Assessing the importance of the FDR — Fixture Difficulty Rank — in Fantasy Premier League
Fantasy Premier League admins are now providing a new data point for each game week, with the intent of helping FPL managers build their squads based on the difficulty of the team’s fixture. But, is it really helpful? If helpful, how can it be used as an advantage in building squads?
Let’s start with the definition provided by FPL help articles:
The Fixture Difficulty Rank (FDR) is based on a complex algorithm developed by Fantasy Premier League experts. A set of formulas process key Opta data variables, along with each team’s home and away form (past six matches), in order to generate a rank for the perceived difficulty of each Gameweek opponent. The FDR is designed to help FPL managers plan their transfer and team selection strategy and is reviewed on a weekly basis.
It sounds pretty cool here, but it is important to critically assess whether FDR influences the points yield from players and teams in a particular game week. For that, I have built some spreadsheets and conducted a few analyses.
FPL Game Week 2 — FDR Basics
I don’t have data on GW1 for FDR, but I was able to tabulate enough data on GW2 to run this analysis. Once a GW is completed, it is impossible to go back and acquire FDR data for that GW. So, I have to document the FDR data before the GW matches are played.
You will notice that the FDR scores are not distributed evenly. In GW2, there were 7 teams in FDR1, 3 in FDR2, 6 in FDR3, 2 in FDR4, and 2 in FDR5.
How to Connect Points to FDR
Next order of business is to develop a methodology to relate points earned in GW2 to FDR scores. For that, I have chosen the following metric for calculating the points:
Total Points Earned by Top 5 Players in Each Team
The choice of 5 seems arbitrary, but there is actually a logic behind it. Different teams end up with different number of substitutions, so it is not really meaningful to sum all point-earning players for comparison. Also, FPL managers try to choose from highest scoring players, and the Top 5 in a single team constitute a list of 100 players overall, which seems more than enough for most managers to conduct their picks from. In summary, my decision was to design a metric that would be reflective of a team’s overall performance in that week, without getting into the complexity of managing too many players. Top 5 players seem to make a good case. In the future, I might try out the same set of analyses with Top 3 players from each team and measure the sensitivity of the studies to the player selection.
The chart above is very insightful even before getting into the discussion of the impact of the FDR scores. In GW2, there was a clear chasm between Watford (the best team among the worst 10 teams) and Middlesbrough (the worst team among the best 10 teams). There is no such gap between any other teams compared to the gap between Watford and Middlesbrough. Would an FPL manager use this information alone, without the FDR data, for the Game Week 3 picks?
Now, the next order of business is to bring together FDR information and the Top 5 Player Points information, and assess whether FDR scores had power to predict points yield in GW2.
FDR vs. Points — is there a relationship?
Based on the scatter distribution above, there is no clear indication that lower FDRs lead to higher Points Yield. Except FDR2, other FDRs don’t seem to be indicating better or worse Points Yield. Could we bet on the reliability of FDR2 to produce better yields in GW3? I guess we will be on the lookout for that…
Applying Correlation formula on this 20-point data set yields a Correlation rating of:
-0.060501
What this means: The correlation between FDR and Points Yield in GW2 is very weak or non-existent.
Similarly, a linear estimation study produces a model with a very low R2. This means that, a very small part of the variability in Points Yield can be explained by the FDR metric.
The Verdict
Based on GW2, one can not make a case to use FDRs to maximize Points in that game week. This is just based on one game week, and it is just game week 2, early in the season. Quoting FPL help article, the “FPL Experts” seem to be using complex formulas to assign FDR scores to teams, but those scores don’t seem to be useful at this time. I am really curious to see if this picture changes with GW3.