Expected Points and The Justice Table
The English Premier League Table lies to you!
Always bringing the latest in football and data analytics. This xFO edition is about the Expected Points statistic and The Justice League table. An alternative way to view the best performing teams throughout the season. This article also provides context, through the expected points statistic, to the surprises within the Premier League table and as to why some teams deserve to be in the position they are, or deserve to be at lower or higher league position.
Introduction:
The EPL table lies. This may be a very controversial statement since most fans believe the league table is the most accurate in determining which teams are the best and the worst. Fans tend to believe that the team in first place is the best team in the league, that the one at the bottom is the worst, and that every other team in between is ranked based on their performance throughout the season. Although this might be true in some cases, most of the time it is not. With xG, we have come to understand that teams who create the highest quality goal scoring opportunities tend to win matches. However, this is not always the case. Teams with the lower xG sometimes win matches due to a combination of scoring low quality chances and their opponents missing high quality chances with high a xG value. This measures how a team performed during the match but it cannot measure how the team performs throughout the whole season.
xG itself cannot measure team performances over a period of time. It cannot help explain whether a team has been playing well or not during that same time period. What happens if a team is consistently creating high xG scoring opportunities but fails to capitalize on them? It may be that this team is playing well but may not be gaining the points they deserve since they are not being clinical or capitalizing in front of goal. Or, what if a team is consistently conceding goal of the season contenders who tend to have a low xG? It may be that these teams have less points because they have been unfortunate, consistently conceding amazing goals, which cannot really be defended against, that have low xG values. If xG cannot help in measuring these situations in football, what can be the correct statistic to measure such?
Expected Points, or xP, is the ideal statistic to measure a team's performance throughout a period of time. xP does this by taking the xG of all the shots taken by the two teams in a given match and calculating the expected points these teams would have taken from the game given the chances they created in it, vs. the chances they conceded. xP helps in contextualizing which teams have truly been better throughout a period of time rather than just in an individual match. It can also help contextualize which teams have been more fortunate than others throughout the given time period and help conclude whether a team deserves to be in a higher or lower position in the league table.
Expected Value:
The expected value of a variable is the “long run average value of repetitions of the experiment it represents.” In other words, it is just the expected value of an action when the action is repeated hundreds or thousands of times. An example of this is rolling a normal dice, as explained in the book The xG Philosophy by James Tippett. The expected value would be 3.5 as this is the average value of rolling the dice hundreds of times. It is impossible to get an exact 3.5, but after rolling ones, twos, threes and so forth a different number of times, the average of the rolls of the dice will be 3.5. Expected value works the same way as xG of an individual shot. For example, it is impossible to score .75 of a goal, but if different players take the same exact shot hundreds or thousands of times, the shot will go in .75 or 75% of the time. This is the way to calculate xG, using past data to explain the expected value or percentage at which the current shot will go in.
Referring back to expected points, when the shot data of the home and away team is collected and inserted into a xP algorithm/equation, it will result in the amount of points each team could expect to achieve in a game. In other words, if the same game were played numerous times with the exact same shots and scoring opportunities, the home team could expect to take away a certain number of points. In the same way, the away team could also take away a certain number of points. Remember, the expected value or the xP might be technically impossible to get from a league match (where a win equals three points, a draw one point and a loss zero points), but given that the game is being repeated numerous times, these values may occur. Let us look at an example using Understat.com for the shot data, and an Expected Points simulator to find the expected points for the two teams in a given match.
Leeds vs Arsenal:
On November 22nd Leeds played out a 0-0 draw against Arsenal. Usually 0-0 draws are very boring with little attacking football being played. However, this 0-0 was different as Leeds created great chances, and Arsenal created two or three chances. However, neither team was able to find the back of the net. Below is the shot map of the match.
Leeds created enough high-quality chances to have an xG of almost two. Arsenal did not create enough opportunities yet ended up with a xG of 0.71. Both teams left with the same amount of points. Knowing what xG entails, it is clear that Leeds were the better team creating better and more scoring opportunities for most of the second half than a 10-man Arsenal. However, what would be the xP that Leeds could have expected to take away if this game were played out in the same exact way hundreds of times? As seen below, Leeds (Team A) could expect 2.23 points while Arsenal (Team B) could expect 0.56
points. Similarly, Leeds would win that game 67% of the times, Arsenal would win 11% of the times, and a draw (which is what happened) would occur 21% of the times.
The chart above shows that Leeds were unfortunate not receiving more points than they should have. Of course, the players could have been more clinical in front of goal as they had high quality chances. This is why football is so unpredictable. Despite Leeds having been so dominant throughout the match, they obtained two points less than they should have and Arsenal got one more point than they should have.
This analysis can be carried out with any team(s) for any given amount of time. The only data that is needed is the xG of all the shots a team creates in an individual match, as well as all the xG a team concedes to their opponent in that same match. Once this xG data is collected for all the games in one time period and then individually inserted into an expected points simulator, like this one, the simulator will give you the xP each team could have expected to take away from that match. Adding up all the xP from the individual matches results in the accumulated xP the team could have expected to accumulate during that given time frame. This accumulated xP is what can be compared to a teams actual number of points to see if their performances deserve a higher or lower league standings. Following these steps with different teams will result in the possibility of creating a justice table where teams can be ranked based on their xP during the analyzed time period.
As an example I took all the xG data from Leeds United’s first eleven matches of the current EPL season which resulted in an xP of 14.75. Leeds currently have 14 points which signifies that they have a number of points close to what they could expect if the games were replayed hundreds or thousands of times.
