Community

Members Only – Premium Dilemmas: which £11m+ players to own?

In the limited budget game of FPL, sadly we can’t own all the premium players. Some will try – World Chess Champion Magnus Carlsen has at times this season owned four premiums and done very well (currently 223 OR) – but most advocate a balanced approach favouring two or three £11m+ players.

In this article I focus on the six players costing over £11.0m – Salah (£13.3), Kane (£12.6), Aubameyang (£11.4), Sterling (£11.3), Aguero (£11.2), and Hazard (£11.1).

I use xG Delta to examine their goalscoring ability. First a quick recap on xG:

The idea behind xG is to quantify the likelihood of a goal being scored from a particular chance. The higher the xG, the better the opportunity.

For example in the most recent round of fixtures, Romelu Lukaku’s goal against Newcastle had an xG of 0.61, meaning it should be scored 61% of the time. Whereas Sergio Aguero’s goal against Liverpool had an xG of only 0.05, meaning it should be scored 5% of the time.

Intuitively we know that Aguero’s was a much harder chance, that is reflected in the xG numbers.

What xG Delta does is measure how the reality compares to the probability. If a player has 12 goals from an xG of 9, then he has scored 3 more goals than we would expect from a player of average ability. He has an xG Delta of +3.

Over a short period of time this could simply be due to a run of good form, but players who consistently exceed their xG can be regarded as expert finishers.

Become a Member and access our data

Memberships for the 2018/19 campaign are now available for the price of just £10.

Join now to get the following:

  • Plot your transfer strategies using the fully interactive Season Ticker.
  • Get projections for every Premier League player provided by the Rate My Team statistical model.
  • Use Rate My Team throughout the season to guide your selections and transfers.
  • Get access to over 130+ exclusive members articles over the season.
  • Analyse our OPTA-powered statistic tables specifically tailored for Fantasy Football Managers.
  • Use our exclusive tool to build custom stats tables from over 100 OPTA player and team stats.
  • View heatmaps and expected goals data for every player.
  • Use our powerful comparison tool to analyse players head-to-head.

CLICK HERE TO BECOME A FANTASY FOOTBALL SCOUT MEMBER

 

 

 

TopMarx Fan of Fantasy Football and Monty Python. "Archimedes out to Socrates, Socrates back to Archimedes, Archimedes out to Heraclitus, he beats Hegel. Heraclitus a little flick, here he comes on the far post, Socrates is there, Socrates heads it in! Socrates has scored! The Greeks are going mad, the Greeks are going mad! Socrates scores, got a beautiful cross from Archimedes. The Germans are disputing it. Hegel is arguing that the reality is merely an a priori adjunct of non-naturalistic ethics, Kant via the categorical imperative is holding that ontologically it exists only in the imagination, and Marx is claiming it was offside. Follow them on Twitter

124 Comments Post a Comment
  1. markdmwilliams
    • Fantasy Football Scout Member
    • 7 Years
    5 years, 2 months ago

    Fantastic Article!

    1. Geoff
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      So so strong!

    2. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      Thank you 🙂

  2. Members Only - Premium Options
  3. Juhlers
    • 11 Years
    5 years, 2 months ago

    Great article indeed...

    So who's selling Kane? 😀

    1. Geoff
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      I'm considering it..

      1. rnrd
        • 9 Years
        5 years, 2 months ago

        Problem is that Spurs calendar is both very appealing AND quite congested.
        So you probably really need spurs cover and because there are a lot of games in a short period of time (which is not the case for Liv or City), and this makes spurs MIDS no real viable options imo

    2. DMil
      • Fantasy Football Scout Member
      • 9 Years
      5 years, 2 months ago

      I was considering getting him for Aubameyang when Son left but I’ve scrapped that plan now. Looking at a spurs midfielder for that nice run instead

  4. petergriff
    • Fantasy Football Scout Member
    • 14 Years
    5 years, 2 months ago

    When speaking of who you’re going for, please say who you have already.

    Not spoken of much is that Aubameyang’s away stats are much worse than his home stats.

    I have Haz, Kane and Aubam.

    I’m tempted to go Aubam > Salah as part of a -4 this week.

    Then do Haz > Sane/Sterling next week for City’s great run.

    1. A.T
      • 13 Years
      5 years, 2 months ago

      Blimey I really fancy Aub this week. I certainly wouldn’t be taking him out for a hit. I don’t how bad these away stats are but surely all players suffer from this, no?
      I just see it at as a great attacker against a poor defence and a history of the home team struggling against this opposition.

      1. petergriff
        • Fantasy Football Scout Member
        • 14 Years
        5 years, 2 months ago

        Hmm, yeah i guess many players are weaker away than they are at home.

        To answer your question, Aubam has:
        -xGI 5.75 in the last 4 home games (highest for all forwards);
        -xGI 2.79 in the last 4 away games (2nd highest for all forwards).

