Fantasy Football

Performance analysis of Goalkeepers in Fantasy Premier League (FPL) and assessment of where to find value

Following a recent post I wrote about three-way defensive rotation strategy in the Fantasy Premier League (FPL) game, I was asked on Twitter to do the same analysis for two-way goalkeeper rotations, the results of which can be seen by clicking here:

In contrast to the three-way defensive study, the goalkeeper research did not eliminate premium (e.g. expensive) goalkeepers from the analysis. I took this decision because the difference between a premium and cheap goalkeeper in FPL is the matter of just ¬£1m, and I know a lot of managers favour putting money into their goalkeeping positions. However, the definition of ‘favourable’ remained the same; a ‘good’ fixture was one against a team that I determined (relatively subjectively) to have weaker attacks, thus increasing the potential for clean sheets.

The problem with this is that the underlying assumption is that clean sheets are the ultimate goal for goalkeepers in FPL. Of course, they do deliver the most points, and are a significant contributor to the bonus point system too, but last season’s most successful FPL goalkeeper, Tom Heaton of Burnley, was perceived to be so good because of his ability to rack up save points against even the strongest attacks in the division, most memorably against an exasperated Zlatan Ibrahimovic at Old Trafford. The basic premise of my analysis was that goalkeepers should be avoiding good attacks, but I thought it was worth testing this in case it wasn’t true.

The question I was looking to answer was:

If my goalkeeper is facing a top side, is it worth dropping them from my starting line-up in favour of a goalkeeper with a better fixture (e.g. against a weaker attack)?

Methodology

I took the FPL data from the last three seasons (2014/15 – 2016/17) and divided the teams into three tiers:

  • Top six finishers
  • Mid-table finishers (7-13)
  • Bottom seven finishers

In doing this, I created a basic assessment of which teams were strong, weak, and mid-range. I then summed the totals of a number of metrics for each group when they played the other groups; for example, the total number of saves made by goalkeepers of top six teams when they played other top six teams, and when they played mid-table teams, and when they played bottom seven teams.

I then created a baseline which was predicated on absolute equality. This assumed that all goalkeepers in every game would score the same number of points, make the same number of saves, keep the same number of clean sheets, etc. Considering the different sizes of the groups, the share of events was distributed as follows:

Capture III

These percentage shares were the baseline, or where we would expect to see the events distributed in a vacuum, where there are no external factors. Using this baseline, we are then able to see which group of goalkeepers are over- and under-indexing against the baseline.

Analysis

The matrices below show the percentage points over and under the baselines for four metrics: FPL points, FPL bonus points, saves made and clean sheets.

Capture

The main points of interest are:

  • Clean sheet and save distribution conforms to conventional thinking. The better the opponent, the more saves and fewer clean sheets a goalkeeper can be expected to produce, and the better the team the fewer saves and more clean sheets a goalkeeper can be expected to make. Tom Heaton is likely to make more saves against the bigger teams, but David de Gea is likely to make fewer than him against those same teams.
  • Points distribution shows that bottom seven goalkeepers – the real budget options – will score a lot fewer points than their baseline against the top 13 teams, but when playing a bottom seven team they are just as effective as a top six goalkeeper. In other words, if you’re playing a team that is in the relegation zone, you are not going to get more points from Hugo Lloris than from Matthew Ryan (assuming Brighton are also a bottom seven side come the end of the season – sorry Seagulls fans).
  • The over-indexing of bottom seven goalkeepers vs. bottom seven teams in points (+1.62) is the one outlier from the data. Based on the trends seen in the clean sheets and saves data, and elsewhere in the points data we would expect the value to have been between -2.01 and +0.97 percentage points against the baseline; considering the performance of top six and mid-table goalkeepers against their contemporaries (+0.12 and +0.15 percentage points respectively), we should have expected something around the baseline, but instead they over-indexed significantly.
  • The driving force behind this appears to be the massive share of bonus points attributed to bottom seven goalkeepers when playing other bottom seven teams. In this metric, they over-index by +8.29 percentage points, taking a total of 19.3% of all bonus points awarded to goalkeepers over the last three seasons.

