant of a factor as many perceived. Project Restart displayed little effect on teams’ general ability to win (or at least earn points) at home. It will be interesting to see a shift in the dynamic (if any) as players become increasingly familiar with and accustomed to the "backyard soccer" atmosphere of playing on the pitch without the background noise and hubbub. Ultimately, though, the question for almost all Premier League fans--and clubs--is when we will be able to watch the matches at the stadiums, standing and cheering right next to our fellow supporters.
For now, though, simply having live sports will have to do.
Source: Premier League (premierleague.com)
GitHub Repository: https://github.com/ethan-allavarpu/projectrestart