Soccer metrics to evaluate players

I haven't found any such statistics for the MLS, probably something that companies sell. On a related sidenote though, there was an interview with Sigi Schmid on World Football Daily a while back, where he talked about having his assistant manager track such statistics on a computer during the game. He then uses that information to figure out whether to sub a player off or not if they're not getting enough touches, or has a bad passing percentage on the day. I wonder how widespread such an approach is in the MLS. Yallop doesn't strike me as a coach who is that interested in such things (more of an old school guy I think). Not saying whether that is a good thing or bad thing, by the way, just thought it was interesting.

 

Sigismund Schmid? Who woulda thunk it?

You're very right though, I doubt that the old school guys (and not just Frank) are going to take well to some newfangled method of directing practices and games. We have heard any number of times (even on these boards) that soccer does not produce the large amount of statistical data, as baseball does, for making the calculations necessary. That's one way of rejecting the approach.

 

You should check out Match Analysis, which sells statistics and near real-time delivery of them to teams and the media.
>> http://www.matchanalysis.com
>> Match Analysis records and verifies over 2400 distinct events
>> per match. Using proprietary technology, every touch by every
>> player is catalogued precisely, synchronized against video feeds,
>> and stored in an expansive video database of international football.

They started in 2000 and claim to be the market leader in the US.
>> http://www.matchanalysis.com/news/article24.jsp
>> Match Analysis signs with 13 of 14 MLS clubs (Apr 15, 2008)

Interestingly, they are headquarterd in Emeryville ...

 

I believe the creator/founder of that company used to post on the San Jose boards back around 2000 or 2001.

 

Well, the Quakes are listed as a client, as is most of the rest of the MLS. Of course, how each team uses such statistics probably varies quite a bit. In the end, whether or not a coach uses such technology doesn't determine success. I personally don't think new school is better than old school or vice versa. Either can work depending on other factors.

 

Match Analysis rocks! These guys are great. They posted some free statistical stuff when they first started. They count touches, passes completed, passes not completed. Forward passes, back passes, and so on.

-- From a previous post --

1. Number of touches

2. Shot creation

3. Ability to retain the ball

4. Balls won per 90 minutes
---------------------------------

We are missing:

A. Forward or lateral passes versus backpasses. Maybe we want to count these as passes that advance the ball upfield. Sometimes, a player will pass backwards to a teammate who is in a better position to make a quality attacking pass.

B. Headers won. This has to be not just getting your head to the ball, but either controlling it, passing it to a teammate, or as a separate category, a head clearance that breaks up an opposing attack.

What do you think? We should kick these ideas around until we get something a bit more useful.

Go Quakes!

- Mark

 

Just because a team subscribes to this service does not mean they will improve. It all depends on how the team uses the stats. You know there is an old SW programming adage that "you cannot improve what you do not measure." But I also subscribe to the adage about "lies, damn lies and statistics." I would be interested if some of the stats had anything to do with the loss of any Quake players this last year...or if some of the stats revealed why the Quakes kept certain players ( Convey, Corrales, Elliot, Ribiero )? Did those stats influence decisions on playing time ( Zaher )? If anyone gets to ask Yallop or Doyle about the role of stats on the past year, it would be interesting if they would open up the dialog.

 

You bring up some interesting points. Some of the guys we kept and started over, and over, and over again -- Corrales, Convey, Elliott -- are almost inexplicable. I think cutting Ribeiro would have been an admission of abject idiocy on their parts. So they kept him rather than have to wipe egg off their faces. But they didn't play him much. Antonio turned in some of his best performances for us and then got benched. (Not that he was objectively good mind you, just that he was not as horrible as Corrales or Elliott.) It might be that he got pissed off when he was pulled from the lineup, or it might just be that Frank is an idiot. Ribeiro played better and worked with his teammates better than any of Corrales, Elliott, or Convey. Yeesh! Convey gives the ball away so often and so cheaply it's shocking that he got any minutes after about the eighth or ninth game. Corrales at least had some nice moments here and there, but mostly he was slow, afraid to tackle, and often out of position. Elliott was terrible all season until the last two matches where he improved to mediocre.

But how would stats have influenced Frank and JD's thinking? And we're talking almost exclusively about Frank here. Convey's teammates refused to pass him the ball the first several matches. If Frank were using this system, he'd have benched Convey. But he didn't. He played Convey unless Bobby was injured. Same for Corrales and Elliott.

