More counting

Further to yesterday’s Dempsey post, I was so pleased with the approach that I thought I’d run a couple more players.

First, the team seems equally likely to do well or badly whether Dickson Etuhu plays or not – his ‘present’ vs ‘absent’ scores are almost exactly the same.       You might expect this, I suppose.

But if you believe these numbers (and why wouldn’t you?), while we miss Dempsey when he’s not starting, the absence of Davies or Gera hasn’t tended to be a problem.    Indeed, we are much more likely to suffer an unexpected result with either starting, and (therefore) much less likely to suffer an unexpected result when they’re missing.

(As a reminder, what we’re looking at here are the times when Fulham have either met, exceeded or faltered against Ladbrokes’ expected result for each game in the last two seasons, split by whether certain players started or not.  NB this is league matches only; Gera’s European magic is a different kettle of fish).

What does this suggest?   The beauty of this approach is that it attempts to cover the problem of which games a player is selected in.   We can run all sorts of numbers but if a player is regularly playing in tough matches then he’ll look bad.  But this way, if we’re expected to lose and do, then no skin of anyone’s nose.   It’s a straight form vs expected results methodology.   

The Davies effect goes along with what we saw with our eyes:  he has played much of the last two seasons injured, and only really started to thrive late on in 2010.   His playing through pain may have affected the team more than we thought.  

Gera?   A tale of two seasons:

In 2008/09 we were a much better team without Gera in the side.  This season we’re still unlikely to really miss him, but it’s a vast improvement, again reflecting what we’ve seen with our eyes.

I can’t really run this sort of analysis on many of the other players, as the settled squad means that most have played most games, so you don’t get any differentials to explore.    The information above, though, does suggest that we need to keep Clint around.


Hotspots:  Neither Davies nor Gera; Davies not Gera; Dempsey not Davies; Dempsey not Gera.

15 thoughts on “More counting

  1. I like Gera. He’s a good player who puts a lot of effort in. And I really like us playing with a linkman and Bobby upfront, as we did last year, rather than trying to make Bobby the link man, as we did the year before. But my number 1 priory this year has been for us to get an upgrade over gera. This is such a crucial role that for us to really move forward we need to have the best possible player there. The gudjonsen rumours are very exciting. And if we got him I’d have no problem whatsoever dropping gera. In the other positions it seems clear to me that duff and Dempsey are a class above. Davies is a solid pro, nothing much more. Gera has not looker good out wide.

  2. I understand your methodology but there are two aspects that I am cautious about:

    1) It relies on the idea that Clint is definitely changing every game he plays and that he is the definitive difference. What happens if he has a bad game or if he is not involved in much of the good things that are happening (goals, assists etc).

    2) It would undoubtedly be a pain, but the other question to ask is who is playing alongside Clint. Is a Clint/Davies combo better than a Clint/Gera?

    I agree completely that we Clint is essential for Fulham moving forward. He is an excellent player and this season may be his to really flourish.

    1. Hi Alex

      I don’t know why 1 would be true. That’s why you need a decent sample size, and why I couldn’t analyse players who’ve been regulars. But over 20-30 games these sorts of effects would even themselves out. As you suggest, not entirely down to the players in question, but over enough games the dynamics of the team will be affected by who’s playing and who isn’t, and these results show that – rightly or wrongly – we’ve overperformed more with Clint in the side and less when he’s not. It might not mean anything, but it then again, it might. Given that the results do make intuitive sense, I’m happy enough with it. It certainly doesn’t prove anything, but all of these angles give you another way of looking at things.

      Wow, point 2 is interesting. I’ll have a look.


      1. Good point about sample size. I don’t mean to be too critical I am just wary of statistics in football because I have yet to be convinced that someone has come up with the correct model for statistical analysis for football in the same way as with American sports such as Baseball (which has been completely rocked by the whole sabermetrics movement, but that is a different story).

        Also, I am a massive fan of Gera and Davies, so was trying to convince myself that they are important to the team!!

        1. I know what you mean, but when you say “statistics in football”, well all we’re doing here is looking at the number of times Fulham have over or underperformed against expectations. It’s not really statistics is it? Just looking at things in a bit of detail. If I say that we were much more likely to do better than expected with Clint in the team, well, you can do what you want with that information, but it’s not ‘statistics’, it’s more or less a true thing (depending on whether you accept the methodology). Now, clearly it’s more complicated than that – Gera’s positional change being a key issue, Davies’ injuries, etc – but still it tells us something we didn’t know before.

