Sabermetrics applying to Soccer

[kenntomasch]

They should just lift weights.

[1953 4-2-4]

I
--Athleticism is beginning to trump skill – I ‘d love to see the 40 times of our national team players, vs. those of, say, England or France. Or Vox2 measurements. Speed, flexibility, fitness, these things are key in the modern game.

--Winning teams have great defensive organization. Rarely out of shape, always strong pressure cover, giving teams few chances at goal, win balls all over the field, allow few shots.

--Good teams win head balls in the penalty area on both ends of the field.

--Good teams have quality touches inside the penalty area, and prevent other teams from having quality touches in the penalty area.

--Good teams “hold” the ball – by that I don’t mean dominating possession. It’s that when you HAVE it, you do something productive WITH it.

--Good teams cause turnovers, but don’t commit them. Especially in the middle and final third.

--Teams that switch the point of attack and use the entire field when in possession are better than teams that are “sided” in their ball possession.

--Teams that can complete a higher percentage of on-the-ground passes than the other team (not quantity, but percentage) will do better on average.

--Teams that on average complete longer on-the-ground passes than the opposition do better.

--Good teams convert a high percentage of their numbers-up scoring opportunities. Weak teams have fewer, and don’t convert them at the percentage that better teams do.

--Good teams have coaches who exploit advantageous player matchup opportunities, and are adept at hiding playing matchup weaknesses.

Wow, your ignorance is staggering. Tell Brazil that it's bad to be sided in attacking (RCarlos),or England in the 60's with Moore, or shun heading (Ronaldo & Ronaldinho), or complete fewer long passes, then splice in things like "teams who have 'numbers up' situations win." LOL, do you type that with a straight face? 40 TIMES? Here's a note: they don't measure those at a pro team. They do endurance tests, as well as agility and quickness tests. Ponder that. Sorry, they actually don't test the 40 & bench press in this kind of football.

[kenntomasch]

Yawn.

I was just busting your balls, asshat.

[1953 4-2-4]

Yawn.

I was just busting your balls, asshat.

Huh? Unless Karl Keller is your clueless alter ego, who's talking to you?

You know, perhaps this website can be as unhealthy as crack when you post so much...

[Real Ray]

There are SO many important things in soccer that you can not quantify--ever. Count "how quickly does X player get goalside from the time your team loses the ball? Did he help offensively on the previous play? Is he in position now because he is afraid to ever get into atttack? IS he marking X player on every play, or is he switching with a teammate?" Just one thing, but just an example. The game is too complex to derive any meaningful information other than the very basic of stats. And if you are a coach, and have to look at a chart to have a feel for which of your players has a high passing %, then you shouldn't be a coach.

Really...perhaps you should speak with Arsene Wenger and host of others who..well have moved into the 21st century-lovely irony to be found in your Big Soccer ID, "1953..."
The Easter Road outfit are in discussion with a firm called Pro-Zone about installing their analytical system, favoured by many of England’s leading clubs including Arsenal and Manchester United.

Using cameras in 12 fixed positions, the company will be able to provide Mowbray and his backroom staff with a complete breakdown of every player’s contribution throughout the 90 minutes on match day - and even warn him when one of his stars is in danger of picking up an injury.

While the complete system may be too expensive for Hibs - it costs in the region of £130,000 a year - Mowbray disclosed it may be possible for the Leith club to install one custom-built for themselves.

He said: "The different camera angles cover the whole pitch and everything is downloaded onto one DVD and sent to their headquarters in Leeds.

"It reveals how far a player has run, when he runs, how fast he runs, the number of tackles he makes, how many passes and whether they are forward, back or sideways, the shots he tries and crosses he puts in."

Huh...imagine that...somewhere there is a dvd in London Colney, with a file that can tell us, "how quickly does X player get goalside from the time your team loses the ball?"


http://edinburghnews.scotsman.com/sport.cfm?id=905992004

[Eliezar]

Really...perhaps you should speak with Arsene Wenger and host of others who..well have moved into the 21st century-lovely irony to be found in your Big Soccer ID, "1953..."

