Alternative league table: The Relative Performance

In our series of alternative league tables, one element was clearly missing. We’ve started with the classic league table and added Total Shots Rates to identify high quality, sustainable performance. After all, a dominance in the number of shots created is the best parameter around to establish projected future success in football.

After the that, we’ve looked the same classic league table, but with added PDO rating this time. PDO, as regular readers will no doubt be aware by now, consists of a team’s saves percentage and conversion rate. Both of these tend to regress to mean with more matches added, so high PDO’s reflect non-sustainable high league standings, if performance levels in terms of TSR don’t increase. Low PDO’s reflect low league standings that are expected to improve over time if the same performance in terms of TSR is continued.

Using information from the TSR table, we use historical data to project the number of points that teams are expected to take from the remaining matches of the season and construct a projected final league table. This reflects what the situation at the end of the season will be like if teams are expected to continue their level of performance in terms of TSR and regress to mean levels of PDO.

A new addition here should be the relative performances league table.

This table below is fairly simple and the idea is not new. In a short amount of time, the same concept originated on different places, with Scoreboard Journalism being the first to write about this method. Using bookmaker odds, one can calculate the number of points a team is expected to take from a match. The total number of expected points taken is then compared with the actual number of points won and the fraction of these is the relative performance. An example of calculating a number of expected points and computing the relative performance is found below the actual table.

Let’s take the match odds from Match Day 8 game, Twente – AZ. Averaging the odds for most larger betting companies would see the odds for a Twente win at 1.71, for an AZ win at 4.49 and for a draw at 3.65. This indicates a 1/1.71 chance of a Twente win, a 1/4.49 chance of an AZ win and a 1/3.65 chance of a draw. These chances are 58% for Twente, 22% for AZ and 27% for a draw. The attentive reader immediately notices that these don’t add up to 100, and that has to do with the fact that bookies make their profit too!

So we have to degrade the odds a bit to make them add up to 100. Twente’s true odds would be 58 / (58+22+27) = 54%, AZ drops down to 21% and the draw to 25%. Now it adds up to 100. The expected number of points for Twente is 3* 0.54 + 1*0.25 = 1.87. For AZ this is 3*0.21 + 1*0.25 = 0.88.

Now, we add each team’s expected points over all matches played and compare it to the actual number of points won. The green and red bars present the fraction of expected points that a team has actually won. If this number is over 100%, a team outperforms the bookies odds so far this season. If this number is below 100%, a team under performs compared to expectations.

 

Data provided by Infostrada Sports.

Leave a Reply