The Justice Table:
Data company Smartodds created the expected points tally and called it the Justice Table. Smartodds felt that using this statistic could judge who the better performing teams were throughout a time period and a specific number of games, or throughout the whole season. The same way xG gives a better picture of who performed better in a match, xP conveys which teams performed better throughout the season or over a specific period of time. Teams who have less points than their xP are said to have been unlucky, while teams whose xP exceeds their actual points are said to have been lucky.
This season has been a rollercoaster and is currently showing surprising results i.e. unexpected teams being in the top half of the table. xP can show whether these teams have been lucky or if, on the contrary, genuinely deserve to be where they now are. Understat.com has an xP algorithm embedded in its website that allows filtering the current EPL league standings and shows the statistics of xP. Below we can see what the current league table looks like after eleven games, using the xP statistic.
The best way to make analyses with xP is by comparing the actual league standings against the ‘justice league table’ which will rank the teams based on the xP instead of actual points obtained. This league table can be seen below.
It is important to remember that xP is based on the xG a team creates or concedes to their opponent. This is why the variation between actual points and xP can be minimal or significant. If we focus on the teams in the top six positions of the current league table, there is a team that can be considered lucky. Southampton are currently in fifth place with 20 points, much more than any fan or pundit could have ever predicted. However, in the Justice League Table they are in 14th place. A drop of nine places. There are two reasons that could explain the disparity between Southampton’s actual position and their justice league position. One, they could be scoring goals on opportunities that have low xG signifying they are consistently exceeding their xG in matches, this over the first eleven match weeks. Two, opponents might be missing their high xG scoring chances allowing Southampton to gain unexpected points.
Looking at each individual match, Southampton have been exceeding their xG consistently. In the picture below we can see examples of Southampton scoring 4 goals from an xG of 1 and scoring 2 goals from an xG of 0.79. A combination of Southampton exceeding their xG and their opponents underperforming their xG during league games has led Southampton to obtain the most favourable match results. This has given way to more points than the expected throughout the season so far.
The Case of Brighton:
On the flip side, Brighton and Hove Albion have been playing some very good football but have not gotten the results to back this up. Comparing their position in the league table against the justice league table, Brighton are in 16th and 4th place respectively. This is a position change of twelve slots which is the most by any team in this league season so far. Brighton are playing good attacking football. Nevertheless, their lack of consistency in finding the final pass to create a clear scoring opportunity that would increase their xG and overall chances of winning is noticeable. This factor does not have a direct impact on their xG and consequent xP because the bad passes prevent a goalscoring chance from happening. Since xP depends on xG, Brighton need to ensure that these crucial passes find their intended target therefore increasing the amount of high-quality chances they create.
Brighton have been unlucky a few times because their opponents have outscored their xG. However, in some matches, Brighton have either matched the opposition’s xG or even surpassed it. This can be seen in two of their first three matches where they outplayed Chelsea and had a higher xG, but Chelsea’s individual talents bailed the team out. Similarly, Brighton surpassed Manchester United’s xG who equalized in the 94th minute after a penalty was awarded by VAR. Brighton came away with 0 points from those two games but could have gained 6 altogether. This shows the difference between their xP and their actual points.
It is important to keep in mind that in both Southampton and Brighton’s cases, what happened in their games was not the only influencing factor that impacted the difference between their actual points and xP.
Outside Influence:
The situations of other teams may highly influence the difference in positions in the league table and the justice league table. Although a team may have more xP than actual points, this does not guarantee the team a higher position. A good example of this is Sheffield United. Sheffield surprised everyone last season by finishing ninth in their first season back in the EPL. Once again they have surprised many fans and pundits, this time for their lack of points. Sheffield currently sit in last place with only one point after eleven games, the worst start to any EPL season. Looking at the justice league table, their difference in number of points is the largest of any team, a negative 9.43 point difference between their actual points and xP which means they have 9.43 less points than they could have expected to have. Although they could have expected to take 9 points since the start of the season, this would still only place them in 18th place.
The reason for the huge point difference not resulting in a bigger change in league positioning is that other teams at the bottom of the table have created enough quality chances and held their opponents to less high scoring opportunities warranting more points and xP than Sheffield. It can be said that Sheffield have been unlucky in their games, but still, their overall play does not ensure them a place out of the relegation zone. A Sheffield United fan can hope for good fortune to come around later in the season, but before this happens the team needs to improve their xG by creating better quality chances consistently. This is the only way they will be able to stay in the EPL another season.
Final Thoughts:
xP is a great start in analyzing whether a team is over or under performing. It does not explain the whole story of the team’s successes or struggles since it is based on xG. The factors xG does not take into account will also not be factored into xP. xP does, however, summarize the team’s situation and allows for the decision makers to conduct a deeper analysis and identify what teams should continue to do and/or improve on.
Teams that have been over performing are Aston Villa and West Ham. Looking at their xP and Justice League Table, it can be said that they are deservedly in the top half of the table. Their point difference is both less than one, meaning they have taken the expected amount of points. They have created enough high xG chances to win games as well as limiting their opposition from creating high xG opportunities. The question to these teams is if they will be able to maintain this throughout the season, or if they will drop off. These are two teams that survived relegation last year and were struggling to keep clean sheets. West Ham and Aston Villa consistently gave up high xG opportunities to their opposition last year and were ranked 2nd and 3rd in xGA (Expected Goals Against) which is one of the main reasons they were fighting relegation. Their newfound defensive solidity has massively helped both teams at the start of this season. It will be interesting to see if they will be able to carry on in this way the rest of the season or if they will begin to ride their luck and gain points they may not have expected. Or, if on the contrary, they will begin to drop points they expected to gain.
The Expected points statistic is not some game changing statistic that will influence the footballing world forever. It is rather a way to better contextualize a team's performance throughout a certain period of time. It is used as a jumpstart for other more comprehensive analyses to take place within a professional team setting. The same way xG contextualizes which team played better in a match, xP does the same over a longer period of time.