• Amrit

WTF is Game State and some Germany at the World Cup ramblings



It's probably widely accepted that scoring a chance at 1-0 down is probably harder than burying the same chance when you're a goal up. Game State therefore tries to quantify this property by tracking the goal difference between the 2 teams over the course of the game. For example, if Team A scores a goal they go to a game state of +1 while Team B will go to a state of -1.


This post is mainly about how Game state affects attacking strategies but also a little bit about why Game State is probably the reason Germany didn't look as good in the World Cup.


Why is this important?


If we've learned anything from the World Cup, it's that smash and grab football, the Atleti and Diego Simeone way, is growing in popularity. World Champions France, are also great examples of how Game State can affect tactics. This video shows Antoine Griezmann pulling his team mate back into a defensive position as France are 2-0 up with 20 mins to go.


TSR (Total Shot Ratio) was a metric that people used to love before xG became so popular. TSR tells you what % of the total shots in a game was taken by each team. So if there are 10 shots in a game and Team A has take 7 of them, they will have a TSR of 0.7 whereas Team B will have TSR of 0.3. This allows us to tell which team was more threatening.


TSR is symmetric around 0. At State +1, the TSR will be the inverse of the TSR at state -1 and so on. The boxed area marks the area for which we have lot of data. Higher game states have fewer data points

As seen in the above image, TSR shows us that Teams at a negative state, ie, teams that are a goal or more down tend to create more chances than the teams that are in a positive state.


In the next image, we're plotting the Average Chance Quality against Game state and the trend reverses itself completely. Previous graph told us that Teams that are winning create fewer chances but this next graph indicates that they create better quality chances.




One final graph to look at Conversion Rate, ( just to back up the xG point ). It's easier converting chances into goals when you are in a positive state.



An important point to note in the last two graphs is that there is not too much difference between states +1 and -1. Teams are not drastically affected by going a goal down. Average chance quality and conversion for teams at -1 may be lower. but. as shown below, teams at -1 try harder than teams at any other state in the game.




Okay, so?


The mainstream conclusion as to why Germany crashed out of the World Cup even though they almost created more chances per game than most teams created in the whole tournament is that their average chance quality, or xG, was below average. That's a valid point and the below graphs illustrate this.




However I think that's a lazy explanation. I still have not found the statistical answer as to why they lost but saying that it's because of their low chance quality isn't 100% right.


I've calculated the xG for the tournament including and excluding Game State. Game state has a weak positive influence on xG. That means for teams that are trailing, xG will automatically be slightly lower. As seen in the below graph, Germany were only in a positive Game State for the final few seconds of their game against Sweden. They spent the rest of the World Cup chasing games.


Also important to note that they spent most of their time in either states 0 or -1 and as demonstrated above -1 state is when teams give it their all. Germany's Average xG in states 0 and -1 does not vary too much indicating that their strategy didn't change at all after going a goal down. This is significant because at -1 xG should automatically be lower than at state 0 but the fact that it's not significantly lower is important.


The next graph illustrates how xG for each team drops after I included Game State in the calculations. All the teams listed in and around Germany in are teams that spent a bulk of their time chasing games. Serbia, Switzerland, Brazil and Spain for example did create some good chances and probably shaded their opponents in open play but their xG was affected by the fact that they were a goal or more down for large stretches of their time in tournament.



To conclude, I hope I've shown that Germany's woes in the World Cup are either defence/tactics related and their average chance quality being low doesn't really mean that they deserved to go out. My model doesn't include defenders in between the goal and chance location but deep packed defences against Germany would have further pulled their xG down. So, their average xG being low is probably expected which is either the outcome of or the reason that their style of play is based around territorial domination.


My explanation as to why Germany crashed out is threefold,

  1. Out of the 4 goals they conceded, 2 goals came at the 32nd and 35th minutes and 2 in injury time indicating that it's probably a concentration issue.

  2. Muller + Kimmich didn't show any chemistry or combination down the right side which forced Germany to assymetrically play down the left.

  3. Both fullbacks being pushed up without a mobile DM exposed the CBs to counter attacks


Have feedback? You can find me on Twitter @sgtsaltnpeppa

  • White Twitter Icon