Fantasy Football · Premier League

Fantasy Premier League: Underlying Stats, 2016/17 Gameweeks 1-3

During the summer I wrote an article on the underlying stats of Premier League footballers over the last two seasons and their correlation with the points scored in the Fantasy Premier League (FPL) game. There are many ways to approach FPL, but one method I wanted to utilize more this season was the use of underlying statistics as a guide for picking players and game strategy. With the season now three weeks old, this is a review post to study how the numbers are looking and whether there is any guidance that they can provide.

A few of points before I delve into the data: first, the number displayed for each player is ‘expected adjusted points per minute’, with only players who have completed 96 minutes in the first three games included (96 minutes is the typical length of time attributed to one full game in the Opta stats I am using, sourced via the Fantasy Football Scout members area). ‘Adjusted points’ excludes points that would have been scored simply for being on the pitch (one point until 60 minutes per game, 2 points thereafter to 90+ minutes). The formulas and metrics governing these projections are reiterated in this post but were initially discovered in the summer and are covered in the original blog post, so please refer to that post for a fuller explanation behind my thinking.

Second, it is a reasonable question to ask why I have not displayed these figures against what a player has scored in FPL points. The reason is because I am less interested in what a player has scored but more in what his actions indicated he should have scored. It is likely that in such a small timeframe a player may have scored two goals from just two shots on target, thus making him a hot FPL prospect. However, I am more interested in the fact that he has had only two shots on target and few touches in the final third which means that, expanding his actions over the whole season, he will not score many points. This means that his current hot streak is likely to be a lucky blip and he will start reverting to the mean (e.g. he will stop scoring points soon). I am only interested in finding players that performing well in the underlying metrics because they have the greatest potential for long-term rewards.

Finally, I fully recognise that this is a methodologically unsound experiment. The data is attempting to project forward the performance of players based on such a small sample size of three gameweeks; it is the equivalent of trying to predict the lottery numbers based on the first ball that is drawn. These tables are meant as a guide, an alternative method of looking at players. However, it is very possible (almost certain) that at least some of the players’ will not perform in the future relative to these performance tables, and this should be kept in the front of mind when reading this post and looking to use this research to make decisions. Take it with a very, very large pinch of salt.

(update: one final point, this post is foolishly being written and published very close to transfer deadline day, which may affect the viability and security of starts of some players mentioned here. All information is accurate at the time of writing. I think)

 

Goalkeepers

Season Formula: Adjusted Points = (Clean Sheets*4.79155)-(Touches*0.0266)+(Saves*0.51851)-2.901

Please note that the intercept (-2.901) has been adjusted downwards to account for the shortened timeframe

 

Def1

Currently, Manchester United’s David De Gea (£5.5m, 46.7% ownership) is the only elite keeper at the top of the rankings, in second, flanked by West Brom’s ever reliable Ben Foster (£4.6m, 13.9%) and 2014/15 favourite Tom Heaton of Burnley (£4.5m, 7.1%). The budget Eldin Jakupovic (£4.1m, 20%) of Hull also features prominently in the list, as the budget keepers here have all kept at least one clean sheet and are racking up the volume of saves they are being asked to make. Foster is currently out ahead by virtue having two clean sheets.

The caveat with the goalkeeper data is that team performances (in the form of clean sheets) play a significant role in the formula, so we will not always expect to see Cech and Courtois so low in the rankings, nor no Manchester City representative in the top 15. Eventually, these teams will start to collect clean sheets; this is more probable than, for example, a guarantee attributed to an expensive forward’s ability to score goals so it is worth considering for FPL managers.

For many FPL managers, the goalkeeping positions are the first to come under scrutiny when trying to free up budget, and these rankings show that there are budget keepers who have started strongly. This becomes more apparent when we neutralise the impact of clean sheets in the data by assuming that all goalkeepers have kept the same number of shut-outs (in this case, no clean sheets).

GK2

 

The table above looks at the FPL points-generating potential of all the goalkeepers assuming all teams kept the same number of clean sheets. This table shows De Gea and, interestingly, Middlesbrough’s Brad Guzan (£4.5m, 0.9%) falling significantly, which highlights the fine work their defences are doing in protecting them. The top of the table retains mostly the same names, suggesting that players like Heaton, Jakupovic and Adrian have FPL potential even without clean sheets. Ben Foster also belongs in this category as he sits 4th in the table, and is testament to what can happen when these already very good goalkeepers finally start keeping clean sheets as he sits top of the rankings in the first table.