        Arsenal as a team are also much weaker away than at home too. ARS have had 24 more shots in the box at home than they did away over the last 8 GWs. That's the 2nd most home-bias in the league, behind Fulham.

        For me, i have Kane+Haz+Aubam. I really want Salah. So i have to choose to drop 1 of my 3 big hitters. I can't drop Kane in his form or Haz with his NEW fixture. So i either don't get Salah for another week (already burnt me since selling after his hat trick) or i sell Aubam. It's a tough call.

        1. petergriff
          • Fantasy Football Scout Member
          • 14 Years
          5 years, 2 months ago

          And just to compare Aubam's stats with Salah's:

          Salah:
          -xGI 3.88 in the last 4 home games (highest for all mids);
          -xGI 4.15 in the last 4 away games (highest for all mids).

          Interestingly, then, Salah in total has been less threatening than Aubameyang in total home and away (8.03 vs 8.54). But, Aubam has whm away next who are leaky away but not so much at home (8 BC conceded at home last 4, which is mid-table). And Aubam's next home game, where he normally excels, is CHE who are very tight defensively away at the moment (3 BC conceded, 2nd best in the league)

          1. scubasmithy
            • Fantasy Football Scout Member
            • 8 Years
            5 years, 2 months ago

            Worth considering also that

            - Prem Mids get an extra point for goals
            - Prem Mids can get a clean sheet point
            - If Hazard doesn't blank, he is also a bonus magnet (unlike Salah)

            1. Deulofail
              • 8 Years
              5 years, 2 months ago

              Extra point for goals, but don't forget, 6 fewer BPS points, which makes it much less than 1 point extra per goal on average.

        2. Saka Punch
          • 5 Years
          5 years, 2 months ago

          Also gator in the fixtures though, Auba was totally out of the Liverpool game so this would scew it. Aso bear in mind a number of players coming back from injury, including Ozil, who usually helps Auba's goal threat

          1. Saka Punch
            • 5 Years
            5 years, 2 months ago

            Factor*

    2. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      Well he played Liverpool recently away, and posted poor stats in that game, but good xG numbers against Southampton and Brighton.

      I think you have to look at the teams he was playing when talking about a stat, his home stats boosted by that Fulham match.

      Had he recently played Fulham away and Liverpool at home, his stats might paint a different picture.

      Anyway I think West Ham is a good match for him.

      1. petergriff
        • Fantasy Football Scout Member
        • 14 Years
        5 years, 2 months ago

        Yes, that's a fair point.

  5. Alchim1sT
    • Fantasy Football Scout Member
    • 7 Years
    5 years, 2 months ago

    Had Auba, Kane und Haz ... sold Auba to get Salah back! I don't trust Arsenal long-term while Spurs, Chelsea and Pool have a better schedule over the next weeks. As an extra point: There are plenty mid priced forwards i like ... Arnie, Mitrovic, Rashford and so on

  6. Geoff
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 11 Years
    5 years, 2 months ago

    Best line:
    'And, as a result, will his returns be, dare I say it, more haphazard?'

    1. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      I'm not sure what that says about the rest of the article

  7. ochili
    • Fantasy Football Scout Member
    • 5 Years
    5 years, 2 months ago

    Great article!

    There is one aspect of choosing which premiums to have in your team, that I feel is a bit overlooked.

    I've had the premium midfielders Salah, Hazard and Sterling for a while. I also have Fraser and Richarlison - with no premium attackers.

    When Pogba, Feilpe Andersson and Son hit form, I had trouble deciding which midfielder I wanted to dump. I got Fraser at 5,5 and I still feel he is insane value, despite some tough fixtures. I also got Richarlison early and he had great fixtures. I ended up sticking to my midfield, which in the long run might pay off, but I lost some momentum because my team wasn't flexibe enough to make changes for in-form midfielders.

    If I had gone for 2 premium midfielders and one premium attacker, it would be way easier to make room for either Pogba, Felipe or Son. Having 3 premium midfielders makes your midfield less flexible, unless you are willing to dump players like Fraser and Richarlison when they're in a slump.

    1. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      Some good points there. I'll be interested to see what Mark does as he has the same premium players as you. It proved very successful early on, but perhaps splitting premium players between midfield and attack is now the way to go.

  8. Subzero (-4)
    • Fantasy Football Scout Member
    • 9 Years
    5 years, 2 months ago

    Amazing work mate. Thank you so much for this analysis. This has really helped with my wildcard decisions. Top top work 🙂

  9. Sky Player in FPL
    • 5 Years
    5 years, 2 months ago

    Great article.

    Given a free choice I'd want 1. Salah, 2. Sterling, 3. Hazard

    Currently at on OR of 677 and only have 2 of the above premiums in Kane and Hazard. Think I need to start considering a third soon having just about got away with it over the festive fixtures (thanks in the main to Heung-Min Son). Salah the no.1 choice but the sacrifices required elsewhere are putting me off. Could be a season defining decision, and has been made all the more difficult with City now beginning to turn it on again.

