Ole and xG

Discussion in 'Manchester United Forum' started by Pogue Mahone, Mar 7, 2019.

  1. Mar 7, 2019
    #81

    OnlyTwoDaSilvas Gullible

    Joined:
    Feb 4, 2013
    Messages:
    17,248
    Location:
    The Mathews Bridge
    So, what's the deal then? Have we scored 10 pure bangers out of nowhere that have fluked their way in? Surely the only way a goal wouldn't count as an expected goal is if it goes right through a flapping goalkeeper, or is a calamitous own goal that the attacking team had little involvement in. If a striker gets a shot on target and it goes in, how could it possibly be an 'unexpected' goal?

    It makes a bit more sense the other way. I.e. a goalkeeper making miraculous saves of what looked like a certain goal, of which De Gea frequently does, so that is fairly clear. There was about 4 of those alone vs Spurs.

    But which are these goals that United have scored that aren't part of the "xG' total? I'd be interested to see these 10 extra "holy shit how did they score that?!" goals that we've amassed. Whoever is calculating this stat should provide that context, if they don't already.
  2. Mar 7, 2019
    #82

    vodrake Full Member

    Joined:
    Nov 22, 2012
    Messages:
    2,786
    Correct me if I'm wrong, but XG is based on the average likelyhood of scoring an opportunity irrespective of the quality of the player taking the shot? It's saying that we've scored more than the average player would have been expected to score in those situations.

    So it's saying that our players are better than the average player? In which case, duh?

    Same for the goals conceded. We've conceded less than the expected average goalkeeper would have conceded, because we have a much better than average goalkeeper...

    People are skewing this to say we've just been lucky, but it seems to me to just be saying that we surprisingly actually have good players?
  3. Mar 7, 2019
    #83

    Cait Sith Full Member

    Joined:
    Jun 8, 2014
    Messages:
    1,214
    Messi is literally the only player who consistently scores more than he should and it's completely in line with what every Barca fan will tell you.
  4. Mar 7, 2019
    #84

    calodo2003 Full Member

    Joined:
    Feb 8, 2014
    Messages:
    1,411
    Location:
    Florida
    Not much of an equivalence there, but I get your point.

    I have also looked at ‘Billy Ball’ & the overuse of metrics in baseball in the same light even when my hometown club uses such tactics to their betterment. Feels a bit Pollyanna & reaching.
  5. Mar 7, 2019
    #85

    mav_9me Full Member

    Joined:
    Feb 9, 2009
    Messages:
    4,303

    Thanks for the table. On that we have 58 goals for xG of 53 and GA of 38 for xGA of 40.45. Doesn't appear to be that much of a discrepancy does it?
  6. Mar 7, 2019
    #86

    Raees Legal Guardian of the Football forums

    Joined:
    May 16, 2009
    Messages:
    27,607
    Yeah I get that argument but XG is a reactive stat in that it tells you what has happened before. It can be used as a predictor of some sort I guess but the best way to use it is analyse past performance and say well we did kind of wing it to some degree but it doesn’t 100% confirm we will get back down to earth because we could easily in the next game start creating chances galore and playing teams off the park.

    XG also doesn’t reflect the mental qualities of teams - some teams are great at carving up defences and creating chances but essentially bottle it under pressure - United under Ole have demonstrated nerves of steel which is in of itself an admirable and essential quality.

    To call it a useless stat altogether is silly as I do think it’s one of the better ones to come up in recent years.
  7. Mar 7, 2019
    #87

    mav_9me Full Member

    Joined:
    Feb 9, 2009
    Messages:
    4,303
    A good point I saw on Twitter.

  8. Mar 7, 2019
    #88

    U99ted Full Member

    Joined:
    Jan 1, 2013
    Messages:
    1,715
    Go by xG and how “accurate” it is and we don’t even beat Fulham. When we clearly swatted them aside.
  9. Mar 7, 2019
    #89

    Dion Full Member

    Joined:
    Mar 18, 2010
    Messages:
    3,312
    No actually. Mo Salah had for the last 4 seasons before this one, Leroy Sane has for 4 out of the last 5 seasons, so has Lacazette, Kane does it every year, Song has done it every year but his first at Spurs, Eden Hazard has done it for the last 5 years, Suarez did it for 5 years straight before age caught up with him, Antoine Griezmann has done it 5 years in a row, Nabil Fekir has done it 5 years in a row, Dybala has done it 4 years in a row, Icardi did it 4 years in a row, Ciro Immobile has done it 3 years in a row at Lazio, Higuain did it 4 years in a row. And those are just stats from players who did it in the last 5 years.