Frank only played Ribeiro when he had next to no other options. Once we got Sanchez and ALM Antonio hardly played (until ALM got hurt).

Campos got traded after Hucks got pissed off at him. It might be that other players were miffed at Pablo too, or that Pablo did something he was told not to, or did not do something that Frank required. It's hard to say, but we all saw Hucks yelling at Pablo and then Pablo got traded a couple of days later.

Weaver had already played himself out of the starting lineup when he got traded. Clearly, Wondo works better with the rest of our guys than Cam did. That was a good trade for us. But I don't know if Weaver being benched goes to statistics.

The big problem with our team early on was that Corrales and Convey (then Corrales and Elliott) sucked in central midfield and Frank kept changing the forwards as though that were the problem. I'd have benched Corrales after the third match and tried McD at D-mid. Elliott didn't even play until the third match when he came on as a sub, so I can't complain about that. Still, Convey, Corrales, and Elliott were all awful their first few matches. I'd have benched them and given them another chance later. Both on the theory that you should sit if you suck, and that sitting a match or two gives you a chance to find your mojo, rather than continuing to practice failure.

And we never tried Somma at attacking mid. We were grooming him to play attacking mid, but we played him at forward for four minutes and then cut him?! Clearly, Somma wasn't the reason that we sucked. Corrales, Garcia, Elliott, and Convey were the reasons that we sucked. (At least, the biggest reasons. Hucks and Hernandez being injured hurt too.) And maybe Somma would have sucked at a-mid. He might well have been worse than Corrales, but we should have tried it, and if we had been following these statistics, then we would have. I'd have tried Arturo at attacking mid, and Wondo too once we got him. I certainly would not have kept repeating the same mistakes over and over again.

As for Zaher getting benched, even Frank admitted that Z was playing well. The problem was that Frank refused to bench Convey or Corrales, so instead he benched a guy that was playing well and needed the minutes to get better so he could be our future left back for years to come. Benching Z was massively stupid in every way. Especially given that Convey did not work well with his other teammates, but did work well with Z. We should have played Convey at left mid once Hucks was out and we should have benched Corrales instead of playing him at left back. However, Frank is loath to admit to any mistakes.

The worst part about playing guys who turned in crappy performances over and over is that it demoralizes every other player on the team. If "those guys" can suck but keep their starting position, it's pretty hard to be inspired. If you're behind one of those guys, nothing you can do is going to get you more playing time. You have to hope for injury because otherwise you're on the bench.

We might subscribe to the Match Analysis statistics, but we sure aren't using them. We'd have benched Convey for certain if we were. Corrales and Elliott too. Joe would have been benched earlier in the season. Garcia would have gotten benched. Instead he played every available minute until we traded him.

No. Frank is making decisions based on his gut, his intuition, or some oracle. He's surely not relying on a serious analysis of what's wrong with the team.

go quakes!
we suck!

- Mark

 

The only time I've heard a comment from Doyle / Yallop about these kinds of less conventional stats for player evaluation was in reference to Ivan Guerrero. I think it was something to do with average time per possession (i.e. when he has the ball, how long does he hold onto it). Or maybe it was just possession time (i.e. how long does he have possession per minute played).

Of course they wound up trading Guerrero and I'm not sure he's even in MLS anymore. I thought he played pretty well for the Quakes FWIW.

 

I loved looking at the stats from the Confed Cup this past summer. In particular, the distance and speed metrics seemed helpful as a gauge of fitness and (probably) defensive commitment:

http://www.fifa.com/confederationscup/statistics/players/distanceandspeed.html

One could use this over the course of a season to determine when a high-energy player could use a break in a cup game.

The passing section also quantified Spain's insane ability to maintain possession.

 

Totally agree - I thought Guerrero fit well with the Quakes, and was surprised to see him go.

 

Yeah, I see this analysis as a couple different phases:

1) Measure, and measure correctly (I was shocked at how different some of the stats were from the FIFA site and the ProZone on USsoccer.com)
2) Determine what constitutes success
3) If a player isn't achieving success, do you place him in a different role? Or do you have a better option for that same position?

It seems like the Quakes may encounter a situation where they know a player isn't doing as well as they'd like, but don't have a better option.

 

Re: Soccermetrics to evaluate players

Please report your findings. Seems like it would be rather expensive to invest in programs like those and not use them to the full extent. Of course, the coaches might think they're experienced enough not to rely on tools like that (you can see Mark's influence creeping in on my thinking).