    Basically stuck between wanting the ones's I don't have, while not wanting to lose the one's I do, so think I could even be tempted to prolong the decision and just get Sane in for Son while he's away in order to preserve the rest of the team.

    The FOMO on Salah is huge compared to the rest though. He's the only one that could be a genuine perma-captain for the remainder of the season for those that have him.

  10. Fire Chief Moe
    • 13 Years
    5 years, 2 months ago

    Great article, I currently have Salah and Kane, and it served me well over Christmas, I need to ship out Son, I could move up to Hazard but it would require me to downgrade Trippier to a 4.3m DEF which I am not too keen on doing, especially as I probably wouldn't captain Hazard over the next few weeks, I am considering KDB for Son as a differential, or Eriksen, it's a tough choice.

  11. Make Arrows Green Again
    • 7 Years
    5 years, 2 months ago

    Great article, many thanks indeed.

    My thoughts are interestingly circular on this, because on one level we are breaking down XG and delta but in order to make sense of them we have to marry them up again.

    If we are saying that XG is changes and delta is conversion of those chances, the potential 'stat to rule them all' is therefore just XG plus delta: how many chances a player is getting and how they are converting them?

    This of course just equals goals!

    So, boiling it all down, does this just mean that if a player has scored 6 in the last 6, he is most likely to score 6 in the next six?

    This is when the delta really comes into play on its own. If a player's delta shoots up then we can infer that he is suddenly getting a better quality of chance which favours him as a player?

    A classic example would be if Chelsea were to switch to classic, aerial crossing to Morata's head, instead to trying to thread through-balls into him when, as we all know, he lives offside. If his delta were to shoot up and his underlying stats were to remain the same, we can start to link this to a change of tactics, and vice versa.

    This highlights a key failing of stats, which is that one striker's big chance is another's difficult chance, because he might lack the agility or reaction times to slot in a rebound, or the aerial prowess to accurately head home a cross.

    So based on this article and my thinking around it as a result, I think we should be on the look out for dramatic changes in delta and jumping on these players. What do you think?

    Thanks again - really interesting.

    1. Make Arrows Green Again
      • 7 Years
      5 years, 2 months ago

      *If we are saying that XG is chances

    2. Deulofail
      • 8 Years
      5 years, 2 months ago

      I see what you're saying. It sounds initially that you're saying the delta would increase when the xG increases (better chances = more goals), but of course, that doesn't make sense. What you mean is that the chances are better suited for the striker, so the xG doesn't necessarily increse, but the delta does! I think that's a great idea and certainly has merit.

      My issues would be regarding the (i) sample size and (ii) causation.

      (i) Aguero scoring a hattrick this season is not enough to predict that on another day he will score more goals (sample too small). In which case you would need prolonged increase of xG (compared to usual individual levels), which would mean waiting for a season to be over untill you could conclude that it was time to buy Aguero 😀

      (ii) In the same case, there is no way to know what has caused Aguero's hattrick, which as you say is a failing of using stats here, though I would add "out of context". In theory, stats could explain everything to 100% accuracy, if you had enough of the and a manner to record space-time events with 100% resolution (The Matrix could accurately describe itself with numbers). But we have to realise that stats we have available to us are tools. And your own proposal is bitten in the aris by your own criticism, which is that it lacks the context to make it powerful. We would likely give meaning to the sudden changes in xG delta based on contextual information we receive from the media and other influencing biases. A hattrick married with a good team performance is proof that they have finally gelled as a unit and Aguero will now achieve a record breaking season!!!! (Or is it?).

      So we would be stuck with in interesting statistical surge of xG delta and no way to tether it to reality without painting our own biased stories, prescribed to us by Big Brother himself 🙁

      I prefer xG by itself, so we can allow our biases to take over asap without going through all the rigmarole of interpretation, only to realise that it was all evident from watching the game in the first place (or predetermined by internal bias towards Arsenal players)

      1. Patch
        • 6 Years
        5 years, 2 months ago

        Wow this is deap, I feel like a pheasant but just about holding onto the jist of this thread!

        1. Patch
          • 6 Years
          5 years, 2 months ago

          *deep

        2. Deulofail
          • 8 Years
          5 years, 2 months ago

          It's just poorly written, don't worry.

      2. TopMarx
        • Fantasy Football Scout Member
        • Has Moderation Rights
        • 11 Years
        5 years, 2 months ago

        This is a good understanding "What you mean is that the chances are better suited for the striker, so the xG doesn't necessarily increase, but the delta does!"

        Essentially when a team plays to the strengths of a particular player, his delta should go up.

        I always think a good use of stats is comparing past performance to present performance, if you suspect a positive change in the team or individual player, then this should be reflected in the stats.

    3. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      I think you are right to dig down to individual player qualities, and how those affect stats.

