• Bruin Sports Analytics

Cracking the Fantasy Premier League Process

By: Shail Mirpuri


Source: Fantasy Premier League

Every year over 6 million soccer fanatics take part in the biggest fantasy soccer league in the world: The Fantasy Premier League (FPL). Players face tough dilemmas when selecting their squads by considering where they should allocate their budgets, when they should make transfers, and ultimately whom they should select to lead them to glory. We can make these decisions with greater certainty if we take a deep dive into the data from previous seasons to uncover some fascinating trends. By looking at various performance influencing factors such as form, fixtures, price and position, this article aims to uncover some useful tips and tricks to allow you to gain an edge over your rivals, and rise up the fantasy premier league leaderboard. Not only will we explore the valuable insights derived from analyzing past seasonal data, but we will also demonstrate how statistical methods can actually be applied to test various different ‘theories’ within the fantasy soccer realm.


The Dataset


In order to find out which factors are important when building the so-called ‘optimal’ team in FPL, we need to consider the historical player performance data. To do this, I will consider two main data sources. The first of these consists of seasonal FPL data for each premier league player since the 2005/06 season. With this large sample, we will be able to get a grasp of the longer term trends occurring within FPL. The second dataset consists of gameweek level data from the latest 2019/2020 season. Here, we can consider the short-term trends within the season, and the impacts of this on micromanagement decisions such as transfers made and captaincy choices. Both these datasets consist of data for all the English Premier League (EPL) soccer players registered in the FPL database for a given season regardless of whether or not they played a match. Since there are a lot of reserve and youth players that have had a lack of playing time (and therefore a lack of points), I have filtered the data to only include players who have played a certain amount of time. For the seasonal FPL dataset, players are only included if they have played more than a third of the total season, while for the gameweek dataset they are only included if they have played more than 60 mins in the particular game. These changes were made to ensure that our analysis isn’t skewed by players that are not playing games. We want to know more about the performance of those who have been playing consistently; these modifications help to make this possible.


Where Should You Invest Your Budget?


One of the biggest challenges players face in fantasy football is deciding where exactly to spend their allocated 100 million budget. Any decision made comes at an opportunity cost, so it is important to consider which positions you should look for. In theory, a player that costs more should perform better since the price tends to be determined by the quality of the player. However, we need to consider how much better they actually perform to decide whether it is worth investing in them rather than looking for a cheaper alternative. For instance, if I signed a Goalkeeper for 6.5 million rather than one for 4.5 million, could I be spending that additional 2 million on another position and be getting more points? Let’s find out.


In order to derive these insights, we need to group players into 3 categories: Budget, Mid-Range and Premium. To account for the different price ranges in each position, I decided to use positional price percentiles. For each position (Goalkeeper, Defender, Midfielder, Forward), we will compute the 50th and 85th percentiles for the price. Every player with a price above the 85th percentile in their position will be assigned the Premium category. Those with a price between the 85th and 50th percentile will be put into the Mid-Range category. Finally the remaining players will be put into the Budget category.


Pricing Differences Between Each Position In FPL

From the table above we can see that Forwards tend to be the most expensive position, while Goalkeepers and Defenders tend to be cheapest. We see this in the actual game, where forwards are more likely to get you points due to the larger goal threat they possess. But is this additional price worth spending? Let’s now compare the seasonal performance of each position when broken down by price categories.



The swarmplot above demonstrates some notable differences between each position and price range brackets. First, we can see that premium players in general tend to outperform mid-range and budget players. However, it is important to point out that there are plenty of exceptions to this rule. This means that at times we can potentially save some of our budget and accumulate the same amount of points. Another fascinating insight is that midfielders tend to significantly outperform other positions. From the midfielder’s portion of the graph we can see that the top performers throughout the seasons tend to be from the premium bracket in this position. This indicates that spending extra money on higher quality FPL midfielders can significantly boost a team’s overall performance. In fact, getting a solid group of high quality midfielders into your team can completely redefine your season and propel you up the leaderboard. We will now look at the top 10 seasonal performances over the last 15 years to see if we can identify any positional or price bracket trends that back up the graph above.



From the bar plot above, we see further evidence backing up our initial interpretation. First, out of the top 10 seasonal performances, 9 of these are from players in the premium category. One notable exception is the 4th best seasonal performance of Riyad Mahrez as a part of Leicester’s ‘freak’ title-winning season in 2015/16. Apart from this one-off unpredictable year, it seems that premium players are going to be getting a large chunk of your points season on season. Another interesting observation is that midfielders seem to be the ‘go-to’ position when it comes to exceptional performances year on year. In fact, we can see that the top 9 performances have all come from midfielders, mostly in the premium category. If we look more closely, we can actually observe that out of these 9 midfielders, almost all (with the exception of Kevin De Bryune (#3) and Dele Alli (#9)) are considered to be wide forwards. This may indicate the rewards that can be enjoyed by investing in premium wide-forwards this season. One example of a player to definitely consider is Mohammed Salah. On our top 10 list, he appears not once, not twice, but an unmatched three times with all performances coming in the last 3 seasons. His consistency and threat once again makes him a reliable choice for your team.


Finally, let’s confirm our findings that midfielders are the area to invest your FPL money by conducting a two-sample t-test when comparing the mid-range group of players with the premium group. This will allow us to see if there is a significant difference between the points earned by players in different price-ranges for each position. The null hypothesis for all these tests is that there is no significant difference between the mid-range players and premium players for a particular position. We will conduct this hypothesis test using a significance level of 0.05. After computing the t-statistic for each test, we will compute the p-values to see if there is evidence against our null hypothesis.


Does upgrading from a mid-range player to a premium player have a significant impact on performance?