 

Re: Soccer metrics to evaluate players

Great thread.

I imagine coaches and managers will have to tailor the metrics to their own particular style. After all, soccer is a game of choices. A player might complete a long pass, or dribble three guys, but these may have been poor decisions based on the manager's strategy.

All this is to say it's not as cut and dry as baseball metrics.

One of my favorite measures (which I compile magically in my head) is how comparatively successful a team is with certain players on the field -- regardless of individual performance. (Eg: "For some reason, the Quakes have a better goals-scored-vs-goals-allowed ratio when Player X is on the field than with Player Y.") It's an imperfect measure, but it takes into account the vagaries of the game.

 

Re: Soccermetrics to evaluate players

Yes, it would be very interesting to know.

As for my influence growing on you like a fungus, these guys are pros, they were players. (JD, Frank, Ian, Ronnie, Jason) So they each have enough ego for a couple of petty dictators. You've got to have a ton of self-confidence to go out on the field in front of thousands of critical spectators and compete against players who you know to be good. If Frank or JD ever see or hear any comments like what is on display here, they no doubt think: "Tosser has no f@#$ing idea what he's talking about."

But it might also be an issue of time. They do have a lot on their plates.

And this is a technology that is still in it's infancy. So it's easy for a coach or GM to argue that it is unproven. Still, I think we can say with some certainty, that given that in Bobby's first several matches with us, many of his teammates would not pass him the ball, Frank does not make decisions based on this software. The thing that really shocks me, is that Frank didn't chew out his other players and say: "Damnit! Give Bobby the ball!"

Go Quakes!!
smelling like month old laundry!

- Mark

 

Re: Soccer metrics to evaluate players

Definitely true, though I still believe that there's a lot of potential in the application of these types of methods to soccer. Soccer is just a much harder problem than baseball.

One of the things that Beane and company did with baseball is to actually build a mathematical model that allowed them to predict the average number of runs scored per game based on certain factors like on-base-percentage. You can plug in data from a number of different teams / seasons and verify that the model works. If you can do that, then you know what kinds of things to value. For example, if on-base-percentage is one of the positive factors in the "runs per game" model, then optimize for that. That's one of the reasons the A's were able to acquire under-valued players - e.g. players who had maybe mediocre batting averages but good on-base-percentage, etc.

So without a mathematical model for soccer, how do you even know what measurements are important? There are some things that would seem to be important on the basis of common sense, but w/o a model you are kind of just guessing. I wonder if anyone's got a model for soccer that works to any appreciable degree. If they do, I don't think they'd want to share it with anybody. I'll have to pick up a copy of "Soccernomics" to see if there's anything about this "modeling" business.

Although soccer is a much harder sport to model and measure than baseball, I think that it's gotta be doable. It's just hard, like mapping the human genome or sending a man to the moon. Hard, but doable. There are billions of dollars at stake in the "soccer economy", so I think the incentive is there as well.

 

Re: Soccer metrics to evaluate players

I like it, as well as the other two. This should be an indicator of the ability to begin attack from the back.

 

Re: Soccer metrics to evaluate players

MJ, and excellent list!

On B, dispossessions, you're talking about our guys taking the ball away. What about the enemy kicking the ball over the end line or touchline? Would you count that as a favorable outcome? or just not count that run because the enemy blew it of their own accord?

I think we need to add passes completed to teammates, and also forward or lateral passes complete to teammates. Some of our guys would "pad" their statistics with tons of back passes.

This is good progress!

Go Quakes!!
like overripe fruit!

- Mark

 

Re: Soccer metrics to evaluate players

Hah, if it were only so simple. Also, in terms of what we're trying to model (a formula for winning a soccer match) # of assists and time of possession are not outcomes, they are additional metrics.

In baseball, you have set times when a team is on defense or on offense, so you can model just for # of runs scored, for example, and it's meaningful at least in getting half of the equation. In soccer, the line between offense and defense is blurred. Just creating a model / formula for goals scored isn't all that meaningful because you may be optimizing for offense at the expense of defense and building a team that loses 5-3.

I haven't thought about this a lot, but since you can only score when you have possession, and the other team can't score when they don't have possession, is there a simple correlation between time of possession and success in a soccer match? Obviously there's more to it than that, but that might be a place to start.

If time of possession turns out to be directly related to success, than you can start to evaluate players on the basis of their ability to hold possession for the team - e.g. how often they lose the ball on the dribble, success rate of their passes, etc.