      However this is incorrect:

      "If a player's delta shoots up then we can infer that he is suddenly getting a better quality of chance which favours him as a player?"

      The better quality of chance is contained in the xG, if his delta shoots up it means he is converting more chances than would be expected of the average player. For instance possibly scoring more shots from distance. Shots from distance have a low probability of resulting in a goal, score one or two and your xG Delta will shoot up.

      Take a look at my comment towards the bottom of the page and see if that clarifies things.

  12. Maximus Bonimus Pointimus
    • 14 Years
    5 years, 2 months ago

    Ah good old XG 🙂

    “The idea behind xG is to quantify the likelihood of a goal being scored from a particular chance. The higher the xG, the better the opportunity.”

    There is one crucial addition required to complete this sentence - it should read as follows:

    “The idea behind xG is to quantify the likelihood of a goal being scored from a particular chance, FOR THE AVERAGE PLAYER. The higher the xG, the better the opportunity.”

    And given that there’s is no such thing as an average player, it’s such a dodgy metric for FPL decision making.

    Kaku’s chance would be scored 0.61 times by the average player, 0.25 times for a below average player, 0.75 times for an above average player, 0.95 times for a world class player who’s in good form etc...

    When you add in the tiny sample sizes we need to operate within when managing an FPL team (5, 10, 15 games etc), and the effect of other important variables e.g. player form and confidence, you may as well be banging some tin cans together and using the loudest noise to select a player.

    1. Maximus Bonimus Pointimus
      • 14 Years
      5 years, 2 months ago

      It I totally agree with the concept of the big 6 & nice article 🙂

      1. Maximus Bonimus Pointimus
        • 14 Years
        5 years, 2 months ago

        *but!!!

    2. Make Arrows Green Again
      • 7 Years
      5 years, 2 months ago

      Great points, and, lol, see above 🙂

      "one striker's big chance is another's difficult chance, because he might lack the agility or reaction times to slot in a rebound, or the aerial prowess to accurately head home a cross."

    3. Make Arrows Green Again
      • 7 Years
      5 years, 2 months ago

      Could it not be argued though that if Kane is career 1.25x delta guy, that he is 25% better than "average" and that this delta could be applied as a consistent modifier to all his XG stats?

      What I mean is that in a way delta itself helps to define who's average, who's world class etc etc.

      Aguero's delta is not very good this season. We know he's world class. Could City not be playing to his strengths this season? If they change system and his delta shoots up then we know they are getting the most out of him.

      Having said that, an easier metric may just be 'goals scored' 🙂

      1. Maximus Bonimus Pointimus
        • 14 Years
        5 years, 2 months ago

        Yeah possibly, although I do think even Kane being a career 1.25x guy is still probably quite limited for classic FPL purposes - i.e. who is going to score points in this limited 5/10/15-game window ahead. Kane is arguably entering his prime years at 25, so he could be a career 1.35, 1.45, 1.5 or more by the time he's 29-30, or he could take the Rooney route, and have peaked at 23/24 and his career XG delta could now decline for the next 5 years, its hard to know.

        Given that his game doesn't really rely on pace you'd imagine he should have longevity and hopefully it will be the former... I just don't know that the one stat to rule them all even exists... picking FPL players still has a huge analogue/intuition element that stats are good at summarising retrospectively, but often not so good prospectively!

      2. Deulofail
        • 8 Years
        5 years, 2 months ago

        Good discussion guys. See my post below for an expansion on delta "modifiers" and group averages.

        FWIW I agree with Maximus' skepticism about xG given all the variables at play in football. But I still think it's a great metric, relative to most, including "goals scored".

    4. Magic8Ball
      • Fantasy Football Scout Member
      • 13 Years
      5 years, 2 months ago

      I agree, but as the old saying goes "form is temporary but class is permanent" I view xG as a good way of identifying which players have that natural ability/class to convert their chances. I find that ranking attackers by xG is useful when making selections because it gives me confidence that I'm buying quality and some assurance of decent returns over a 5-10 game horizon.

      Those players who don't finish as well but still score goals due to a higher volume of chances are still very viable, but much more susceptible to low points when playing better defenses or a drop in form of the team as a whole. That's why well-above-xG-average players like Kane are fixture proof, because they typically only need one or two chances, and also provide more consistent returns, because they still score in games with few chances.

    5. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      You make a good point, and I did in fact edit that sentence down as I wanted to keep it very simple.

      Under xG Delta I say: "If a player has 12 goals from an xG of 9, then he has scored 3 more goals than we would expect from a player of average ability."

      So I got the average bit in there 🙂

      I agree that a good chance for one player is not necessarily a good chance for another player.

      It was interesting reading Kenneth Tang talk about beating his friends at fantasy NBA despite not watching basketball:

      “Every match there is 100-something to 100-something, you know? A lot of rebounds, a lot of field goals. Much easier because the stat is a lot more standard, a lot more static. Too easy.”