Huh...imagine that...somewhere there is a dvd in London Colney, with a file that can tell us, "how quickly does X player get goalside from the time your team loses the ball?"


This is so widely publicized you would have to be in a time warp not to notice it.

Rooney's workrate has been majorly touted. The fact that when he was playing for Sunderland, Claudio Reyna had the highest completed pass percent, highest number of completed passes, and most tackles won was interesting. I'm sure the actual coaches had more in detail numbers than what they posted on the website.

1953 is just someone who pretends who knows a lot and claims to develop 'Pro's' all the while just being someone who has never made an informed post on big soccer that I have read.

[1953 4-2-4]

Really...perhaps you should speak with Arsene Wenger and host of others who..well have moved into the 21st century-lovely irony to be found in your Big Soccer ID, "1953..."


Huh...imagine that...somewhere there is a dvd in London Colney, with a file that can tell us, "how quickly does X player get goalside from the time your team loses the ball?"


http://edinburghnews.scotsman.com/sport.cfm?id=905992004

Damn. Impressive. Ok, this is productive, I take back what I said. How's that for being close-minded? ;)

[Karl K]

Damn. Impressive. Ok, this is productive, I take back what I said. How's that for being close-minded? ;)

Oh, and they lift weights at Arsenal too.

Maybe that mind will open a crack more....not.

[Serie Zed]

Jesus, what sort of tool trolls in the statistics forum?

But never mind that...I've just finished Moneyball and was thinking about this thread as I did. btw...congrats to Voros (I'm assuming) for the long section in the book about his findings.

At the great risk of being over simplistic with a group that knows more about this than I do...

One of the (many) reasons Sabremetrics baseball worked well for baseball is that it was relatively simple to objectively chart the result of a given play (at least if you had the time to do it). Player V hit the ball with a velocity of X and trajectory of Y to location Z -- which lets you chart the average result and etc. And those results (offensively) are completely independent of what any other (offensive) player does. Soccer doesn't work that way. Both offense and defense are dependent on what other players are doing at the same time as the players at the center of the action.

I know all that's obvious, but to me it suggests that any sort of charting has to take into account the relative positions and movements of multiple players at any given time. Which is going to be nearly impossible watching on TV. So any serious effort to do this would require an overhead view. You just aren't going to find what you're looking for any other way.

I'm also wondering if a big piece of the goal here (so to speak) is creating metric that judge how well an offense creates time and space in dangerous positions? Beating a player off the dribble, making a great run off the ball, linking three or four one-touch passes together or a single pinpoint pass over distance all work because they create time and space.

So wouldn't it be possible (if you were watching from above) to chart the results of a possession in terms of what it was that generated time and space for the ball? To measure which actions resulted in objective time and space.

The corner kick is a great example. It's dangerous because it's essentially unlimited time and quite a bit of space.

That's not the only measure of course -- there's something in there about finishing too, but that's somewhat easier to measure, especially if you could see all 22 players at any given time.

And if you were really clever (I'm not) couldn't you plug all those dynamic run-of-play metrics into some sort neural network program and tell it to solve for what creates goals?

[fuschia]

I keep reading that it's very difficult to make statistics for soccer. How about starting with simpler things and refining over time? e.g. goals in first halves, then goals per 5-minute period, then goals after a pass, etc.

As a coach, I am interested in that sort of information at least to confirm that my intuition is correct. For example, knowing whether a group of defenders is better than another one. I can pick a line up I believe is the best then compare what actually happened to see if I was correct over a few games.

I look at statistics as a mean to validate or refute my beliefs or even as a place to ask questions. If a group of players allow more goals than another, I must ask "why". Maybe I need to change the training. I just don't have the time or inclination to collect all that detailed info. However, If some one else can do it, that's a great help.

