How Game States Shape Match Statistics
What is a "Game State" and how should it influence the way we view key football statistics? Today on the blog Mark Taylor explains.
The Premiership is one of the most stylistically diverse leagues in the world. In the recent past, long ball teams, such as Stoke and Bolton have happily coexisted with more elegant exponents of the beautiful game, typified by Arsenal. Over the broad brush of an entire season, it is reasonable to conclude that the latter owed much of their points gathering prowess to creative artistry in the midfield, whilst the former relied more on imposing a physical style and an over reliance upon set pieces and corners.
However, an increasing appreciation of the importance of the current score line, usually referred to by the catch all phrase of “game state” has led to the realization that within each sides preferred style of play there is a degree of flexibility. In short, there are situations when even the most cultured of sides are forced to play much more direct football.
What Is “Game State”?
Game state is most easily understood in terms of the current score. A side that leads often becomes more defensive as they try to protect their advantage, especially as their opponents increasingly push forward in search of an equaliser. Therefore, the on-field actions of both sides can undergo a dramatic change due partly to the changed situation.
Goals make obvious and dramatic alterations to the current game state for individual matches, but the steady ticking of the clock also slowly alters how each side views the game. Different talent levels between sides and other factors, such as home field advantage mean that it is rare for two sides to enter a game with an equal chance of victory. One side will almost always be favoured. Therefore, even in the absence of goals, the pregame favourite will often be more eager to win the match to claim a more representative share of their pregame expectation, especially as the match wends its way to a scoreless conclusion.
So, game state is constantly shifting and with it the kind of things each side is attempting to do on the pitch.
The 2011/12 season, where data was released by Manchester City, still provides the most accessible and thorough collection of football data available, so the following examples relate to that season, although the general principles are universal.
Crosses provide the starkest example of how this most rudimentary of tactic grows in importance and use for virtually every side as they attempt to retrieve what they consider to be an unsatisfactory current score line. Each and every side which played in the 2011/12 EPL played proportionally more crosses into the box than was usual for them in matches where they trailed, compared to games where they held the advantage.
Tottenham, in their last season under Harry Redknapp exhibited a strong and persistent tendency to cross more often in games where they struggled. In the plot below, the vertical axis represents the volume of crosses made by Spurs and the horizontal one uses a composite figure to describe the average game state they experienced over the course of each match, where the higher the figure the more comfortable Spurs were within the game.
If you were lucky enough to be doing well against Spurs in 2011/12 and this could also include holding them to a draw at White Hart Lane late in the second half, then you could reasonable expect the host to go to one of the league’s default responses and bombard the box with proportionally more crosses than usual.
Crosses appear to be a universal default approach, but any attacking or defending trait that is strong enough to be identified at an individual game level for individual teams can provide an insight into the tactical approaches and alterations that managers habitually make during matches to improve or maintain a position.
To pick some particularly strong team traits from 2011/12, Spurs, along with Swansea, as well as crossing more, also attempted statistically significantly more dribbles when they wanted to improve their game position. Arsenal hit more long balls when they were in a good game state and, by contrast Stoke went longer more often when they were trying to improve theirs. Two sides were using the same tactic, but possibly to very different aims. Stoke probably employed the long ball as a primary means of scoring and Arsenal used it as a counter attack/pressure relieving tactic.
Swansea’s quantity of final third passes attempted in their inaugural Premiership season also appeared intimately tied to game state. When they felt comfortable within a game Swansea retreated into their own half, but when chasing a result they ventured much further up-field. In common with Spurs, such was the rigidity of Swansea’s passing approach, their final third passing frequency characteristic is starkly displayed, even on a game to game basis, where short term random variation and individual players performing unexpectedly badly can have an overwhelming effect on intent not being achieved in reality. Swansea’s passing in 2011/12 was undoubtedly skilled, but also largely predictable.
Everton shared Swansea’s passing preference in 2011/12. They too were much more active in trying to find a colleague in and around the opponent’s penalty area when they were doing badly and also in keeping with the Swan’s, they were happy to allow their opponents more freedom deeper down the pitch when Everton led.
A game between these two adherent’s to a game state defined passing approach, therefore could produce a textbook example of how a combination of score line and time remaining can shape what we are most likely to see played out on the pitch.
Everton vs Swansea: December 2011
The December game at Goodison, was a closely fought affair. Everton were the strong pregame favourites and a slightly better than 60% chance of winning the game gave them a predicted long term, average points haul from such a match of around 2.1 points. Swansea, as the visiting underdog could only expect to return long term with an average of 0.68 points.
The first hour was goalless and Everton’s points expectancy for the game slowly leaked away, as a home win became gradually less likely and a draw more so. Swansea, by contrast saw their expected points average gradually rise from the 0.68 points they could have expected to gain before a ball was kicked.
Just prior to the only goal of the game, scored by Osman on the hour, the hosts, Everton were in a position from where they would likely gain, long term just 80% of their hoped for average at kick off time, compared to Swansea whom now enjoyed a likely, long term haul that was equal to about 125% of their pregame total. The match was going well for Swansea and relatively poorly for Everton, but that was about to change.
Osman’s goal dramatically changed the game state, that had until now been modified merely by time elapsing. Once goals joined in as well, game state for Everton leapt in unison with Swansea’s falling.
Over the course of the game, Everton had experienced an hour of mildly disappointing game states, followed by 30 minutes plus injury time of extremely good game states. Swansea had seen the opposite. If we now go to the 442 Stats Zone, we can see how each side reacted in trying to shift the game state and ultimately the result more in their favour.
Of the two sides, Swansea was the side which showed the least flexibility over the season in how they dealt with changing game states and this one dimensional, yet effective outlook can be seen, even in a single game.
Over the whole match, Swansea attempted 110 passes into the final third and during the first hour, when a draw was an acceptable result, they carried out just 41 such passes. The hour represented 63% of the available playing time and Swansea had used that time to play just 37% of the final third passes they would attempt over the whole game. When Osman’s goal left them chasing the game, they played the remaining 63% (69 passes in total) of their total final third passes in the remaining 37% of available playing time.
Swansea’s Lack of Adventure In Favourable Game States in the Opening Hour.
Swansea’s Increased Urgency In Unfavourable Game States in the Final Half Hour.
Swansea’s entire footballing philosophy during their time spent playing in the Premiership, has laudably revolved around passing and ball retention and they invariably were towards the head of the possession tables. Statistically, Everton appeared much less one dimensional, but they also modified their desire and ability to pass in the final third as their game state, first gradually and then dramatically, altered.
In the first hour of slowly deteriorating game state, they attempted 73% of their final third passes. Proportionally more in terms of the 63% of time elapsed. They then played just 27% of such passes in the final half hour, as they protected a lead for the remaining 37% of the match.
This is an idealised example, where one side in particular alters their approach in response to the current game situation, but the predictability of many team’s tactical adjustments in other areas, such as the amount of crosses attempted, highlights often the existence of a structured response from many sides to different match situations.
The predictable reaction of sides, such as Swansea means that opponents can usually be confident that they know what to expect from the Welsh side. It is the sides which fail to produce neatly correlated on-field action charts in different game states, usually the highly resourced teams with deep squads, who can produce a multifaceted response to protecting or retrieving a game situation that are likely to cause the most problems to opponents.
Game states pervade every minute of a football game. They influence a team approach and therefore, also dictate the kind of things that individual players are being asked to do for the benefit of the side as a whole. In short, game state provides the context that helps us to more fully understand all of the newly available statistics that are now ready to view even in real time.
Read more of Mark's work on his The Power Of Goals blog
And follow Mark of Twitter: @MarkTaylor0