      Tin cans in FPL, very true.

  13. Rasping Drive
    • Fantasy Football Scout Member
    • 14 Years
    5 years, 2 months ago

    I think Joe summed it up quiet well on the scoutcast when he mentioned (paraphrasing) whichever of the big hitters you have in your team, stick with them unless their stats drop considerably. Doing the hokey-cokey, chasing last week’s points, will only lead to a permanent residency in the nearest asylum.

    1. Make Arrows Green Again
      • 7 Years
      5 years, 2 months ago

      Wibble 😉

  14. tbhogal
    • 14 Years
    5 years, 2 months ago

    Guys, which moves are best for my team let me know pls? 2Ft's 2.3 itb...

    A) Patricio for Guaita & FAnderson for Sane
    B) Lovren for TAA
    C) Lovren for TAA & FAnderson for Eriksen

    Patricio (Hamer)
    Digne, Lovren, Alonso (AWB, Bennett)
    Pogba, Anderson, Rich, Salah (Hojberg)
    Rashford, Kane, Jimenez

  15. Az
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 14 Years
    5 years, 2 months ago

    Absolutely fantastic article TopMarx indeed!!

  16. Mingo
    • Fantasy Football Scout Member
    • 7 Years
    5 years, 2 months ago

    There's been some excellent discussion on a fantastic article, thanks very much Topmarx. It's nice to see content of this nature and the subsequent discussion in the members section, more of it please!

    I made a choice to go with Auba & Kane from GW15 & 16 respectively, this was a long term view based on their roles within the team, their importance to it and with their run of fixtures. They've been a great source of points the last few weeks, the only issue has been choosing the top point scorer each week. The area where I've really suffered has been midfield. I chose to go with Sterling at what will probably prove to be City's worst run of the season, coupled with no Salah has led me to slide 100k places over Xmas.

    I have to rectify this by getting Salah in, I don't want to lose Sterling as I think he will be an important asset over the next short run of games. I'll be leaving the team alone this week, I've been downgrading my bench places to cheapies over the last few weeks to claw back as much money as I can. After this weekend has passed I think I will be moving Auba, Snodgrass & Alonso to Rashford, Salah & Bednarek, Digne can then become TAA or another suitable option (maybe Pereira) in the following 2-3 weeks. If I do the same move with Kane then I can do Alonso to TAA and keep Digne.

    It feels wrong to get rid of Kane or Auba but no Salah and Liverpool's upcoming fixtures will kill me.

    1. WATERMELONS
      • 6 Years
      5 years, 2 months ago

      Im in the exact same predicament. It kills me to take out one of these two premium strikers for Salah!

  17. Deulofail
    • 8 Years
    5 years, 2 months ago

    Great article. Really very useful. But personally I think it's tarred by your statement that "Everyone will regress to the mean". If this were true, it makes the whole analysis of historical xG deltas meaningless, since over- or under-performing your xG would always be temporary.

    This seems to be the most persistent and confusingly fuzzy myth about xG and FPL stats in general. It's used to confirm biases when it fits, (e.g. "this feels unstainable"). (Not to mention that the word "regress" seems to have different meanings in the article and for different people in the comments). It suggests that being good a shooting necessitates that you will regress to being average at shooting, or that being poor entails that you will improve.

    The reason the extreme values will probably appear to regress to a group mean (note: This doesn't mean they will reach the average, but merely to become less extreme), is because the numbers are extreme for the individual player. In other words, you regress to your own mean. Each player has a cetain level of ability, which, once you roll the dice enough times, will be evident in the data. This is because the data represent the individual actions of an individual.

    Simply belonging to wider categories (groups) does not make the data representative of those categories because you are missing the data from all the other constituents of that group. (You would need to compare data from one group to data of the same group under different circumstances (e.g. a different season)).*

    However, since we cannot roll the dice enough times, we can only use common sense to predict regression, and that may be foolhardy anyway. The simple act of taking a penalty or having the ball rebound off your nose into the net has a massive impact on your xG delta, and I'll be damned if I can control all the factors that went into these kinds of moments.

    You need a lifetime of evidence from each player to be able to start to predict what kind of delta they should be achieving, but of course they are only in peak form for a few years. There just aren't enough goals scored in a season for us to predict xG delta regression to a mean. And there's no reason to believe that all players are averave, but temporarily playing worse or better than average (regarding every single metric).

    I'd be interested if anybody could offer a method for predicting xG delta trends. Personally, I treat it all with a massive pinch of salt. xG itself is useful, and then I basically use my own mental multiplier based on how good I think a player is, which is essentially guesswork based on the eye test, conversion rates, contextual information about player and team confidence and club news, and if I must, a little bit of historical xG delta 😉

    But really, this level of detail is not powerful enough to have a positive effect over more basic, casual observations about who looks like they will score goals. In fact, it can often muddy the water in my experience.