[Carrot]

Maybe this is totally oversimplifying things, but why not start out with a +/- rating for players, as with hockey? There are, of course, some limtiations to this statistic; most obvious is the fact that good players on bad teams will have low ratings and marginal players on good teams will have good ratings. However, I seem to recall that they've developed a corrected +/- rating that takes this into account.

The good things about this system:
1. It's simple, one stat, and while not perfect, gets at the main question: is your team better when this player is on the field?
2. It doesn't get into possesion, shots on goal, number of dumps taken at halftime, or any other nmber of stats with unknwon utility--it measures whether your team scores more or scores less when the guy is in there--any coach would be interested in that.
3. It does (in its own way) work much better than purely offensive stats for midfielders/forwards (if the guy scores a lot but also sits up top all game, never comes back to defend, and his team gets scored on a lot because of that, the rating will show it) and better than pure defensive stats for defenders (if a guy is only a fair on the ball defender, but moves up well, sends in a lot of dangerous crosses leading to goals, and helps his team in that way, the rating will show it)
4. You can also use this stat to look at lineup differences and see how well the team does when certain combinations of players are in (player X + Y= team does well, player X + Z=team sucks).

Just one man's opinion. Fire away.

[ChrisE]

Maybe this is totally oversimplifying things, but why not start out with a +/- rating for players, as with hockey? There are, of course, some limtiations to this statistic; most obvious is the fact that good players on bad teams will have low ratings and marginal players on good teams will have good ratings. However, I seem to recall that they've developed a corrected +/- rating that takes this into account.

The good things about this system:
1. It's simple, one stat, and while not perfect, gets at the main question: is your team better when this player is on the field?
2. It doesn't get into possesion, shots on goal, number of dumps taken at halftime, or any other nmber of stats with unknwon utility--it measures whether your team scores more or scores less when the guy is in there--any coach would be interested in that.
3. It does (in its own way) work much better than purely offensive stats for midfielders/forwards (if the guy scores a lot but also sits up top all game, never comes back to defend, and his team gets scored on a lot because of that, the rating will show it) and better than pure defensive stats for defenders (if a guy is only a fair on the ball defender, but moves up well, sends in a lot of dangerous crosses leading to goals, and helps his team in that way, the rating will show it)
4. You can also use this stat to look at lineup differences and see how well the team does when certain combinations of players are in (player X + Y= team does well, player X + Z=team sucks).

Just one man's opinion. Fire away.

http://www.bigsoccer.com/forum/showthread.php?t=107927&page=3&pp=15

[Carrot]

Thanks. Looks like I'm about 10 months behind the curve.

[scu84]

Folks,

I'm having a philosophical coaching difference with a fellow coach -- and I am attempting to prove my point statistically. I have just spent the last 3 hours looking to back up my 'thesis' (with no luck), and hoping you can help.

I am attempting to get our team to take more shots... thus I am very encouraging that ANY opportunity to shoot within the box (given a reasonable angle) should be taken (ie. behave selfishly). My fellow coach is stressing looking for a more open, or better positioned teammate -- and is beginning to influence players to look for a better option first, then think about taking a shot.

I would like to prove (what I think I know from playing) that "good things" happen when shots are taken (deflections, second opportunities, stretches defense, and of course goals) and show that the correlation between the number of shots taken and goals scored is high. Or show that teams with a higher average number of shots attempted per game, win more.

Any ideas on how to get this data to back-up my thesis? I would take data for any league (MLS, EPL, Serie A, Bundesliga, etc.).

Thanks!

