Quantifying Playing Styles in the Hero ISL - Part Two

Last year, I took a shot at trying to use some basic box stats to identify a coach's playing style. I wanted to repeat that analysis this year but also improve the methodology a little.

There were a couple of problems with the methodology I used last year,

1. The first issue is that I arbitrarily defined 12 categories within which each team/coach should fall. There might be more/fewer categories and the selection of that number 12 wasn't really backed by data

2. The algorithm I chose also left a lot of room for interpretation over what exactly each category represented. I think I got it right, but a more objective measure would be better.

Because of the 2 reasons mentioned above, the majority of last year's post was about creating various 2x2 graphs with different metric combinations to interpret the results. So, this year, I wanted to try a method that allowed me to compare and visualise ALL the numbers at the same time - a "Principle Component Analysis".

When you have "n dimensions" (n number of stats to compare), Principle Component Analysis can be very useful, because it's hard to see through so many numbers. PCA allows you to take 10 numbers and simplify them to 2 or more numbers.

This sort of analysis is mainly useful when you're hiring a coach to ensure that there is continuity and progress and not random, uninformed choices.

So, onto the results then. Below is a graph of every manager to have taken charge of a Hero ISL game between 2017-2019. (Robbie Keane was in charge of only 1 game and, therefore, is a big outlier).

[How to read the graph - Names that are closer together are more similar in style and the red line indicates what stat influenced positioning in which direction]

Takeaways

1. Sergio Lobera is special. Albert Roca and Cesar Ferrando are the closest to him, which makes sense. What makes Lobera different is his ability to create passing teams that are incredibly successful in attack.

2. Eelco Schattorie does what he says he will. His teams is rated very well for number of passes, touches and crosses made. Positive intent.

3. Josep Gombau, Miguel Angel Portugal are not similar in terms of style to Lobera, which is very interesting. My hypothesis is that their teams have tried to be quicker on the transition and more direct but the main difference seems to be in making a lot of passes/crosses without creating the corresponding number of chances

4. Phil Brown and Pradhyum Reddy - Phil Brown's tactics were essentially the same as Reddy's last 2 games but the model thinks they are very different. The gap between the 2 is down to the fact that Brown's side were far more successful in attack - probably owing to the fact that they had most of their foreigners either suspended or injured in the first half of the season.

5. Steve Coppell, Rene Meulensteen. Ashley Westwood, Joao de Deus and Avram Grant have sides that don't like to register big numbers when compared to their peers, which indicates that they are reactive coaches that look to pick and choose their moments.

6. The most interesting grouping is the one right in the middle - Carles Cuadrat, David James, Ranko Poppovic and Nelo Vingada. This group don't have any defining traits but their success depends a lot on efficiency, via keeping the opponents down and maximising their own opportunities. Two of them got it right, two of them didn't.

Another interesting aspect that is worth looking at is the inter-season change in playing styles. For example, how was Lobera's team different in 2017-18 in comparison to 2018-19?

Takeaways

1. BFC have regressed in attack. Albert Roca and Carles Cuadrat had almost the same teams but there was a significant dip in their attacking production last year.

2. Phil Brown and Pradhyum Reddy switch places. The 2 Pune coaches seem to have exchanged positions in this graph in comparison to the first one. I have no explanation for this, really. Weird but maybe it points to how similar they are in terms of style.

3. Miguel Angel Portugal had a terrible start with Pune. Huge divergence from his Delhi performances.

4. Josep Gombau is very similar to MAP. That continuation in style and mentality would have served the Delhi players well.

5. John Gregory's emphasis on crosses didn't work out too well. Huge difference in the way his teams played in his two seasons - mainly a dramatic increase in crosses per game and the equally dramatic drop in goals per game.

Got feedback? I'm always willing to listen, send it over to @SgtSaltnPeppa on Twitter.