    But as I say, lots of great information in this article. Many thanks for your contributions. If I'm wrong about anything I've said, I will gladly take on board any rebuttal. There are probably hundreds who will read this who know more than me. Speak up!

    *I'd like to think that the confusion comes from xG being a measure which inherently compares an individual to a group ("How many goals would an average player (group) score from the shots taken by this player (individual)"), but it seems that the same regression myth applies to all "underlying" (and overlying?) stats.

    1. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      I agree with most of what you say.

      What I meant was not "Everyone will regress to the mean", but "Everyone will regress to their own mean". I hope that is clear from the points I make in the rest of the article?

      I think if we look at 5 seasons of worth data, look at the variance from season to season, we can get a decent idea of what we can expect to happen over the current season.

      I agree with you - it's not going to be 100% accurate and judgement will have to play a part. I guess that's where the skill lies.

  18. jtreble
    • 7 Years
    5 years, 2 months ago

    Great work TopMarx.

    1. Deulofail
      • 8 Years
      5 years, 2 months ago

      😀

  19. Greyhead
    • Fantasy Football Scout Member
    • 5 Years
    5 years, 2 months ago

    Awesome work as always Top Marx. Salah, Haz and Auba for me plus a big sofa for whenever City or Spurs play.

  20. LosBlancos
    • Fantasy Football Scout Member
    • 6 Years
    5 years, 2 months ago

    Great article!

  21. Jacket Potato
    • 7 Years
    5 years, 2 months ago

    Really top article with solid analysis, has helped me to make my transfer decision this week and will influence over the next few weeks, thansk

  22. JohnnyRev7
    • 13 Years
    5 years, 2 months ago

    I have Sterling, Aubameyang and Kane (Pogba and Eriksen) and 2FTs.

    I want Salah, but just feel it's a GW too early for my proposed transfers and the balance of my team because of fixtures. So I'll do a simple upgrade (non playing to playing) and have 2FTs next GW when the fixtures are more obvious.

    And hide behind the sofa when LIV play.

    Top marks to TopMarx for a top article. 🙂

  23. TopMarx
    • Fantasy Football Scout Member
    • Has Moderation Rights
    • 11 Years
    5 years, 2 months ago

    There are some really interesting points that have been made, I've been thinking a lot about how we use xG.

    A criticism of my own article - and actually not just mine but most of the articles I read in the media - is that we/I am rather lazy with the term 'finisher' used in the discussion of xG. Here's why:

    "players who consistently exceed their xG can be regarded as expert finishers."

    Is not necessarily true. The problem is the definition of the word 'finisher' and whether it accurately describes what xG is telling us.

    Let's compare Coutinho with Aubameyang last season, who played 14 and 13 matches respectively:

    54 shots // 15 SiB // 7 goals // 4 goals from outside box // 3.51 xG // +3.49 xG Delta // Coutinho
    31 shots // 28 SiB // 10 goals // 0 goals from outside box // 8.63 xG // +1.37 xG Delta // Aubameyang

    Coutinho takes very few shots from inside the box and is very good at shooting from distance. According to the xG model, a shot from outside the box has a very low probability of resulting in a goal. Therefore you can take loads of shots from outside the box and it won't increase your xG by very much. Coutinho's 4 goals from 39 shots outside the box is impressive and helps give him an xG Delta of +3.49.

    But would you really describe Coutinho as a better 'finisher' than Aubameyang?

    No. I don't think that because he's good at scoring from distance that makes him a better or more clinical finisher than Aubameyang.

    "Clinical finishers are forwards that specialise in their accurate shooting ability. They are "clinical" in that they need few opportunities to score a goal ...They can be identified by their high goal to shots ratio."

    Using that definition Aubameyang with 10 goals from 31 shots is a much better finisher than Coutinho with 7 goals from 54 shots.

    When the term 'expert finisher' or 'clinical finisher' is used in relation to xG it means a player who scores more goals than the xG model would expect from the average player. I think that definition is different to our everyday use of the term 'finisher'.

    1. Deulofail
      • 8 Years
      5 years, 2 months ago

      Top marks for self-criticism. It pays to leave room for improvement so you can gain the extra points in the reflection and further research section. 😀

    2. TopMarx
      • Fantasy Football Scout Member
      • Has Moderation Rights
      • 11 Years
      5 years, 2 months ago

      For those interested, here is the data for the six players discussed in the article, shots in the box (SiB) only (data for the past five season, source: understat.com ):

      211 SiB // 52 goals // 38.04 xG // +13.96 Delta // Hazard (0.25)
      409 SiB // 105 goals // 91.12 xG // +13.88 Delta // Kane (0.26)
      293 SiB // 72 goals // 60.67 xG // +11.33 Delta // Salah (0.25)
      442 SiB // 91 goals // 91.09 xG // -0.09 Delta // Aguero (0.21)
      232 SiB // 45 goals // 47.52 xG // -2.52 Delta // Sterling (0.19)
      400 SiB // 107 goals // 109.83 xG // -2.83 Delta // Aubameyang (0.27)

      The number in brackets at the end = goals / SiB.

