We've talked previously on this site on the importance of performance at certain game states. Think of a tied game state (evens) as the most crucial, followed closely by performance at a crucial game state and followed then by the slops (blowouts).
Talk of game states has been evident on this blog since June when I looked at EPL strikers, but it's only since I was able to record teams' total shots, shots on target and goals at certain game states that things have begun to get interesting.
You guys know where to find previous articles on game states, and if you don't I want this piece to be a broad overview of the uses and fine detail which becomes evident when we look at a teams performance by the various game states.
For this example of how a large amount of information that can be teased out from game state data I shall use Arsenal as an example. A league wide, detailed article will appear once we have more completed fixtures. Oh, and when I have the chance to breakdown the W's numbers (West Ham, Wet Brom and Wigan)!
Arsenal Total Shots By Game State
Nice and all, and it is usefull to look at the raw numbers. Here is a very basic viz to help us a little:
Arsenal's Goals For And Against
|Goals For||Goals Against|
These are Arsenal's goals for and against numbers at every game state they spent time at this season, no real need to visualize these numbers again. But what we can do now that we have the shots totals and the goals totals are things like this:
Scoring%, Save%, Pdo% and TSR at certain game states
It is from the two very basic data sets listed previously (shots and goals) that we are able to find pretty good information like this,
Note Arsenal's scoring% at Evens(tied Game State), or their strong TSR number, again at Evens.
Now, if we strip away some of the the less frequent game states and focus on the close game states, those being -1, Evens and +1, we can reduce some of the noise and focus on the most frequent actions.
81.25% of the total shots that have taken place in Arsenal's fixtures have taken place at the Close Game States.
Close game State
|Shots For||Shots Against||Differential||Scoring%||Save%||PDO||TSR|
|Close Game State||308||225||83||10.71||89.34||100.05||57.78%|
Arsenal are outshooting the opposition at CGS and their PDO would indicate that Arsenal's performance is neither lucky nor unlucky.
By grouping these close game states together we protect around 80% of the sample whilst also evaluating what is really going on with a team when the game is in it's most crucial and frequent states.
Performance at Evens (Tied Game State)
If we refine the data even further and settle on just teams' performance at a tied game state we are left with the most crucial game state with 48.57% of all game time having been played at evens in the PL in2012/13.
A potential issue with focusing only on a tied game state is that just 48-50% (ish) of the shots sample remains. Relevance v sample size concerns?
Arsenal in 2012/13's PL
As mentioned this is just 50% of the data. Even strength/tied game state performance should be the most accurate information we can gather on teams' performance, but I thing Close Game State tells us some things too.
A lot of of the numbers and methods above may look complex, but infact they are simple enough. It may be worth noting that not all of the above information may be useful, but I personally feel much of it is.
Thoughts on Close GS and Game Tied GS?
As a quick bonus feature I thought I'd borrow some of the above Game State methods and apply them to two players: Walcott and Wilshere.
Going really, really deep into the data we can look at specific players and how they have performed at certain game states. I'll just post the raw numbers, not much need for viz's at this point.
Walcott numbers when he is on the field of play and at no other time.
|Walcott shots||Share of total shots||Walscott SoT||Share of SoT||Gls||Shots Scoring %||SoT scoring%|
These are Walcott's shot totals and conversion %'s broken down by game state. I think there are many issues with small sample sizes here.
|Shots For||Shots against||Gls for||Gls against||On Filed Sc%||On Field Save%||On field PDO||TSR|
A slightly different take here for Jack Wilshere. These numbers are solely for when Wilshere was on the field of play and help us understand some of what was happening at certain game states.
I particularly like On Field PDO and TSR although the players personal ability is trapped within the teams ability to a great extent.
I will probably release individual teams' game state data if and when I can between now and seasons end. June will probably see a lot of words on this topic when I look at this topic on a league wide basis.
If there is any interest in this topic I may well dump the raw data someplace and let people pick over it.
Any thoughts or reactions, please let me know?