FPL Analysis

I have always been a big fan and a player of this beautiful game for at least the past 6–7 years.

FPL is probably the next most interesting thing after the Premier League. It is FPL that takes us as close to the manager’s seat as possible. So, who is the cheapest alternative to Sanchez?, who is an able deputy to Costa?, when does Harry Kane peak?… These are the questions that the best FPL Managers have answers for.
I have tried to have a crack at it with statistics and Machine Learning. So let us first look at some interesting insights that I have been able to observe from the data. It could be called as Exploratory Data Analysis (EDA).

I have always felt that it is quite easier for players to score points against easier oppositions when compared to a top of the table clash (But mind you, Premier League is unpredictable)

The above graph suggests that Hull City, Swansea, Sunderland and Watford are some of the easier teams against whom players tend to get more FPL points. I find that fair enough since they seem to struggle this season. But wait, Sunderland have always been known for their tough relegation battle and once we reach Game week 30, I guess it would be the time for FPL managers to make Defoe their captain (Not a statistical prediction though :P)

Next would be Home advantage. I’m sure that every single premier league side would fear playing WestBrom or Swansea in an away fixture cause such is the impact they have in their home games with such passionate fans cheering them.

Again, the above plot also suggests that average points obtained by players in a home fixture is slightly more when compared to an away game.

I have always felt that the midfield as a whole tends to get more points when compared to other positions because midfielders carry both defensive and offensive responsibilities and are hence awarded for both. The same is illustrated in the graph below.

The ICT (Influence, Creativity and Threat) index seems to be an important metric. The index seems to be a more accurate attempt at evaluating the impact a particular player has had on the game. Hence, a midfielder with a higher ICT index, can be trusted to give good points in the near future. More on the ICT index later.

Next up, I would like to bring more insights for each position.

The above graph shows the ranking of goal keepers in terms of average points obtained by them in each game. We can see that there are many keepers from smaller teams in the Top 10. Tom Heaton, Lee Grant and Ben Foster tend to get a lot of points but they have less number of clean sheets when compared to keepers such as Courtois, Lloris, De Gea etc.. who are from stronger teams. The smaller team keepers get more points, courtesy the saves that they make during the games and we can see that Tom Heaton, Lee Grant and Ben Foster make more saves in a game when compared to keepers from top teams (Let’s not forget that top teams have a very good defence and they prevent oppositions from taking many shots at the keeper). I would suggest to have Heaton or Foster in the team and save up some cash for better outfield players and the existing FPL managers are following that. Close to 40% of the FPL managers have started transferring Heaton and Foster into the team.

Let’s look into the defenders now:

When teams like Chelsea, Tottenham and Arsenal are doing well, it isn’t a surprise to see their defenders in the Top 10 in terms of average points per game. But what about Charlie Daniels, Baines, Winston Reid and Coleman? The second graph suggests that these defenders have a tendency to attack and have a habit of scoring or playing a pivotal role (assist) in the goals that their teams score. My suggestion is would be to pick them as the 3rd and 4th defenders in the team as wouldn’t cost much but the ROI is really high and we can have defenders from top teams for the purpose of clean sheets. Teams like WestBrom and Bournemouth tend to score a lot through their defenders.
So defense from a stronger team and a goalie from a relatively weaker team would be a good idea, say Walker and Heaton.

Midfield maestro’s are up next

I have always felt that we need something more than the goals and assists to rate them. The above 2 graphs are an example for the same. Like mentioned earlier, ICT (Influence, Threat and Creativity) is a very important metric for midfielders and the above 2 plots explain that. In the first graph we can see that Mane and Hazard tend to have the same value for average goals per game and Mane seems to be better in terms of average assists per game but still Hazard yields better FPL points. From the 2nd graph we can see that Hazard generally play more key passes in a game and has a better ICT index when compared to Mane. I can assure that midfielders with better ICT index and key passes will get you more FPL points.

The Goal hunting wizards and poachers are up next:
Every single FPL manager will want to have strikers who keep scoring and assisting in every single match. There are other traits that are important for strikers. Yes, I am coming back to ICT again.There are few other metrics that define a good striker and help in winning more FPL points. Winning goals scored and Big chances created (Greater the chances created, greater is the possibility to score) are some of the other important metrics.

We can see that good strikers who score more goals tend to create more chances also score a lot of winning goals
We all know football is highly unpredictable and numbers and statistics can never win the beautiful game. But we can always make better and meaningful decisions with the help of numbers.

If you are interested in knowing about the statistical model and the metrics used, e-mail me at : prashanthsekar1@gmail.com