      So by the everyday definition Auba is the most clinical finisher of the six because he has the best goal to shots ratio. However by the xG definition he is the least clinical finisher of the six because he has the worst xG Delta.

      One question I now have is - why does Auba get better chances in the box (according to xG) compared to the other five players?

      At the end of the day, Kane and Auba take a similar amount of shots and score a similar amount of goals, Auba happens to have a higher xG than Kane. So just bear that in mind when comparing them using xG stats.

      1. Deulofail
        • 8 Years
        5 years, 2 months ago

        https://memegenerator.net/img/instances/71451135/the-thing-about-arsenal-is-they-always-try-and-walk-it-in.jpg

        I guess he takes more shots from close range, both scoring and missing more high xG chances than other players.

        1. TopMarx
          • Fantasy Football Scout Member
          • Has Moderation Rights
          • 11 Years
          5 years, 2 months ago

          No to the first bit (Kane takes more) but yes to the second bit 🙂

          1. Deulofail
            • 8 Years
            5 years, 2 months ago

            Not even a higher proportion of his shots from close range? I feel like Kane shoots more from distance (low xG values) than Aubameyang, relative to their respective close-range shot counts. I have nothing to back that up though, besides my tingly senses.

            1. TopMarx
              • Fantasy Football Scout Member
              • Has Moderation Rights
              • 11 Years
              5 years, 2 months ago

              You are right, he does take a higher proportion of shots from distance. But look at the table in my comment above, that only considers shots in the box, Kane still as a Delta much higher than Auba's. Or another way of saying the same thing, Auba has a higher xG from a nearly identical number of shots in the box.

              1. Deulofail
                • 8 Years
                5 years, 2 months ago

                Yeah, possibly because Aubameyang's shots are higher in xG value (closer to the goal) than Kane's on average. Doesn't that make sense?

                1. Deulofail
                  • 8 Years
                  5 years, 2 months ago

                  Also, Kane has 3 goals from penalties in the Prem this season compared to 1 from Aubameyang. I don't have all the historical penalty data, but that could be boosting Kane's delta, assuming pens are included in your xG numbers and delta? I don't know what xG value a penalty is, roughly.

                  1. TopMarx
                    • Fantasy Football Scout Member
                    • Has Moderation Rights
                    • 11 Years
                    5 years, 2 months ago

                    For OPTA the xG value of a penalty is 0.79.

                    They both take penalties, Kane is slightly better than Auba. Kane has 17 successful pens and 3 misses in the PL. xG from 20 penalties = 15.8 so Delta of +1.2

                    Auba is 12 from 17 in PL and Bundesliga. xG from 17 pens = 13.43 so Delta of +1.43

                    Some different but not enough to explain the gap between them

                    1. TopMarx
                      • Fantasy Football Scout Member
                      • Has Moderation Rights
                      • 11 Years
                      5 years, 2 months ago

                      *
                      -1.43

                2. TopMarx
                  • Fantasy Football Scout Member
                  • Has Moderation Rights
                  • 11 Years
                  5 years, 2 months ago

                  You mean within the penalty box his shots are closer? I don't think that's true... I will look at the shot location maps.

                  There are other factors that affect xG:

                  - Angle of the shot
                  - Did the chance fall at the player's feet or was it a header?
                  - Was it a one on one?
                  - What was the assist like? (eg long ball, cross, through ball, pull-back)
                  - In what passage of play did it happen? (eg open play, direct free-kick, corner kick)
                  - Has the player just beaten an opponent?
                  - Is it a rebound?

                  Ultimately Auba fashions better chances for himself in the box, either due to his ability as a footballer or thanks to his teammates, or both.

              2. Kiwivillan
                • Fantasy Football Scout Member
                • 9 Years
                5 years, 2 months ago

                As a complete aside a shot on target includes one that the keeper saves. Some players have better ability at on target placement and stats don't cover that. Hell I see it in my social league. 2 players will hit the target 10 times and one will manage to hit straight at the keeper more than the other

                1. Deulofail
                  • 8 Years
                  5 years, 2 months ago

                  Jesus seems a prime example of that. He can score 4 goals on his day, but on other days he'll just get 10 shots on target.

                  1. Deulofail
                    • 8 Years
                    5 years, 2 months ago

                    (exaggeration on the 10 shots on target, but he does have high accuracy and low conversion despite an amazingly low mins per xG)

                2. TopMarx
                  • Fantasy Football Scout Member
                  • Has Moderation Rights
                  • 11 Years
                  5 years, 2 months ago

                  You will probably be a fan of "post shot xG" or "xG2".

                  xG doesn't look at any info after the shot is taken, it only considers factors leading up to the shot.

