Fantasy Football

Fantasy Premier League: Player Rankings, Gameweek 8 2016/17

This is the first of what I hope will become a regular series of articles which use a methodology I have developed to identify the players with the most potential for points in the Fantasy Premier League (FPL) game. For those new to this methodology, here is a brief overview of the research I have been doing and how I have come to develop this ranking system:

  1. I correlated FPL points from 2014 to 2016 with key underlying metrics for each position, which through a simple linear regression analysis gave me equations which explain what should be happening each game from an FPL perspective. The key word here being ‘should’; over the course of the season actions will generally revert, but on an individual game level there is no telling what will actually happen. This method is a way of ranking players by their probability of performing through assessment of their on-pitch actions, or what I’m calling form, over the last four games.
  2. I then rank each club based on the forthcoming six fixtures. This is done by understanding the propensity of the opponents to concede big chances and the propensity of the midfielder or forward’s team to create big chances. For goalkeepers and defenders, the reverse is taken into account.
  3. For the goalkeepers and defenders, clean sheets play a significant role in assessing their form, which is a team action covered by the fixtures ranking. Therefore, in order to get a true view of a player’s form regardless of their team’s activity, I neutralise the clean sheets metric by giving all players a value of zero for this metric.
  4. The overall rank over the player is discovered by simply calculating the average between the form (individual) rank and the fixture (team) rank.

Here’s an example to illustrate my methodology. Let’s take Christian Eriksen of Tottenham.

  • From gameweeks 4 to 7 his form suggested he should be scoring 0.0439 points for every minute he was on the pitch (not including appearances), which earns him a form rank of 9th. Whether he scored this amount is not of interest, what is important is that he should have done based on the equation for midfielders.
  • Tottenham in that time created 7 ‘big chances’ according to Opta, or 1.75 per game. In that same period, Tottenham’s next six opponents from gameweek 8 onwards conceded an average of 1.875 big chances per game. Therefore, we calculate Tottenham’s attacking potential over the next six as 1.813, or the average of them both, which earns him a fixture rank of 7th.
  • Taking the average of the form rank (9th) and fixture rank (7th) gives him an overall rank score of 8. Only 7 midfielders earned a better rank score than this, so his overall rank amongst midfielders is 8th.

Please note that only players who have played more than the equivalent of more than one full game in the last four qualify for the rankings.

It should also be noted for anyone looking to use this analysis as the gospel that some players will over-perform against the model and others will under-perform. The research is designed to increase probability, not provide a definitive answer.

 

Top 10 Goalkeeper Rankings: Gameweek 8 Onwards

gks

Note: 11 names are shown due to a tie for 10th place

What is notable about the goalkeepers highlighted is that the strong individual performers have bad fixtures coming up, and those with good fixtures are not strong on individual performances within the context of FPL. We are left with a true form vs. fixtures dilemma; is it better to back the Manchester United defence to keep a clean sheet which will be the majority of David De Gea’s points, or back the form of Jordon Pickford to collect save points and hope for a clean sheet that is unlikely to arrive?

 

Top 20 Defender Rankings: Gameweek 8 Onwards

defs

Note: 21 names are shown due to a tie for 20th place

Watford’s set-piece taking wing-back Jose Holebas is the name at the top of the rankings as his individual form is strong and his club have good potential for clean sheets in the next six games. Chelsea’s switch to a 3-4-3 against Hull has propelled Marcos Alonso into the wing-back position for his club and up to second in the rankings; whilst Chelsea’s potential for clean sheets are average, he is the top performing defender based on the individual metrics when he has been on the pitch.

 

Top 20 Midfielder Rankings: Gameweek 8 Onwards

mids

Note: 22 names are shown due to a tie for 20th place

Liverpool’s extremely favourable fixtures and excellent individual performances mean that a remarkable six of their players make it into the top 20 midfielders list, with Roberto Firmino at the summit. Tottenham similarly have three players in the top 8. A surprising name to be found near the top of the rankings is West Brom’s James McClean, but he has registered a shot on target every 49 minutes in the last four games, a rate bettered only by Firmino and Kevin De Bruyne at the summit of the list. Another unexpected name to appear is Wahbi Khazri after being frozen out under David Moyes at Sunderland, but strong underlying stats in his first start against West Brom and post-match words of praise from his manager may mean that he comes back into the FPL reckoning soon.