    Messi has scored 165 league goals in the last 5 seasons with an xG of 137. Kane has 120 with an xG of 101. That's almost the exact same level of over-performance over the same time period. Eden Hazard has over-performed even more than both.
    Last edited: Mar 7, 2019
  10. Mar 7, 2019
    #90

    Rish Sawhney New Member

    Joined:
    Mar 15, 2015
    Messages:
    378
    Location:
    State College
    Well Barca should fire him then. Clearly it can't last.

    EDIT: Also, since it looks like Liverpool and City are also overperforming their xStats does that mean their form can't last either?
  11. Mar 7, 2019
    #91

    FlawlessThaw most 'know it all' poster

    Joined:
    Oct 26, 2005
    Messages:
    27,393
    Seems like the top players as well as the top clubs in the PL outperform xG. Funny that.
  12. Mar 7, 2019
    #92

    Cheesy Bread with dipping sauce Scout

    Joined:
    Oct 16, 2011
    Messages:
    35,852
    Location:
    Scotland
    True. If he'd, say, lost two more games then he has, and if he'd drawn a couple more, then we'd likely still be happy. And in many ways last night, if anything, almost makes his run even better in that the one game he's lost was in a two-legged tie we eventually emerged from successfully anyway.
  13. Mar 7, 2019
    #93

    Teja Full Member

    Joined:
    Aug 17, 2014
    Messages:
    164
    Location:
    Boston, MA
    I think right now it's just too small of a sample size. We have outliers like our PSG smash and grab skewing the numbers too much. I think xG is very valuable but it definitely needs more context here (the fixtures we went through incl. Pool, Chelsea, Arsenal, PSG 2x are expected to be tight cagy affairs where we don't cut these teams open as we want). We've also had the 5-1 after Mou left where the xG for us was definitely not 5.

    So far, the football passes the eye test, the whole squad is behind the manager and morale is high. These things can't be quantified but they're probably the most important things a manager can do for the squad and that by itself deserves a permanent contract.
  14. Mar 7, 2019
    #94

    Ace of Spades Full Member

    Joined:
    Jan 24, 2013
    Messages:
    3,204
    :lol:
  15. Mar 7, 2019
    #95

    Anustart89 Full Member

    Joined:
    Jun 6, 2013
    Messages:
    8,317
    If you take a hundred shots from the edge of the box and a little bit to the left you'll maybe see that 10 times out of 100 it goes into the net. That makes the next shot from that position in similar circumstances have an xG of 0.10, ie 0.10 goals expected from every shot in that situation, or one goal scored per ten shots from that position.

    What that means is that if a team takes ten of those shots in a game (and creates no other chances) and score two goals, they will have amassed a total xG of 1.0 (0.10x10) while having an actual goals scored value of 2. In that situation, one interpretation is that the players are over-performing and would given enough time regress back to the mean (ie score one goal instead of two), while some others say that it's fairly natural that a team with above average players will outperform the average number.

    What this essentially means for us is that we've scored more goals than we would have been expected to based on the amount and quality of chances we've created. If we had scored fewer goals than our xG value, then it would have meant that we'd missed a lot of good goalscoring opportunities (but depending on how big the xG number is, it would have been an indication of how many chances we create as well).
  16. Mar 7, 2019
    #96

    Dion Full Member

    Joined:
    Mar 18, 2010
    Messages:
    3,312
    The important clarification here is the "you" is every shot that Opta has measured being taken from that spot. That means every Ronaldo pot-shot, every Jonjo Shelvey swipe etc. not just every time Messi does it.
  17. Mar 7, 2019
    #97

    Dion Full Member

    Joined:
    Mar 18, 2010
    Messages:
    3,312
    Spurs should clearly sell Kane and Chelsea should ditch Hazard. They were really lucky they consistently outperformed their xG by so much for half a decade.
  18. Mar 7, 2019
    #98

    NoPace Full Member

    Joined:
    Jan 20, 2014
    Messages:
    5,460
    I think it shows that we don't create enough chances and we give up too many because we don't control games.

    But that's less on Ole and more on us only having 2 world class players in Pogba and De Gea and then Shaw, Lindelof and Herrera are up to the standard. The forwards are still young (Rash, Martial) and inconsistent (Lukaku) but we should back them, but we need 4 starters rather than having an aging Matic, technically poor Smalling, Ashley Young at RB and Lingard/Mata whoever you want to call our RW.

    Plug those last 4 players into City's team. What's their XG?