This discrepancy in bonus points for goalkeepers in bottom-of-the-table clashes is the main surprise in the data. It is even more surprising when we consider that they are not more likely to make more saves in these games than they would against the top teams, nor are they more likely to keep more clean sheets against them than the better teams. The only logical explanation I can come up with at this point is to suggest that there are fewer outfield players performing exceptionally in these games, thus increasing the goalkeepers’ relative standings on the pitch.

It is clear though, from the following chart which shows the points per million distribution, is that this anomaly is a key driving force behind the bottom seven goalkeepers providing the best value for money when they play their contemporaries.

Capture I

The analysis of points per million shows:

  • You will get the most value from a goalkeeper when they play for a poor team and are playing another poor team (0.91 ppm).
  • However, irritatingly, they provide the worst value for money when playing mid-table teams relative to goalkeepers from the top 13 sides (0.67 ppm, lower than 0.72/0.73 ppm for better teams).
  • In terms of value, you’re not going to find much difference between top and mid-table goalkeepers, regardless of who they are playing.
  • The least amount of value is found when a goalkeeper is playing a top six side, and it actually makes little difference whether they are a top goalkeeper themselves or a relation candidate (0.64-0.66 ppm)

The most disappointing point here from a FPL manager’s perspective is the second bullet point. If the 0.67 ppm a bottom seven goalkeeper gets from playing a mid-table side was closer to 0.72 ppm it would bring them up to par with all other goalkeepers when playing top 13 sides. It would then highlight the 0.91 ppm they score against relegation candidates as the significant outlier and the driver of true value. Unfortunately, the data shows that to take advantage of this potential bonus point bonanza, you must take the hit of lower value when playing mid-table sides.

Summary

I approached this analysis aiming to discover:

If my goalkeeper is facing a top side, is it worth dropping them from my starting line-up in favour of a goalkeeper with a better fixture (e.g. against a weaker attack)?

The answer to that is an unequivocal ‘yes’; both in terms of points scored and points per million, it is much better to have your goalkeeper facing the weakest team possible. It also is of benefit in terms of total points scored to have a goalkeeper from as good a team as possible. The original suspicion that it might be better to have a weak team’s goalkeeper going up against a top side in order to rack up save and bonus points has proved, at an aggregated level, unfounded.

However, FPL managers are always on the look out for value, and it turns out it isn’t anymore beneficial from a points per million perspective to have a top six goalkeeper than a mid-table goalkeeper. Yes, they will score more points, but the increase is proportional to the cost regardless of the opposition.

The bottom seven goalkeepers are the most interesting because that’s where a lot of us look for value, and in terms of ppm you are no better or worse having them play a top six attack than you would be with a goalkeeper from a better team. But, and here is the rub, you will be worse off when playing a mid-table team, but much better off when playing a bottom seven team as they bring in the bonus points.

With this in mind, it seems the best compromise may be to select cheap goalkeepers from teams you expect to finish mid-table, either as number one / number two combination (Foster/Myhill, for example) or as a rotating pair. However, the information is open to interpretation, so each manager is going to need to make a compromise based on how they read these numbers and the prioritization they place on goalkeeper budgets.

Advertisements

2 thoughts on “Performance analysis of Goalkeepers in Fantasy Premier League (FPL) and assessment of where to find value

  1. Hi, great article here, seeing your 2 Way GK Rotation and 3 Way DEF rotation, i was thinking maybe it’s all about rotation in FPL, with 3-4-3, we can have 3 Way rotation with 2 premium DEF + 2 Way Rotation with budget DEF, combine with 2 Way GK Rotation, it’ll make squad selection much easier in defence. But, i was thinking maye you can make 2/3 way rotation for Captain, basically the same with 2 Way rotation for GK, but it’ll check with the team that you think will most likely leaking goal. With that we have 2-3 player that is nailed on, and another 5-6 for flexible bandwagon

    Cheers,

    Like

    1. Personally I like to have the big, heavy-hitting players as transients in my side, with the captain especially a player that ‘chases the form.’ I believe the value of a defensive / GK rotation is it reduces the need to spend transfers on the fringes of your squad, instead being able to focus on bringing in the explosive players as and when form dictates.

      Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s