 

Top 15 Forward Rankings: Gameweek 8 Onwards

fors

Note: the source of this data lists Watford’s Issac Success as a forward, whereas in the FPL game he is a midfielder. I will update the tables in due course.

The list of top strikers is littered with players who are not sound investments due to their insecurity of minutes. If we ignore these players (plus the injured Harry Kane), then Success’ Watford teammates Troy Deeney and Odion Ighalo occupy the 3rd and 4th positions on the list by virtue of their fixture list, making them appealing FPL assets in the coming weeks. A notable name missing from the top of the rankings is Marcus Rashford despite appearing to have nailed down a starting berth at Manchester United.

The hot topic as I write this is whether to ditch Ibrahimovic from the squad. Removing the unsecure players from the list, the Swede sits at 5th in the rankings, weighed down by his fixtures against solid teams. The data suggests that it would be a hasty move to get rid of him, but from a personal perspective three things count against him:

  1. The mass hysteria around his continuous blanks at home to Leicester (GW6) and Stoke (GW7) mean that he is going to drop in price, probably by £0.2m or £0.3m to well below his original £11.5m value. His premium tag means he has a long way to fall, so time is not on a FPL manager’s side
  2. The aforementioned fixtures, including Liverpool, Chelsea and Arsenal in the next six, not to mention a surprisingly-resilient-against-the-­big-teams Burnley, mean that his opportunities are likely to be more limited in the coming weeks. There are forwards with better fixtures coming up.
  3. The £11.5m most managers have invested in him can be useful elsewhere in the squad, and the falling price means that bringing him back in if it looks like he’s hitting form will be a lot easier if his price dips to £11.0m than when it was up at £11.9m.

In my opinion, despite his underlying stats, now is a good time to sell Ibrahimovic.

 

If you have any questions, I can be found on Twitter @artemidorus_1.

Advertisements

4 thoughts on “Fantasy Premier League: Player Rankings, Gameweek 8 2016/17

  1. Good article, look forward to the updates. How do you deal with the quality issue of the previous Four fixtures.? For example a team who had their previous Four fixtures against Hull, Bournemouth, Stoke and Middlesborough are likely to have better team and individual statistics than if the last Four opponents were City, Utd, Tottenham and Liverpool

    Like

    1. Thanks! I haven’t yet incorporated the type of weighting you’re talking about into the model. I have thought about it but haven’t explored yet how I want to do it. The problem I encounter quickly is that when trying to assess the strength of an opposition, as you mentioned, one recognises that this is influenced by the strength of their opposition. For example, if Man City are creating five big chances per game we assume they are strong, but that might be because they’ve played weak opposition, but the numbers will say that the opposition are weak because they’ve played strong opposition like Man City, and I quickly get into a horrible feedback loop. What is needed is a method of identifying the baseline and using that as the weighting, but I haven’t yet devoted the mental energy to solving that (could be something as simple as bookmakers’ odds – might as well piggyback on their analysis).

      The other weakness in this model that I’m not yet happy with is the even weighting attributed to individual form and team fixtures. I instinctively feel that a 50-50 split is not correct, nor do I feel that its necessarily consistent across positions. Historical data similar to that which I used to develop the form formulas hold the key, but that was a big piece of work and I haven’t had the time to embark on another yet. I get the feeling that I’ve stumbled unknowingly into a quest to develop a perfect predictive algorithm which may yet take years (considering my lack of free time). In the meantime though, I feel it’s worth sharing my work-in-progress because I feel that it’s interesting at least.

      Like

  2. Nice piece but your info regarding ibra a bit off.

    It’s very easy for a player to rise above his base price as thresholds are lower, these are irrespective of percentage ownership

    but once a player drops to his starting value, then the further drops are a function of percentage ownership.

    This is evidenced by how fast Ibra dropped from 11.9 to 11.5 – he only lost his rises. same with hazard dropping only 0.1 below his base price.

    As of now Ibra is -43.4% towards a drop and requires over 110k transfers to drop to 11.4. That’s because now the no. of ntis required to drop are a function of ownership percentage, which is very very high with ibra given his ownership in the dead teams.

    There is no way he will drop more than 0.1 before deadline.

    Like

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