Football stats definitions
Explaining the popular metrics developed by the football analytics community.

In this first football article, we place words on the numbers of football analytics. The first part of the article gives the actual definitions, while the second shows an example for the 2012-13 Premier League season.

Eventually, I play with the numbers to understand some surprising facts about the relegated teams of last year.

Definitions of metrics

When I searched for football numbers on the Web, I found lots of blogs using many abbreviations to describe the stats of the game. I found a list of definitions on the StatsBomb blog, but I wanted to do my own list to see whether I understood everything correctly. So I summed up all the metrics for which I discovered a clear explanation in blog articles.

Here are the definitions:

Metric Abb. Meaning
Points Per Game PPG actual result
Total Shot Ratio TSR game dominance
PDO PDO relative shot/save efficiency (regarding other teams)
Goal Ratio GR goal scoring ability
Shots On Target Ratio SOTR ability to create serious attempts
Conversion Ratio CVR ability to score goals from attempts
On Target Conversion Ratio OTCVR ability to score goals from serious attempts
Save Ratio SVR ability to save serious attempts from opponent
Cards Per Game CPG fair play

Then, to compute these values we use the corresponding formulas:

Metric Formula
PPG mean(points_for)
TSR shots_for / (shots_for + shots_against)
GR mean(goals_for)
SOTR shots_on_target_for / (shots_on_target_for + shots_wide_for)
CVR goals_for / (shots_on_target_for + shots_wide_for)
OTCVR goals_for / shots_on_target_for
SVR (shots_on_target_against – goals_against) / shots_on_target_against
CPG mean(cards_against)

Premier League (2012-13) example

To show the real appearance of the data, let’s apply these definitions to the last Premier League season (2012-13). The dataset I use is a sample of different databases from Soccerway and Football Data UK. For more information about the process, see the programming articles about football stats extraction. Below is a table with all the values.

<td>Manchester City</td>



<td>Tottenham Hotspur</td>



<td>West Bromwich Albion</td>

<td>West Ham United</td>

<td>Swansea City</td>

<td>Norwich City</td>


<td>Stoke City</td>


<td>Aston Villa</td>

<td>Newcastle United</td>


<td>Wigan Athletic</td>


<td>Queens Park Rangers</td>
1 Manchester United 2.342 0.542 1.154 2.263 0.579 0.177 0.323 0.831 1.158

You can play with this table by many ways, and ask yourself some questions about what happened last season:

  • Try to guess Stoke’s spot in the fair play ranking
  • Reading’s conversion rates are excellent (3rd CVR, 2nd OTCVR). They scored 1 goal in 4 shots on target. Despite this, they got relegated. How is it possible?
  • Many people spoke about a miracle for Wigan’s cup title last year. Indeed, it was the first time we saw an English first division team being relegated and winning the cup in the same season. Were they as weak as it seems? Was their situation comparable to the ones of QPR or Reading?

The last two questions are open ones, and they do not have a unique answer. However, we can use the numbers of the table to find some lines of research:

  • Reading actually were very good finishers, but they were outmatched very frequently by the opposition (lowest TSR) and thus were given very few opportunities to shoot/score in most matches. Further investigation of the data table also shows that Reading were 2nd-worst in SOTR, which suggests shot quality problem.
  • Wigan’s record in the league seems weak, with only 9 wins and 20 defeats. On the one hand, they were less effective (lowest PDO) than any other team in the league, and also encountered big difficulties saving opponent attempts (lowest SVR). On the other hand, they were league-average in terms of game dominance (10th TSR), attempt creation (9th SOTR) and attempt conversion (12th CVR). Unlike Reading’s inability to shoot, what cost Wigan their Premier League spot is maybe not their attack, but the lack of efficiency of their defense.

The solutions I proposed are just lines of inquiry and should be verified with tactical analysis, and further statistical analysis. However, we were able to find them without watching a single game, only by having a look at the the metrics.