In August I made a post on my analysis of xG data and luck in FPL in the 2017/2018 season. That post was well received however there were two points raised
- Just one legend was included in the analysis and a greater sample was required to be conclusive.
- FPL Review also forecasts points based on bookmaker data however that was not considered as the algorithm was not running last season. This meant the usefulness of forecast from the site was unverified.
For this mid-season analysis I have included 15 legendary managers- 6 of whom are past winners with consistently high performance. All the data was collected after the completion of GW18.
I have also included the forecast functionality of FPL Review and considered luck from a dual perspective based on the following:
- Pre-match Bookmaker Odds & the FPL Review Algorithm: This aspect of the analysis determines the expectations for each player in each game based on odds. If a player with high projections based on the odds gets low returns it can be considered a good choice but unlucky result. Check out the Team Planner Tool to make use of these player performance forecasts now.
- Post-match xG Analysis: This data is based on the situations a player found himself in during a match, for example if a usually effective player misses 5 sitters it can be considered unlucky. The xG data is collected from understat.com. This data broken down by player for the season can be found on the Player Database page.
Both perspectives can be used to determine luck/over-performance in isolation though a blended version gives a more overall perspective so was used for this analysis.
The groups selected for this analysis were as follows:
- Group 1: A random selection of 25 managers [IDs 100000 to 100024]
- Group 2: A league of 11 active but non-notable managers
- Group 3: The top 15 managers overall worldwide by the end of GW18
- Group 4: 15 pre-selected legendary managers
How These Groups Were Analysed
Each GW for every team was investigated using both the pre-match odds forecast and post-match xG analysis methods. Deviation from the expectation of these expected values was considered luck.
The Season Review Tool was used to carry out this analysis and it can be used to review the performance of any team in FPL and determine their fair expectation for the season.
The immediate stand-out point is that the least lucky of the top 15 sides in FPL (+95pts of over-performance) was luckier than any team in the other 3 groups. There is little doubt that the worlds top 15 have all experienced extreme fortune, the most over-performing was the top placed team in the world at over 200 points beyond fair expectation.
It should be considered however that the worlds Top 15 have also made extremely high point expectation decisions (even better than most Legends) which combined with luck leaves them where they are. Some of the legendary managers have made decisions that leave them worthy of being in the mix at the very top, the highest point expectation Legend is a famous Finnish manager who was the only Legend used in the analysis from last year and is followed by thousands of users.
The data very clearly indicates that the Top 15 have experienced incredible amounts of luck and also shows a trend of more luck being required the higher you move up the ranks.
While the Top 15 have experienced a 1.75% over-performance that largely comes from the first 5 GWs, when several popular player choices among veterans produced exceptional performances. From GWs 1-5 the Legends experienced a 6.25% boost from over-performance however since then collectively they have under-performed by 0.07%.
The results conform to last seasons analysis, to reach the upper echelons both extreme skill and extreme luck is required. Targeting beyond a top 10,000 finish is not particularly realistic without accepting that some fortune will be required. The Season Review Tool will allow you to visualise this for yourself.
One interesting trend is that with luck/over-performance seemingly being capable of being measured then it appears that skill or decision quality is also measurable. It also appears that a teams points projections based on pre-match bookmaker odds and the FPL Review algorithm provide a very strong indicator of team performance.
The plot above shows just how well bookmaker odds and FPL Review algorithm forecasts FPL performance for the Randomised 25 team group. There was a correlation factor 0.862 between the Forecast and Total points for this group and 0.938 between the Forecast and xG points.
You can implement these forecasts into your own team using the Team Planner Tool. Please note it requires realistic and non-biased user inputs for expected minutes for full effectiveness.