          Real statistics is difficult, but Colin’s getting there and I suspect the information is better than people think. The problem is getting buy in. In the US a lot of what Bill James said was verifiably true in about 1977, but it still took Michael Lewis to write a best-seller in 2004 or so for people to wake up, and even then there was a lot of resistance until the teams using these methods were winning. And these were 100% provable facts! People just weren’t interested, their eyes being much better judges than “some geek in his mother’s basement”, etc. Ultimately though you want to get as much information as you can and try to keep an open mind.

          1. FWIW, I had Gera, Dempsey and Davies at almost identical win responsibilities, with less than a point (that’s league table points) difference between them.

            One interesting thing, though (and I haven’t fully implemented this in my stats yet), is set pieces. I’ve put together a +/- tally for the number of set pieces each player has won/lost, and an approximate number of team goals scored/allowed because of those. Dempsey came out very positive (almost two goals added), Davies slightly positive (~0.2 goals added), and Gera slightly negative (0.2 goals…um…lost? added against?). Dempsey was also positive in 08/09 (+0.5), while the other two were negative (around -0.5).

            Like I said, I’m not sure what the end result of all of this is yet, but it could potentially be an argument in favor of Clinton.

  3. This analysis is excellent, but why did you leave Duff out? His place in the side might be “safe,” but I’d be curious whether it should be! Likewise, it’d be interesting to see how he partners with any combination of Dempsey/Davies/Gera. But regardless, this makes for fascinating reading.

  4. Nice work/ I am afraid most of my feelings/perceptions are more subjective. Flat out, we are a better team with Dempsey on the pitch. He adds an element or surprise that is sorely lacking with our squad. Another issue about Fulham that gets little attention: We are old, relatively slow team that just does not seem to have youngsters emerging. Also the transfer gossip seems to have us running after more older players. This is not a good formula for the future.


  5. I think it might not be just about being a better team with/without a player starting. I think it might be Roy doing a bit of man management and a psychological thing.

    I think Clint get upset by being on the bench, so when he comes on or gets a start he plays like his life depends on it. When he gets a run of starts he goes a bit stale.

    For Gera, I think he is the opposite. When he does not get game time, he get nervous and stops believing himself (all very subjective from viewing in very bad seats and dodgey internet feeds.) Last season he got little game time and when he came on for short periods he froze with fear. This season he was given good game time, courtesy of ‘lesser’ Europa games, and started to lift the fear and believe in himself (and team mates believing in him) and started to give us the results.

    Don’t get me wrong, I am a maths/stats man, but correlation is not always causation.

  6. I like all three players. I love your numbers. I believe your numbers prove Dempsey’s desire and a burning/winning attitude.

    Don’t get me wrong. I am sure the others want to win.

    But did you ever choose up sides on a team? Certainly, you pick the most talented guy first, or the fastest. But how do you pick the next few players? Do you pick the guy who wants to win; who always wants to win in everything (in scrimmages, drills, races, small things)? That winning attitude spreads to the rest of the team.

    Dempsey is tough. He reminds me of Jerry Sloan (now Coach of the Utah Jazz) who was a 2 time all pro for Chicago Bulls (despite his marginal talent). Every NBA player hated playing against him, but every player wanted him on their team, and every team he played for had a winning record, regardless of talent (including his own).

    Does Dempsey’s attitude make the difference in the numbers you crunched?

  7. I am a really big baseball fan and no offense but I have to agree that numbers like this are a good reason why we need more numbers to crunch in football. Actually, this exact statistic has been used for my team (the Seattle Mariners) to highlight the supposed importance of a future Hall of Fame player who, frankly, sucked this year (Ken Griffey Jr.) and who was more or less forced to retire at the beginning of June. Right after he retired, the team went into a giant tailspin (not that they were doing particularly well before) which eventually led to the manager being fired.

    In this case, there actually is a correlation between Griffey’s retirement and the M’s performing badly, as Griffey apparently poisoned the clubhouse against the manager and everybody started to underperform. However, that is not a mark of Ken Griffey Jr. actually being a good player or worth 10 victories in the standings as some fans have tried to point out. That’s the real problem with this stat you bring up: teams win games, not players, and although player performance is integral to a victory, there are lots and lots of potentially spurious variables at play here.

