Should clubs gather their own data for analytics?
Ted Knutson of Statsbomb tweeted an interesting thread yesterday:
Football clubs seem to have the mentality that if they are using the same data as everybody else they won't gain a competitive advantage from it.
There is perhaps a kernel of truth in this for scouting data, but only a kernel. Firstly not everybody is operating in the same market, and secondly not all clubs have the same requirements for a player. But yes if the new Messi pops up it will very clearly and very obviously be viewable by everyone.
But for a club looking to use analytics this view seems very short sighted. To me it seems to miss the point of gathering the data in the first place.
Whilst player and opposition scouting and post-match analysis are arguably the most fun parts of the rise of analytics the main day to day competitive advantage is surely through analytics informed coaching.
And if you are going to be informing your coaching through analytics you surely want to make sure your data set is accurate.
Your competitive advantage should be gained on the practice field not on your inhouse dataset or xG model. Unless you can find a fundamental flaw in the commercial datasets that you can exploit you are better trusting the data and using it to your advantage.
The problem, and beauty, with football is that so much rests on a single 90 minute period. Multi-million pound decisions are taken on tiny samples of data that are subject to huge swings of fortune and statistical noise.
Analytics informed coaching allows you combine the large "real life" datasets with your own training ground data.
Say I've got a young player like Dominic Calvert-Lewin on the training ground wanting to improve his finishing. Traditionally I'd get the cones out, get the reserve goalkeeper warmed up and practice, practice, practice.
But do I know that my drills are actually making him better? Do they represent real life situations with active defenders and top class goalkeepers? I've never been tackled by a cone (well I have actually but decent footballers aren't).
Is it a technique issue, a movement issue, the type of chance we are creating?
With analytics informed coaching, we could look at video data of where he was moving, the xG of the chances he was receiving and rather than (or in addition to) practising finishing we should be working on increasing the average xG of the shots he is taking.
Larger clubs could gather their own training ground data to see if their training is actually working. Is Calvert-Lewin now making the correct runs to increase his xG from each shot? In a game you might only get this scenario once or twice, but in training you can replay the move 100 times. When matchday comes you can see if the training has paid off. Has he made the run you worked on? Regardless of whether the pass made it to him or he tucked the chance away, that is the variance you can't control. All you can do is increase your chances.
To get that detail you need to use a big data set, and it doesn't matter if everyone else is using the same one.
The value is in how you work with the data not the data itself.
There is perhaps a kernel of truth in this for scouting data, but only a kernel. Firstly not everybody is operating in the same market, and secondly not all clubs have the same requirements for a player. But yes if the new Messi pops up it will very clearly and very obviously be viewable by everyone.
But for a club looking to use analytics this view seems very short sighted. To me it seems to miss the point of gathering the data in the first place.
Whilst player and opposition scouting and post-match analysis are arguably the most fun parts of the rise of analytics the main day to day competitive advantage is surely through analytics informed coaching.
And if you are going to be informing your coaching through analytics you surely want to make sure your data set is accurate.
Your competitive advantage should be gained on the practice field not on your inhouse dataset or xG model. Unless you can find a fundamental flaw in the commercial datasets that you can exploit you are better trusting the data and using it to your advantage.
The problem, and beauty, with football is that so much rests on a single 90 minute period. Multi-million pound decisions are taken on tiny samples of data that are subject to huge swings of fortune and statistical noise.
Analytics informed coaching allows you combine the large "real life" datasets with your own training ground data.
Say I've got a young player like Dominic Calvert-Lewin on the training ground wanting to improve his finishing. Traditionally I'd get the cones out, get the reserve goalkeeper warmed up and practice, practice, practice.
But do I know that my drills are actually making him better? Do they represent real life situations with active defenders and top class goalkeepers? I've never been tackled by a cone (well I have actually but decent footballers aren't).
Is it a technique issue, a movement issue, the type of chance we are creating?
With analytics informed coaching, we could look at video data of where he was moving, the xG of the chances he was receiving and rather than (or in addition to) practising finishing we should be working on increasing the average xG of the shots he is taking.
Larger clubs could gather their own training ground data to see if their training is actually working. Is Calvert-Lewin now making the correct runs to increase his xG from each shot? In a game you might only get this scenario once or twice, but in training you can replay the move 100 times. When matchday comes you can see if the training has paid off. Has he made the run you worked on? Regardless of whether the pass made it to him or he tucked the chance away, that is the variance you can't control. All you can do is increase your chances.
To get that detail you need to use a big data set, and it doesn't matter if everyone else is using the same one.
The value is in how you work with the data not the data itself.
Comments
Post a Comment