Sabermetrics applying to Soccer

[kenntomasch]

Tom Eaton gets some pub (http://www.mlsnet.com/content/03/analyze0808.html).

[beineke]

Originally posted by JG
The Cunningham one obviously shouldn't count...the others are subjective.

Of the 19 remaining goals, only 5 are headers.

If we're asking what to do on a corner kick, I think that the other four are important cases to include. Based on your descriptions, it does sound as if the plays evolved from a corner.

My impression is that the great majority of corner kicks are attempts to cross the ball for a header on goal. But even with this small amount of data, that doesn't seem to be a very productive approach. Almost certainly, the success rate is higher for other options ... short corners, or crosses to knockdowns, dummies, flick-ons, or volleys.

Thanks for doing the follow-up on this.

[mpruitt]

Since I haven't really posted anything substanitive in terms of a hypothesis to throw out there.

I believe that the biggest difficency in terms o the stat tracking of soccer games is the Percentage of Possesion statistic. As it has been mentioned before possesion does not always corrilate to gaining points, or goals. I think that a far more accurate way would be to track some kind of team turnovers. Who really cares if a team is trying to bunker down, holding the ball and passing the ball along its back four. I know that this might be a little bit more difficult to track than say a turnover in basketball, but based on a team level I think that it could give a lot of insight.

Perhaps it would be even more difficult and pointless to track for an individual player because people taking on guys one on one is rather rare, but I would be wiling to bet that teams, no matter their style would have a better won/loss/tie record if they had a Plus Dispossesion record than a Minus one.

Give the flukey break neck pace of soccer, obiously one wrong move can change a game. But no one's really keeping statistics as to how many wrong moves teams are making. Statistically speaking, I'd be that some of the best teams in the world would have this ration be very high, rather than 10 David Regis' on the feild.

Edit: PS, nice job by Tom really forcing the issue with Peter Hirdt.

[beineke]

One more corner kick goal this week ... Magee sailed it over the pack, and Jolley went backdoor to send a volley into the net.

Yet again, a corner goal is not a header, and yet again, Chicago is the victim.

[JG]

Originally posted by beineke
By the way, if you asked a hypothetical stats grad student about this, he might consider a Poisson model instead of the simpler Pythagorean Formula.

You're probably right about that...I tested on the last five years of premiership results and the mean error in "point percentage" using a poisson model was 3.1%, which is better than I could get using variants of pythagorean and goal ratio formulas. It seems to do a better job with the teams at the top of the league.

I stumbled across a pop-scienceish probability book from the UK today...they had a couple pages about soccer...talked about using poisson distributions to predict results as above, also mentioned a study by a British newspaper showing that the team scoring first in a premiership match won 69% of the time, plus a dutch study about how teams perform playing 10 vs. 11, which could presumably be used to determine when a defender is better off commiting a red-card foul than allowing a sure goal.

[mpruitt]

http://www.bigsoccer.com/forum/showthread.php?s=&threadid=64954

still would really like to see a better venue to discuss some of these things in depth. please post in that thread if you agree.

[KenC]

Sorry, only had a chance to glance at this thread, but why wouldn't you try to break down possession to possession in the attacking half to eliminate the odd results, and/or look at possession in the box.

[beineke]

Originally posted by JG
You're probably right about that...I tested on the last five years of premiership results and the mean error in "point percentage" using a poisson model was 3.1%, which is better than I could get using variants of pythagorean and goal ratio formulas. It seems to do a better job with the teams at the top of the league.


I've seen the same kind of thing when I looked at some 20-year-old NFL data. By assuming a Poisson distribution for scoring, you could model things pretty well.

This kind of result -- obtaining good information from an overly simplistic model -- is why I've become more optimistic about the feasibility of soccermetrics.

[mpruitt]

Originally posted by beineke
soccermetrics.

And doth! A name was born.

[RichardL]

I'm sure it could be useful but as I'm sure has already been mentioned somewhere in the previous 8 pages, it's knowing what measurements would be useful and also crucially, knowing how to interpret them. I'm sure many are aware that an old dinosaur at the FA many years ago studied hundreds of matches and found that most goals come from moves of three passes or less, and deduced from that that the long ball game, with it's emphasis on the minimum of passes, was the way to go. It would be far too easy to look at a dry stat like passing completion rates for two central midfield players and assume the one with the higher percentage is better - of course the one with the lower stats could be a creative midfielder trying to pick out players in the box, while the other could be a ball-winning fetcher & carrier type player who only plays safe passes. It would take a staggering amount of breaking down of stats to ensure that only like for like comparisons are being made.

[mpruitt]

Originally posted by maxim-1
http://www.bigsoccer.com/forum/showthread.php?s=&threadid=64954

still would really like to see a better venue to discuss some of these things in depth. please post in that thread if you agree.

I really think if to be productive at these types of discussions or to maybe gain some greater insight it would be best to do that in our own forum. please pose on this link and maybe we can make that happen.

[beineke]

Originally posted by RichardL
It would take a staggering amount of breaking down of stats to ensure that only like for like comparisons are being made.