    So it’s not just a sample size issue, it’s an issue of not controlling for when Dempsey is being used, for example. As a reserve, I can take a strong educated guess that he was used primarily in situations where his speed and heading ability would shine and his predilection towards walking the ball upfield was not too big of a deal. Of course he’s going to shine in those situations. The problem is that that doesn’t necessarily mean he’s going to help Fulham just as much or at all if used on the first team the majority of the time (which allows opponents to game-plan against him to a much greater degree than if they aren’t sure whether he’s going to be on the pitch or not).

    Anyway, going back to baseball for a second, it wasn’t as though Bill James came out in the late 70s, sent out a couple Baseball Abstracts, and then nothing else happened until Moneyball was published. The baseball stat community continued to move forward. One particularly important event was STATS, Inc. tracking hit locations. That led to, essentially, the creation of entirely new baseball statistics and an ability to quantify fielding in a way that we were never able to do before (apologies if this is over non-baseball fans’ heads; I am coming to a point here). That’s actually something that fans of the EPL and the League should start doing: compiling their own stats for the games, stuff way beyond goals, assists, and fouls, and use that to get an economic appreciation of the game.

    If you pay me just 50,000 pounds a year, I will do it for you! Do I hear 40,000? 30k? A satellite dish and a subscription to Setanta?

    1. “As a reserve, I can take a strong educated guess that he was used primarily in situations where his speed and heading ability would shine and his predilection towards walking the ball upfield was not too big of a deal. Of course he’s going to shine in those situations”

      But that bears no relation to how he was used at all – he invariably gets long runs in the team until he’s injured, then long breaks until he’s not. I strongly believe that, while the above digging *proves* nothing, it does outline when the team has surpassed itself and when not, and over a number of games there are clearly patterns. Your baseball example is very different, not least because I’m not looking at WLD, I’m looking at performance vs expected performance.

      Colin is tracking all you suggest for football and more, don’t worry (much of it covered in the Fulham Review)! The Times Fink Tank (Castrol Index) is working in roughly the same way. The point is that while James/Palmer et al were on the scene early, and John Dewan’s firm did make progress, they were thoroughly unable to make use of this beyond the odd consulting job. Sandy Alderson took note and took advantage in the 90s, but otherwise the whole body of work was ignored by 99.9% of those who could have cleaned up if they’d opened their minds, which is staggering when you consider how powerful their information was. It’ll never be like that for football as the stats aren’t nearly so strong/obvious (team game), but still there is something to be gained by having a little dig around.

      1. You make some excellent points. The one thing I do want to go in on is that basketball is in some ways almost as much a team game as football but in its Americanization it’s statted out to hell as well. I guess the biggest difference there is that the teams that have used a lot of sabermetrics-based analysis haven’t been able to Moneyball things (Seattle in the mid-2000s used a guy who wrote a book similar to the old Baseball Abstracts, only for hoops – that’s the only team that comes to mind off the top of my head). That may be in part because the stat-tracking still isn’t sophisticated enough and as such the numbers that are produced don’t tell as much as they could.

        I’m glad to hear that people are working on advanced stats in football. That’s probably the one thing keeping us Americans from really getting into the sport. ;)

      2. Sorry for the double reply, but noting that Dempsey is used in spurts and then kept out for spurts is interesting, too, in its own right and another troublesome reason why partial performance may not be indicative of future, full-time results. I don’t think it’s *quite* so open and shut, as ISTR Hodgson defining Dempsey’s role as a 2nd-teamer (although especially with Europa, the reserves got to play quite a bit). Still, stamina is a potential issue; one wonders, for instance, how he’d be under a Mark O’Neill.

        My overall point, anyway, wasn’t that you had to downgrade Dempsey’s performance for matchup-based reasons, it’s that the fact of not playing a full season means you have variables you need to account for above and beyond sample size. Breaking it down as to whether Fulham overperformed or underperformed helps to take out some of the strength of schedule issues, but one wonders, too, if Dempsey’s team performance might have been helped out by him playing a larger portion of his games against teams who get frustrated by Hodgson’s shell and take bad risks. I don’t know the answer to that. Home/road is also partially covered by the numbers that were used but Fulham as a whole did seem to be more impressive on the road (the month of January notwithstanding). With baseball and basketball there are so many freaking games that things tend to even out even when a guy misses a quarter of the season. I don’t know that that’s so much the case with football.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your 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