    Sane-Aguero-Lingard
    --DeBruyne-Silva----
    --------Matic---------
    Young-----------Walker
    --Otamendi-Smalling-
    -------Ederson--------

    and suddenly the defense looks shaky on the ball, Matic has way too much room to cover and while Lingard's composed finishing means his numbers aren't awful, he can't beat a man out wide and is blunting City's attack's on that wing.
  19. Mar 7, 2019
    #99

    Tarrou Full Member

    Joined:
    May 13, 2013
    Messages:
    14,173
    Location:
    Canberra
    To be honest we have rode our luck a little bit under Ole.

    We have still performed brilliantly, but there are two/three results that could've easily gone the other way.
  20. Mar 7, 2019

    Anustart89 Full Member

    Joined:
    Jun 6, 2013
    Messages:
    8,317
    Exactly, so in my world any above average team should be outperforming their xG, which means that doing precisely that shouldn't be used as a criticism of a team, but rather the criticism should be directed at a good team failing to outperform xG for/against.
  21. Mar 7, 2019

    harms Way Staff

    Joined:
    Apr 8, 2014
    Messages:
    18,183
    Location:
    Moscow
    Messi is one of the very few people who completely shits on xG stat throughout many seasons. It's not surprising as his game is beyond any rational explanation.

    Aguero is a great example of how xG work though. I'm sure that all of us can agree that he is a brilliant finisher, but his most important quality is not actually kicking the ball into the net, but getting into the most dangerous positions consistently. A difference between a good and a bad finisher is that, quite often, a good one will be better positioned before actually taking a shot.

    Ronaldo is an even better example. His shots to goal ratio was questioned for years before the xG was even invented, but no other player can get into the goalscoring positions as often as he does (not even Messi). And no one would say that he is a lousy finisher, or even an average one.
  22. Mar 7, 2019

    bosnian_red Worst scout to ever exist

    Joined:
    Aug 13, 2011
    Messages:
    36,504
    Location:
    Canada
    As you would expect when you play PSG/Spurs/chelsea/arsenal/leicester/palace all away and then PSG/Liverpool at home. Not to mention the injuries. Weve done terrific whatever way you look at it.
  23. Mar 7, 2019

    Dion Full Member

    Joined:
    Mar 18, 2010
    Messages:
    3,312
    I see you haven't read my other post yet. I would be interested to hear your thoughts on it.

  24. Mar 7, 2019

    adexkola Arsenal supporter

    Joined:
    Mar 17, 2008
    Messages:
    36,017
    Location:
    The CL is a glorified FA Cup set to music
    Supports:
    orderly disembarking on planes
    What if you're a Bayesian though?
  25. Mar 7, 2019

    Rish Sawhney New Member

    Joined:
    Mar 15, 2015
    Messages:
    378
    Location:
    State College
    So according to this table, the top 4 looks as follows:

    No Team M W D L G GA PTS xG xGA xPTS
    1 Manchester City 29 23 2 4 76 20 71 72.84-3.16 21.25+1.25 68.94-2.06
    2 Liverpool 29 21 7 1 64 15 70 59.07-4.93 21.91+6.91 63.56-6.44
    3 Tottenham 29 20 1 8 56 30 61 48.36-7.64 36.70+6.70 48.35-12.65
    4 Manchester United 29 17 7 5 58 38 58 53.19-4.81 40.45+2.45 47.29-10.71

    What does this say about Liverpool (who have xPTS of 63.56 but have 70 points) and Spurs (who're overperforming even more than we are)?

    Yeah if you ignore all the other people that overperform their xG.

    See:
  26. Mar 7, 2019

    Schneckerl Full Member

    Joined:
    Jun 17, 2016
    Messages:
    2,510
    I think that's interesting and wouldn't discount it as nonsense. The record after Ole took over is just insanly strong and I don't think maintaining a 100 point pace is realistic and expected, so it has some merit. However even if we see some slight drop in the goals scored and conceded it's good enough to contend for trophies. Doing a fantastic job regardless and a massive improvement from Jose in all aspects.
  27. Mar 7, 2019

    jem Full Member

    Joined:
    Jan 14, 2010
    Messages:
    6,201
    Location:
    Toronto
    I agree - it's even worse with baseball. I think it's also just a lot of people making up stuff to try make themselves appear relevant. All you need to do is watch United now, and compare it with the Mourinho era and the difference in quality is there for all to see (but that's the thing: you actually have to watch in order to see it.)
  28. Mar 7, 2019

    Sarni nice guy, unassuming, objective United fan.