                  Post shot xG considers what happens to the shot and where it ends up.

                  Have a read here http://thepowerofgoals.blogspot.com/2018/01/after-shot-xg2.html?m=1 and https://www.infogol.net/en/blog/analysis/expected-goals-who-is-the-premier-leagues-best-goalkeeper

                  So far I've only seen it used in discussion of keeper quality.

      2. Ruth_NZ
        • 9 Years
        5 years, 2 months ago

        "So by the everyday definition Auba is the most clinical finisher of the six because he has the best goal to shots ratio. However by the xG definition he is the least clinical finisher of the six because he has the worst xG Delta.

        One question I now have is - why does Auba get better chances in the box (according to xG) compared to the other five players?"

        I am glad that you decided to look at this, TM.

        A finisher is a player that rounds off moves and turns them into goals. That's the classic meaning anyway. When people say 'Chelsea played well but lacked the ability to turn all their possession into goals' they are saying that the finishing was lacking. I'll try not to mention Morata. 😉

        Now, if you examine this you will see that part of what we generally mean by 'finisher' is THE ABILITY TO BE IN THE RIGHT PLACE AT THE RIGHT TIME. To 'get on the end of things' and turn good play into goals. Could be a midfielder like Lampard that times his late runs into the box brilliantly. Could be a wide forward like Salah that makes runs that - in his case because of his extreme pace - enable a big chance to be created rather than someone shooting from distance. Or could be a striker like Aubameyang who has a knack of anticipation and a reaction speed that gets them to the ball where others might be a yard short.

        This produces an odd contradiction because the number of chances and 'big chances' that player gets will be more than others. BECAUSE THEY GET THERE. If Morata is a yard short of the finishing position it doesn't count as a 'big chance missed' in the xG model. It doesn't count as any chance at all, in fact it probably goes down as a misplaced pass. But perhaps Aubameyang, with his lightning instincts (Aguero is another), manages to get there. He is at the high end of excellence by even managing to do that so if he misses some of those it's not surprising, he did very well to get there at all. But for him it goes down as a chance or big chance missed if he fails to convert it.

        This is absolutely key to understanding Salah's inability to do well in the BPS. He misses too many chances. But what the BPS model fails to take into account is that for any other player than Salah many of those chances wouldn't materialise at all. SALAH HIMSELF IS RESPONSIBLE FOR MANY OF THE CHANCES HE GETS in other words. It could almost be said that he creates his own chances, or helps to. And this affects xG modelling when applied to an individual player as well.

        Which is all a roundabout way of saying that xG isn't a complete or definitive measure when applied to an individual's play. You have to apply some understanding of what the player/team is doing to it. Which is the underlying question that your article is examining, if I understand it right?

        1. TopMarx
          • Fantasy Football Scout Member
          • Has Moderation Rights
          • 11 Years
          5 years, 2 months ago

          It's definitely what I'm getting at - trying to understand what these xG numbers mean for different players.

          It makes perfect sense what you say about how certain players - due to their footballing ability - are able to fashion better chances, and therefore get a higher xG as a result. Strikers will low xG and low Delta are not worth paying much attention to (Morata).

          I think what I discovered in researching this article is the importance of making informed comparisons between players. For instance to highlight the Kane v Auba stats in the post above:

          409 SiB // 105 goals // 91.12 xG // +13.88 Delta // Kane (0.26 goals per SiB)
          400 SiB // 107 goals // 109.83 xG // -2.83 Delta // Aubameyang (0.27 goals per SiB)

          They are so closely matched on SiB and goals in the box that the disparity in xG is somewhat irrelevant. At least it shows that we can expect Auba to get a better xG than Kane and still get a similar amount of goals. Although it's not a perfect comparison (it never is!) some obvious factors are that Auba played for Dortmund and has a new manager this season at Arsenal. This seasons data:

          53 SiB // 12 goals // 11.47 xG // +0.53 Delta // Kane (0.22 goals per SiB)
          46 SiB // 12 goals // 13.27 xG // -1.27 Delta // Aubameyang (0.26 goal per SiB)

          It seems the trend is continuing despite the change in circumstances for Auba. I guess we will continue to see Auba get into (create) good goalscoring chances and miss a number of them. That period in the early season when he scored 10 goals from 10 shots https://www.premierleague.com/news/929716 was clearly an anomaly.

          It's been interesting to see where each players 'normal' is.

          Now, just on the definition of 'finisher', despite Auba's ability to get himself good chances (and a higher xG as a result), I think the fact that he doesn't finish as many of these good chances should count against him. If he fashioned those chances AND scored a higher percentage of them, then we would be talking about a truly special player imo.

  24. Nesrevi
    • 7 Years
    5 years, 2 months ago

    Well this didnt help me at all, i was wondering who to put in my team, but after reading this i want them all :)..

    Great piece