Goalkeepers: what to watch out for

Player: Adrian, West Ham (£5.0m, 3.5% ownership)

The fall of De Gea in the rankings when team stats are neutralised shows the impact that clean sheets can have on a goalkeeper’s FPL performance. With this in mind, it is still too early to write off the elite keepers because their team performances will accumulate over the course of a season. The data also shows that under favourable circumstances (e.g. a good and/or lucky rotation policy) the budget keepers can combine excellent personal performance with the occasional team clean sheet to good effect, as has been the case early on with Heaton, Jakupovic, and especially Foster.

A player who seems to sit between both these categories is Adrian of West Ham. The Hammers defence, whilst not very solid have potential for clean sheets as they displayed against Bournemouth, but even without them Adrian will continue to perform in goal. For the price range, he may represent an option for anyone looking to save some money when transferring out a £5.5m elite keeper.

 

Defenders

Season Formula: Adjusted Points = (Clean Sheets*5.029)+(Attempts from Set Plays*0.176)+(Touches Final Third*0.015)+(Shots*0.447)-(Tackles*0.104)-(CBI*0.032)-3.86

Please note that the intercept (-3.86) has been adjusted downwards to account for the shortened timeframe

Defender1
As with the goalkeepers, team performance plays a significant role in the defenders’ ability to shine in FPL. It is no surprise then to see four of both Manchester United’s and West Brom’s defensive line in the top ten as both have kept two clean sheets to date. At the top of the pile is Middlesbrough’s Daniel Ayala (£5.0m, 0.2% ownership) and Everton’s Ashley Williams (£5.0m, 7.5%), both of whom are up their by virtue of a clean sheet and a willingness to take shots. However, with both of these players it is noted that they have completed just one game for their clubs, so the sample is smaller for these two than for others.

As with the goalkeepers, we would expect the course of the season to revert to type with elite clubs’ players generating more clean sheets and rising up the rankings. However, it is interesting to see which defenders have the best potential on an individual level, so as we did with the goalkeepers, we have removed the impact of clean sheets for the following table by assigning all players no clean sheets.

 

Defender2

When we performed this manipulation with the goalkeepers, the impact on the rankings was limited. However with the defenders, the changes at the top of the rankings are pronounced. The absence of clean sheets mean that there is no longer any Manchester United or West Brom players in the top ten, the previous leader Daniel Ayala falls to 10th place (although this is still high, suggesting he is going to be a strong FPL asset this season), and previous 2nd place Ashley Williams falls out of the top 20.

Crystal Palace’s Joel Ward (£5.0m, 1.1% ownership) is the new highest ranked defender by virtue of his five shots on goal, a decent volume of touches in the final third (58) and limited defensive activity which in this formula costs the player. Watford’s Jose Holebas (£4.5m, 0.2%) has similar stats and remarkably sits 6th in the ranking for touches in the final third. Investment in Watford’s defence is limited due to their starting fixtures which have been (and continue to be) tough, but at 0.2% ownership Holebas looks to be a player who is flying under the radar.

The following five on the amended list reads like an FPL hall of fame with stars from previous seasons. The name that jumps out here is Nathaniel Clyne (£5.5m). Owned by 11% of managers but without a clean sheet to date and a blank in his last two outings, the international break may see many managers looking towards some of the higher scoring £5.5m players, which will lead to Clyne’s value falling. His individual performances so far suggest that attacking returns will be forthcoming (six shots, 82 touches in the final third), and if Liverpool can start collecting clean sheets (insert own joke here) he could become a valuable asset.

The final word is on a player I know I will be asked questions on, so I’ll nip it in the bud here. John Stones (£5.0m, 33.5%) is causing a tremendous degree of frustration from what I can see on Twitter, although the assumption is that as soon as Manchester City start keeping clean sheets he will become a valuable asset. It is true that Manchester City should be keeping clean sheets, but on an individual level when the clean sheets are removed and the performance of the player is all that is taken into consideration, Stones sits 59th out of 82 defenders. However, it should be noted that this is only two places below Luke Shaw (£5.6m, 25%), a current FPL darling due to his clean sheets.

