The inspiration to this discussion came from the Tottenham v Arsenal match last weekend, where Arsenal heavily domianted possession (68% v 32%), shots (12 v 6) and sent in 34 crosses. Despite seemingly dominating the on ball metrics and appearing to have been unlucky to come away losing 2-0 if you only looked at the post-game stats, watching the game tells a different story. Are there any ways of measuring the dominance of a match and do they require the ball?
Arsenal didn’t ever really seem like scoring and Spurs didn’t ever really feel under threat of conceding. It’s worth mentioning here that for the majority of the game Spurs were winning after scoring with their first open play shot, which means that Spurs didn’t need to overextend themselves. From then on the responsibility for trying to attack fell to Arsenal, whilst Spurs could soak up pressure and hit on the counter if they so choose. Which they of course did.
Spurs had a game plan and their intentions were clear, they also executed this game plan extremely well appearing comfortable the whole game. Arsenal maybe had a game plan and their intentions were not so clear, it’s not clear if they executed it and arguably they did not since they lost the game whilst appearing desperate. Intent or execution of intent are potential candidates for measuring dominance.
So in this particular match it seems the difference was the defending team had a game plan, executed it and so were comfortable. This is different to a potentially similar scenario where a possession heavy team is creating lots of chances and seems to score at a will. Think Manchester City, Barcelona or Bayern Munich against lower skilled opposition. They have a game plan, execute it and are comfortable when they’re playing at their best.
This highlights the distinction between traditional box-score statistical dominance and the eye test. Controlling the ball doesn’t mean you are controlling the match, rather it’s the control of space on the pitch and particularly the areas of importance.
For example without the ball:
- Spurs were winning
- The only way they would lose is by letting Arsenal create chances
- That’s most likely if Arsenal get the ball near Spurs’ penalty area
- So Spurs decided not to let them do that (very successfully)
Or with the ball:
- Barcelona are controlling possession and are actively trying to score
- Perhaps to go ahead or increase their lead
- They need to create chances
- To do so they need to get control of the ball near the opponent’s area
Both scenarios end up with the team that can control the area on the pitch they want to seemingly dominating the game.
I feel like a flow chart would be appropriate, something like this makes sense. There are lots of options I haven’t put here since I just wanted to get the main point across. And the word dominant is used optimistically here, there are of course exceptions to everything in specific circumstances.
- Do you want the ball? YES
- Can you get the ball? YES
- Do you want to score? YES
- Can you create chances? NO
- Not dominant
- Do you want the ball? NO
- Are you conceding chances? NO
In terms of quantifying this match dominance, the questions in the flow chart roughly correspond to existing metrics.
- Do you want the ball? Gamestate and style dependent (Eg. winning, losing or drawing and possession/counter attack)
- Can you get the ball? Gamestate, style and possession dependent (Eg. pressures by area of the pitch and passes completed, attempted)
- Do you want to score? Gamestate dependent (Eg. winning, losing or drawing)
- Can you create chances? Gamestate, style and space dependent (Eg. chances created, xG, shots, passes into the penalty area)
- Are you conceding chances? Gamestate, style and space dependent (Eg. chances conceded, xGA, shots against, passes conceded into own penalty area)
Notice all of these questions are gamestate dependent, which shows how important it is. Gamestate usually leads the game plan of a team, if you’re losing you want to score and if you’re winning you want to not concede. Gamestate is easy to quantify, so could be used as an approximation for intent or combined with a style metric to further outline what a suggested game plan for a team would be in a particular scenario.
The other dependencies involve possession, space and style. Possession could be important, but not on its own. Pitch control models are great at understanding which team controls which area of the pitch at any given point of a match. You could go a step further and introduce a possession quality weighting to create an possession value model, which would reward control of the pitch closer to areas of importance.
Combining these two together would involve creating a time series of a match that weights the pitch control model or possession value model of each part of the match by the gamestate. For example, if you are winning, then the possession value of the ball near the opponent’s goal isn’t so high since you don’t need to score. If you are losing then the possession value in the opponent’s half is much higher, as you need to generate chances to get back into the game.
This is a really long way of saying that if you can execute your gameplan successfully and comfortably, arguably you are the dominant team in that period of the match. And just because you have the ball a lot, if you can’t do anything with it then it’s not that great, I’m sure we can all agree there.