This won't ensure only like-for-like comparisons, but you could do pretty well by tracking a few simple things:
Location of a touch (e.g. attacking third)
Type of pass (e.g. crosses)
Direction of a pass (e.g. square)

This doesn't account for defensive pressure and offensive support, but it does provide a way to filter out "creativity." Lower-percentage passes are only a good move when you're pushing the ball forward or when you're in the attacking third of the field. By excluding those cases, you should be able to make a pretty accurate comparison of midfielders' ability to guard possession.

[microbrew]

Originally posted by JG
I stumbled across a pop-scienceish probability book from the UK today...they had a couple pages about soccer...talked about using poisson distributions to predict results as above, also mentioned a study by a British newspaper showing that the team scoring first in a premiership match won 69% of the time, plus a dutch study about how teams perform playing 10 vs. 11, which could presumably be used to determine when a defender is better off commiting a red-card foul than allowing a sure goal.

What's the book's name? Probably what would be more interesting is to get the bibliography of the soccer section and see the sources. There seems to be a lot of work already done, and I hate to reinvent the wheel everytime.

The dilemma of a red-card vs. a sure goal is one aspect what this discussion is about: making decisions based on more than just instinct, intuition and/or "common sense".

[mpruitt]

bump. this is an important thread with important ideas and should be read and continued

[microbrew]

Chances for getting a forum don't look good...

http://www.bigsoccer.com/forum/showthread.php?
s=&threadid=64954

I think this topic is just too esoteric, and the all the math speak is off-putting or just over many people's head.

There must be someplace- a mailing list or something on this topic somwhere. If not, Bigsoccer can be a pioneer. Build it, and maybe they'll come...

[Real Ray]

I think perhaps the place to put this topic in is the Coaches forum. I think you'd get a good input from people re: their perceptions/view on what some numbers show.

For instance, I've started looking the idea of height in midfield and the role it plays not just winning and possession but the more subjective "level of play" and styles. And comparing it to MLS.

Looking at Arsenal for instance Vieria is 6'3" Ljunberg 5'9", Parlour 5'10", and Pires 6ft.

Liverpool w/ Gerrard, Riise, and Hamaan, have three 6ft and over, with Danny Murphy at 5'8".

Man Utd. is small by comparison, as Giggs, Scholes, Keane, and Butt are all under 6ft.

But they all that one big guy up front: Henry, Heskey and Van Nistelrooy at 6'2".

Looking at the Fire and Metros as examples:

Lisi 6-1, Williams 5-5, Clark 5-10 and Guevara at 5-11. Up front they have no one over 6ft-Mathis and Moreno at 5-10 and Magee at 5-9.

The Fire have Armas at 5-7, Whitfield 5-11, Mapp 5-10 and Marsch at 5-11. Razov at 6-1 and Ralph at 6-0 are on top.

At the bottom of our tables is Dallas with:

Vaca 5-6, Pareja 5-8, Stone 5-9, and Davis 5-11.

Up front: Cerritos 5-9 and Rhine 5-10. Hmm...

[mpruitt]

yeah it's disapointing people haven't have seen fit to post for a new forum about soccermetrics. i'm going to do a little more reasearch as to what rssf.org or whatever the heck was about. microbrew, i really liked your ideas about what that forum would represent, especially because i'm interested in this stuff but i'm a dope when it comes to math.

i just think it's disapointing because it appeared to be a general topic that people were pretty interested in then this thread kind of died because of the disorganization that comes with having a topic like this just in one thread.

i presonally would be a big fan of continuing to press for a new forum or sub forum on here. i just prefere the set up of a message board to that of a listserver and it'd be nice to start something like that on these boards.

maybe this would have better luck in the coaches forum, but if we were looking to do non-traditional "outside the box" stuff i'm not sure if that would be the best place for it.

[microbrew]

The below paper is where the author looks at the scoring of goals in the World cup with a poisson distribution, using a data set of 232 games.

Chu, S. (2003), "Motivating the Poisson Process Using Goals in Soccer." INFORMS Transactions on Education, Vol. 3, No. 2, http://ite.informs.org/Vol3No2/Chu/

[JG]

Originally posted by microbrew
What's the book's name? Probably what would be more interesting is to get the bibliography of the soccer section and see the sources

http://www.amazon.com/exec/obidos/ASIN/0198526636/qid=1061499505/sr=2-1/ref=sr_2_1/002-8819408-2287220

If I have a chance I'll get the book from the library and check out the bibliography.

I'm pretty sure that the red card article is:

G. Ridder, J.S. Cramer, P. Hopstaken, Journal of the American Statistical Society, 1994, 89, 1124-27.

For people with access, the link is:

http://links.jstor.org/sici?sici=0162-1459%28199409%2989%3A427%3C1124%3ADTTETE%3E2.0.CO%3B2-R

[mpruitt]

a little cross polination posting
This speaks to the question of what sports statistics should be. Are they to be only an objective record of what occurred and nothing more? Or should they be made as “smart” as possible, to be used as much for analytical as for historical purposes?

From Mr. Elias' latest column. (http://www.mlsnet.com/content/03/analyze0822.html) I wonder if he's been sneeking peeks at this board?

On a different note. Great find by someof you guys in regards to some stuff that's already been written on this subject. However, as mircobrew suggested for possiable topics of a forum relating to this stuff, probability and statistics for math dummies threads would be a massive help to people like me. Because I'm an absolute moron when it comes to numbers and math when I see Poisson distrabutions I think of being back stage with Bret Micheals.