    Joined:
    Jan 21, 2004
    Messages:
    50,653
    Location:
    Krakow
    It goes for our non-wins as well. We could have scored against Liverpool and were unlucky with injuries, we should have probably beaten Burnley.
  29. Mar 7, 2019

    Anustart89 Full Member

    Joined:
    Jun 6, 2013
    Messages:
    8,317
    This paints a cracking picture, since you can see that players who are widely regarded as the best around are consistently over-performing their xG. To then draw the opposite conclusion with regards to teams, ie that xG over-performance from a team is bad, which people tend to be doing with the regression to the means argument, is beyond ricidulous.

    What you want to be doing, as a team, is to amass a set of players who can rack up high numbers in the xG department, but still outperform it. If you've got a high xG but are under-performing in terms of goals scored, then your finishing isn't good enough. Now that might win you a league anyway if you're creating so many chances that you'll outscore the opponent anyway, but by buying better players you're not necessarily changing the outperformance ratio, but you're pushing the actual goals scored and xG up to higher levels.
  30. Mar 7, 2019

    bosnian_red Worst scout to ever exist

    Joined:
    Aug 13, 2011
    Messages:
    36,504
    Location:
    Canada
    They dont include penalties into the total, they separate I'm pretty sure. So really its like 36ish expected goals compared to 39 actual. (Big deal right).
  31. Mar 7, 2019

    adexkola Arsenal supporter

    Joined:
    Mar 17, 2008
    Messages:
    36,017
    Location:
    The CL is a glorified FA Cup set to music
    Supports:
    orderly disembarking on planes
    I assure you it's not impossible to do both.
  32. Mar 7, 2019

    AndyJ1985 New Member

    Joined:
    Apr 11, 2016
    Messages:
    8,954
    Yet another silly stat for boring people to argue about
  33. Mar 7, 2019

    Dave89 Full Member

    Joined:
    Nov 30, 2007
    Messages:
    17,553
    Speaking as an auditor, it's not a statistic, it's an estimate. Estimates can say whatever you want them to say.
  34. Mar 7, 2019

    harms Way Staff

    Joined:
    Apr 8, 2014
    Messages:
    18,183
    Location:
    Moscow
    This is a good article on the matter, better than anything that I can come up with on my own:
    https://www.optasportspro.com/news-...als-and-the-repeatability-of-finishing-skill/

    Why the positioning (and the xG score) is more important than the G/xG ratio. Despite the fact that better forwards will score more consistently than worse ones, obviously.
  35. Mar 7, 2019

    Dion Full Member

    Joined:
    Mar 18, 2010
    Messages:
    3,312
    None of that really addresses the issue that Messi is far from the only player who consistently outperforms his xG. It's actually very common among forwards at top clubs.
  36. Mar 7, 2019

    Dave89 Full Member

    Joined:
    Nov 30, 2007
    Messages:
    17,553
    Apologies if I've missed this, but we should do a quick sense check.

    Does the sum of say every premier league player's actual goals less expected goals equal close to zero? If the xg goals figure truly was expected, then over such a large sample size (200 starting outfield players per gameweek, plus say 50 subs, over 38 game weeks = 9500 data points), then the total goals actually scored in the league should be very close to the expected goals.

    If it's not, it's a shit metric by its own definition.
  37. Mar 7, 2019

    kundalini Full Member

    Joined:
    Aug 26, 2004
    Messages:
    4,377
    I believe most of the models have used the data from last 3+ years for the top European leagues, PL, La Liga, Bundesliga, Serie A and Ligue 1 to come up with their figures. Over the course of that period of time (3 to 5 years) the number of goals scored equals the expected goals total. In any one season, the real goals scored total may be slightly higher, or slightly lower than the expected figure.
  38. Mar 7, 2019

    Revan Assumptionman

    Joined:
    Dec 19, 2011
    Messages:
    38,288
    Location:
    Munich
    We have better xG stats under Ole than under Mourinho. However our run of points is better than City's last year, while only an idiot will claim that we are playing better than City last year.

    xG actually agrees with what we see in the pitch, but disagrees with the table (we aren't ever going to get 102 points in a season playing like this, which is what Solskjaertable says).
  39. Mar 7, 2019

    Dave89 Full Member

    Joined:
    Nov 30, 2007
    Messages:
    17,553
    If I understand that correctly, they've confirmed that, working backwards, the models fit the source data used (which you would expect), but for the first full season of new predictions, did it fit?
  40. Mar 7, 2019

    Revan Assumptionman

    Joined:
    Dec 19, 2011
    Messages:
    38,288
    Location:
    Munich
    Can you please help me then to publish the next paper? I am desperately in need for better results than other competing models.