Every FPL manager knows the feeling of frustration following a bad week- especially while league mates seem to be enjoying all the luck. It can be very difficult to differentiate between skill and luck over a short period- though the general expectation is that this will even out over the long term.

Looking at just how far ahead the top managers in the world are you have to wonder- just how skilled are they?

After GW36 of the 17/18 season FPL Review calculated Expected Points data for selected teams and performed an analysis on 4 groups of selected teams in Fantasy Premier League. The Review Performance tool effectively performs this analysis on a week by week basis.

The groups selected for this analysis were as follows:

  • Group 1: A random selection of 50 managers
  • Group 2: A league of 10 active but non-notable managers
  • Group 3: The top 10 managers overall worldwide
  • Group 4: Legendary Manager

Expected Points: How this data was calculated

Before digging into the results and what can be taken from them, I will give a brief explanation of how these Expected Points were calculated.

First, consider the following scenario: A player has a one on one chance with the goalkeeper 15yards from goal- historically let’s say 60% of these are converted, therefore his expected goal value is 0.60xG. Should he score he is determined to be over-performing by 0.40xG and should he miss he will be under-performing by 0.60xG. Effectively this data is analysing the situations players found themselves in during a match and determined the average expected result, based on probability. .

This process was repeated for every chance over the season and the probability of a clean sheet for the time each player was on the pitch was deduced. FPL Review modifies xG data slightly to reflect the impact ‘Quality’ appears to have [and it’s not as much as you’d expect] .

From the performance data expected goals, assists and clean sheets per player for each GW could easily be deduced and a bespoke algorithm generated xBonus and calculated the final points expectation. Any managers season could then be reviewed and a comparison made between the teams total Expected Points and the actual point tally. .

When ‘Expected Points’ are lower than the actual points this is considered Over-Performance (lucky). When ‘Expected Points’ are higher than the actual points this is considered Under-Performance (unlucky).

The Results

The first plot below illustrates the points total of each team in all 4 groups and their corresponding luck- according to the FPL Review algorithm.

Total Luck Analysis

From the data it is quite clear that luck doesn’t really even out over the season for each team. Over 70% teams in the random selection ended up within 50pts of their expectation, but extreme cases were found with over 100pts in difference both positive and negative. The mean and median luck of the 50 team sample were both within a few points of the expectation- which indicated that the analysis was reasonably balanced.

Nothing too notable was found within the active league selection. It fit in with the random 50 team dataset and most teams experienced either mild good or bad luck.

Elite Luck Analysis

Among the top 10 teams in the world the data gets particularly interesting, as shown above. Every single team experienced a very significant amount of fortune- the average amount being about 130pts. Of the other groups analysed, none contained a team that even reached 130pts . No team in the top 10 had less than 50pts of good fortune. Some teams received extraordinary luck over the season- one of which even was just over 250pts more than would typically be expected. However it should be noted that generally all 10 teams had very high point expectations- so it was not just dumb luck that got them there.

These results from Group 3 are very notable and clearly illustrate the recipe for a successful FPL season: Very High Performance + Very High Luck.

Group 4 is just one manager- a legendary manager noted amongst several online communities. This manager has had notably consistent year-in year out strong performances and his inclusion is to illustrate what a good manager with non-notable luck appears as. The legendary managers ‘Expected Points’ for the season is extremely impressive and better than a significant amount of the top 10 in the world, however he still managed to be 130pts away from a top 10 position by GW36.

Elite Team Table

With over 6 million players last season it is inevitable that of the thousands within the 99th percentile of skill will exist a sub-group who are can be considered among the 95th percentile of luck. Quick maths would suggest that 3000 managers out of the 6 million would be in both the extreme skill and high luck ranges.

Conclusions

So for the everyday manager what does this mean and what is the benefit of this data?

If you want to succeed in FPL it is worth knowing where you stand in terms of luck- relative to your competitors. It is very possible and reasonably common for two managers to make equally skilled decisions and yet have a gap of 100pts or more by the end of the season. It is possible that rather than making bad choices your strategy is good and your rivals have experienced very good fortune. Conversely this data can indicate if a change of tact is required- maybe taking all those hits is a bad idea after all. The concept is that this analysis can assist a manager in being less reactionary to the result and more aligned to actual performance and FPL potential.

If you wish to investigate this further feel free to try out the Review Performance tool on the FPL Review site. Any team in the world can be investigated using this tool.

Thanks for reading!