Defenders: what to watch out for

Player: Jose Holebas, Watford (£4.5m, 0.2% ownership)

The examples of John Stones and Luke Shaw highlight that, as with the goalkeepers, hunting clean sheets have to be the priority for FPL managers, so it is not advised to look beyond the elite clubs for at least two assets. Even if a defender is underperforming in all the other important metrics in the above formula, he can be comprehensively redeemed if he is part of a unit that is keeping clean sheets. For the budget options where clean sheets are not guaranteed, the second table in this section which focuses on the individual performance can provide a useful barometer on where non-clean sheet points may come from. With this in mind, some key differentials to look out for are shaping up to be Joel Ward, Daniel Ayala, Jonas Olsson, Kyle Naughton, Adam Smith and, one I will be keeping a close eye on considering his performances against tough opposition, Watford’s Jose Holebas.

 

Midfielders

Season Formula: Adjusted Points = (Passes Received Final Third*0.03363)+(Shots on Target*1.9876)+(Touches Final Third*0.01897)-5.85

Please note that the intercept (-5.85) has been adjusted downwards to account for the shortened timeframe

Midfielder1

The most notable thing about the list of leading midfielders is the absences from the top 20: De Bruyne (#22), Eriksen (#25), Mane (#34), Martial (#35), Firmino (#36) and Alli (#57) are all names we’d expect to see near the top of the list of most effective midfielders. However, it is not the case that they have been usurped by a group of unknown budget players. The sub-£6.5m players in the top 20 – Mirallas, Deulofeu, Redmond and Townsend – were all names heavily discussed amongst the FPL community on Twitter before the season started, although the emergence of Kevin Mirallas (£4.4m xx% ownership) so high in the list is perhaps a little unexpected.

The 50%+ ownership of forwards Sergio Aguero and Zlatan Ibrahimovic has put a strain on many squads’ midfield positions (although that may change with Aguero’s forthcoming ban), with a typical maximum of two elite midfielders able to be accommodated. The lack of form from the aforementioned big names has been helpful by narrowing our options, however there are still plenty of big names near the top of this list to make FPL managers second-guess their choices. Despite this, the two stand-out candidates for inclusion based on this data appear to be Liverpool’s Philippe Coutinho (£8.2m, 30.6%) and Edin Hazard of Chelsea (£10.2m, 28.1%).

Whilst the clubs that are featured at the top of this list may not be a surprise, and a great many names are also those familiar with previously successful FPL seasons, there are still differentials to be found. Glyfi Sigurdsson (£7.4m, 5.7%) and Willian (£7.4m, 6.9%) have already experienced price drops despite having the 3rd and 4th best underlying stats amongst the midfielders, and Juan Mata (£7.5m, 2.6%) is assumed to be on his way from Old Trafford despite having the best figures of any of the Manchester United midfielders. Slightly further down the list, Liverpool’s James Milner (£6.5m) has strong stats despite being fielded at left back for some of the season to date; he is currently owned by just 0.9% of FPL managers.

Midfielders: what to watch out for

Player: Josh King, Bournemouth (£5.5m, 0.8% ownership)

There are few budget options present here, however if we look just slightly outside the top 20 we find Bournemouth’s Josh King (#23). At £5.5m and owned by just 0.8% of players, he probably remains firmly under the radar despite a goal in gameweek 3. Bournemouth’s fixtures are looking good despite a trip to Manchester City in gameweek 5, so King could prove to be an effective differential.

 

Forwards

Season Formula: Adjusted Points = (Shots on Target*1.69668)+(Touches Final Third*0.03415)-5.67

Please note that the intercept (-5.67) has been adjusted downwards to account for the shortened timeframe

Forward1

Written off by many before the start of the season, Wayne Rooney (£9.0m) has just 6% ownership but is leading the forward rankings, very closely followed by Diego Costa (£9.6m, 11.3%) who was also projected to have limited impact this season, although this was because he was expected to be transferred out of Stamford Bridge. Both Sergio Aguero (£13.1m) and Zlatan Ibrahimovic (£11.7m) are owned by more than 50% of FPL managers, which is justified by their presence near the top of this list; Ibrahimovic especially has his doubters due to his abnormally high conversion rate which is expected to fall but he remains high in the list in part due to having more touches in the final third than any other player. As an Ibrahimovic owner myself, these stats have convinced me that there is no need to move him out just yet.

The most eye-catching name on this top 20 list (which is frankly not short of a few surprises) is Shinji Okazaki of Leicester in 4th place. At £5.9m and owned by just 1.7%, he is outperforming last season’s main man Jamie Vardy (curiously absent from this list with his England partner, Harry Kane) and the player who was bought in to compete for his place, Ahmed Musa (£7.4m, 7.9%). Despite this, doubts persist over his potential for time on the pitch – he is yet to complete 90 minutes this season – and with Leicester pursing Islam Slimani from Sporting Lisbon, it may be that Okazaki is not a sustainable differential.