The table above shows the key metrics computed during these hypotheses tests. We can clearly see that opting to select a premium goalkeeper instead of a cheaper mid-range one doesn’t produce significantly more total points over a season. On the other hand, we can see that the most significant difference between the Premium group and Mid Range group occurs in the Midfielder position. This is indicated by the test having the lowest p-value and highest t-statistic. The low p-value means that based on what we have observed over the last 15 years it is highly unexpected if it were the case that premium and mid-range midfielders perform equally in FPL. Therefore, we would use this as evidence to suggest that premium midfielders perform significantly better than mid-range midfielders.


We have drawn some riveting conclusions about the breakdown of FPL performance by position and price to obtain some actionable insights. Firstly, our findings indicate that you should look to invest a bulk of your 100 million budget in a couple of premium midfielders over any other position. We have also seen that settling for lower-priced goalkeepers may be a better option since the performance difference between higher- and lower-priced players in this position is not significant. Your go-to picks for this season can be established elite midfielders, especially those who play as wide forwards such as Mohammed Salah, Sadio Mane, Heung Min Son and Raheem Sterling.


Form vs. Fixtures


Next, we will consider short-term trends such as form and fixtures that can impact a user’s week on week decision making. These micro-management decisions include figuring out whom to captain on a gameweek or which player to bring into your team for the next few weeks. Form can be measured by looking at the moving averages of various key metrics that are often indicative of performance in FPL including total points earned, ICT index, and Influence, Threat, and Creativity scores. This will allow us to pin-point the most reliable metrics for predicting a player’s future performance. We will consider moving averages of 1, 5 and 10 gameweeks to investigate whether short- or long-term form plays a greater role in predicting a player’s performance for a given gameweek. Apart from considering how a player's form impacts his performance, we will also explore whether favorable fixtures make players perform significantly better. The difficulty of each fixture will be accounted for by converting the opposition team categorical variable into the position the opposition team finished during the 19/20 season. For instance, those playing against Liverpool will have a 1 in the opposition column since Liverpool finished top of the league last season. Under this definition, the greater the value in the opposition team column, the easier the fixture will be since the quality of the team played against is lower.



From the figure above, we can observe some notable trends. First, it can be seen that monitoring a player's performance over a longer period of time is a better way of predicting current performance. In other words, this insight suggests that you should not jump onto a player who performs amazingly in a single gameweek, but rather you should look for longer term consistency when deciding whether or not to bring a player in for a given gameweek. Furthermore, we can see that looking at metrics such as the threat score and ICT index may be better indicators in judging a player’s form over other metrics such as creativity, influence and total points. Additionally, we can observe that the impact of form and fixtures are fairly similar in terms of predicting a player’s score for a specific gameweek. This suggests that we should take a holistic approach when making transfers and captaincy decisions for a given gameweek: specifically, we can consider the fixture difficulty, and the 10 week moving-average for threat score and ICT index. It is also important to point out that all of these factors have a relatively low correlation with total points. This means that they cannot solely tell us what exactly is going to happen in a particular gameweek, but rather they may be used to reduce uncertainty slightly. At the end of the day if the Premier League was so predictable with simple metrics like these, then it would not be the most entertaining soccer league in the world. Despite this, these key insights can allow you to make more informed micro-decisions throughout the FPL season.


Home Advantage


Another interesting aspect to consider especially when deciding who to captain for a particular gameweek is whether or not the player is at home. A captain in FPL is a player that you pre-select to earn double the amount of points for that given gameweek. Having a good captain choice can elevate your total weekly score by around 10-20 points, which can prove to be pivotal to success especially in the long run. Usually, amongst your premium players there are 2-3 options that stand out as viable captain choices. Often people cite that a player is likely to perform better when they are playing at home, and therefore you should always captain these types of players. Investigating whether or not this is true can be easily done using a two sample t-test. Here, we will assume that our null hypothesis is that players who play at home, on average, score the same number of points as those that play away from home. Again, we will use a significance level of 0.05 for this test.


Summary of Home and Away FPL Performances

From the table above we can see that, on average, players tend to perform better when playing at home than when playing away from home. This is reflected in the slightly higher mean score when comparing home and away performances. However, to test whether or not this difference is actually significant, we will need to compute the t-statistic and the corresponding p-value. Using the standard deviation, number of fixtures, and mean points scored, we can find that the corresponding t-statistic is 6.70. Therefore, the p-value for our test will be approximately 1 * 10^-11, which is much smaller than our significance level. This shows us that there is a statistically significant difference between a player’s performance at home and away from home, suggesting that home advantage is a factor impacting weekly player performance. Using this conclusion along with the insight that fixture difficulty impacts performance, one potential strategy is to transfer in a player with favorable home fixtures in the short-term. Apart from this, users should also aim to captain players that have a home fixture over those with an away fixture. Ultimately this can help narrow down captaincy choices for a particular gameweek.


Key Takeaways


Overall, we have seen the power of statistics in making better FPL decisions by considering where you should spend your budget, whom you should captain, and what you should consider when making transfers. We have shown how premium midfielders are vital for a team to be successful over the course of a season. These insights argue that settling for mid-range players elsewhere, and spending heavily in the midfield may be the best bet in cracking your squad-selection dilemmas. In addition, considering the longer-term threat scores and upcoming fixtures of a given player can definitely reduce the uncertainty when it comes to making short-term decisions within the game. Finally, we have shown that home-advantage is a huge factor in a player’s FPL performance, so we should prefer players with favorable home fixtures for the captaincy selection.. Paying attention to these tiny details can add up over the course of a long season to help you conquer your friendly leagues, and maybe even the global one!


Sources: Fantasy Premier League, Seasonal FPL Dataset, 19/20 Gameweek FPL Dataset


GitHub Repository: https://github.com/shailm99/crackingFPL

191 views

Recent Posts

See All