Beyond the elite strikers, most FPL managers need a cheaper 3rd striker, but even this list of top performing strikers not short of budget options does not produce any stand-out candidates: Burnley’s Andre Gray is facing a potential suspension from the FA; Enner Valencia is about to play second-fiddle to the incoming Simone Zaza at West Ham; Shane Long and Charlie Austin seem to be rotating starts at Southampton; Cristhian Stuani is not guaranteed minutes at Middlesbrough;  Sunderland’s Fabio Borini and Burnley’s Sam Vokes are not blessed with illustrious striking history in the Premier League; Fernando Llorente of Swansea is cursed with an awful set of forthcoming fixtures.

The two best budget candidates then appear to be West Brom’s Solomon Rondon (£6.5m, 3.9%) and Hull’s Abel Hernandez (£6.0m, 2.1%). Rondon’s prospects will be boosted by the arrival of Nacher Chadli from Tottenham which should give the Baggies more attacking threat, and Hernandez is the undisputed starter for the season’s surprise package to date, although their ability to maintain this over the course of the season is doubtful.

Forwards: what to watch for

Player: Solomon Rondon, West Brom (£6.5m, 3.9% ownership)

In my view, Rondon is the pick of the budget strikers, although as previously stated there is significant flaws with each of the options; anyone who witnessed West Brom’s toothless attacking display at home to Middlesbrough will testify to Rondon’s. Another option considering the paucity of options in attack is to sacrifice the 3rd striker slot for an ultra-cheap option (Adama Diomende and Modou Barrow are potentials), transfer resources to upgrading the 5th midfielder where there appear to be many more decent options, and revert to a 3-5-2 formation for the season.

 

Two up front?

Thinking about the prospect of sacrificing the 3rd forward in the team for a midfielder is a bold move, but one worth exploring. Below is a chart of the proportion of players who have played in each position to acquire a target Expected Adjusted Points per Minute.

The threshold here is 0.025 Expected Adjusted Points per game, which is equivalent to midfielders such as James McClean, Bojan and Adam Lallana, and forwards such as Solomon Rondon and Enner Valencia. However, from this moment on I will not be referring to the players by name and treating them as data points only. I am looking at the statistical probability of picking a good midfielder vs. striker, and so I’ll make the assumption for the sake of this article that the patterns set here are established and will be repeated throughout the season, though with a revolving cast of main players. In short, look upon this as a theoretical exercise designed to explore team structure and not as a guide for identifying individual players like that which has come before.

Chart1

The data shows that 39% of forwards have 0.025 or higher, whereas only 26% of the midfielders have achieved the same. This indicates that you have a proportionally greater pool of high-potential players to pick from in the forwards, which in turn means that you have less chance of making the wrong call. It should also be noted than just ten of the 28 midfielders in this range (36%) are priced 7.0 and under, whereas for the forwards it is five from 12 (42%), so there is proportionally more ‘bargain’ players in the forward lines who can be risked.

Of course the data also shows that there are no forwards with more than 0.045, whereas there are seven midfielders who meet the criteria, and from a multitude of price ranges too. Therefore, whilst you have a statistically better chance of picking a good forward than a good midfielder, if you’re going to strike it big you’ll need a midfielder to do so.

Advertisements

7 thoughts on “Fantasy Premier League: Underlying Stats, 2016/17 Gameweeks 1-3

  1. Excellent follow up on former underlying stats metric. How about an article on which underlying stats explains clean sheets? With this being so important for GK + DEF?

    Like

    1. Thanks very much. Yeah, I’ve been thinking about explaining clean sheets as you’re not the first person to ask me. Unfortunately, the data source I use (Opta, via Fantasy Football Scout) does not feature many ‘conceded’ stats other than variations of goal attempts. I could take each entry for each team’s proactive stats and match them against the opponent to get what I need, but the double gameweeks (DGW) are bundled together. So, to quote a hypothetical example, I know that Everton played 1000 passes in DGW 34, but I can’t say how many ‘passes conceded’ were attributed to Opponent A and how many were for Opponent B. If you, or anyone else reading, has any idea where I can get Opta stats broken down by game rather than gameweek, then please let me know so I can crack on with improving my formulas!

      Like

      1. you’re defenitely more into the regression analysis than I am, but isn’t there any value in just taking total CS for each team and try to explain by total big chances conceeded etc. and see which underlying stats are significant?

        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