[EvanJ]

http://www.mlsnet.com/MLS/stats/index.jsp?club=mls
Shots, shots on goal, goals, and points in the standings for the 12 MLS teams so far this year:
Chicago Fire: 301 shots (144 on goal), 41 goals, 39 points
CD Chivas USA (expansion team): 265 shots (102 on goal), 27 goals, 14 points
Colorado Rapids: 236 shots (103 on goal), 29 goals, 28 points
Columbus Crew: 280 shots (97 on goal), 21 goals, 27 points
FC Dallas: 329 shots (143 on goal), 39 goals, 36 points
D.C. United: 307 shots (137 on goal), 39 goals, 38 points
Kansas City Wizards: 322 shots (141 on goal), 43 goals, 41 points
Los Angeles Galaxy: 317 shots (139 on goal), 31 goals, 35 points
MetroStars: 307 shots (135 on goal), 38 goals, 33 points
New England Revolution: 283 shots (150 on goal), 42 goals, 45 points
Real Salt Lake (expansion team): 245 shots (107 on goal), 23 goals, 19 points
San Jose Earthquakes: 264 shots (126 on goal), 36 goals, 44 points
Every team has played 23, 24, or 25 games.
Only the New England Revolution have gotten a majority of their shots on goal.
Correlation between shots and goals: 0.6226384173
Correlation between shots on goal and goals: 0.904453275

[happyforever]

scu84,

Sorry to say it, but you will not find anything!
What you are suggesting is to play attacking football while your fellow coach is more on the defensive side. The way to distinguish which teams play offense and which once are more on the defense, you simply need to look at the total goals scored in every match. I discoverd that the teams total is completely uncorrelated to the teams rating (strength). As such attacking football is nor better nor worse, however there is still a tricky thing about it; being an attacking team means that there will be less draws. Since draws only give one point versus wins three, you may want to go for a bit more of an attack.

Another very interesting factor that I discovered is that it is totally wrong to say that teams that are high scoring have good offensive players and bad defensive players. A very interesting example is the German goal keeper Oliver Kahn. He is one of the best keepers in the world, but all his matches are high scoring. Thing is that if you look at his playing style, he plays very offensive. He never holds the ball long in his hands. He directs the ball in straight attacking positions. The match simply goes a lot quicker. More happens in the same 90 minutes, which inevitably leads to him seeing more goals in his own net than other keepers.

Another issue is why try to win 2-0 with 90% certainty with the risk of having the match ending up in 1-1 at 10% chance, while you can have a "sure" 1-0 victory. Of course doing the first on average will give you 0.8 goals more than your opponent, but if the goal difference is of no importance at all and just the win counts, which it is in the Italian Leagues, then why risk?

[numerista]

Any ideas on how to get this data to back-up my thesis?

Afraid I don't have a source handy (big book published by the USSF, lots of articles by prominent coaches), but it cites a study that found that, in order of effectiveness, the best options in the box were:

1. Shoot
2. Dribble
3. Pass

That said, if you look at any of the world's great players, he (or she) is likely to threaten to do any of the three. Defenders can't give him space because he might shoot, they can't collapse too fast because he might sidestep them or draw a pk, and they can't leave his teammates open because he might find them.

[soccercoachguru]

Wow guys
Glad to see this thread is still alive.

really enjoyed the +/- thread as well.

been busy started Top 10 Soccer Academy (www.top10soccer.org)

named for the lists I have developed over the years a set of top 10 lists ranking individual skills (importance in game play) - individual skills (order to be taught) - team skills - and mental skills

will try to get longer post up in the future - am working on a book on the subject

[DCMike]

I haven't seen Actim mentioned yet in this thread. They seem to be an OPTA competitor, and now the "official" stats for the EPL. See http://www.pa-sport.com/en/products/actim.html and http://www.premierleague.com/fapl.rac?command=forwardOnly&nextPage=actimStatistics

More on Opta is at http://www.sportingstatz.com/.

Anyone care to compare and contrast OPTA and Actim?

Anyone know if Opta/Actim like stats for MLS are available to fans anywhere? Matchtracker seems to be talking about something similar, but I've never seen anyone publishing such detailed stats / ratings for MLS.

[jmacnak]

so has anyone actualy come up with any formulas that hav worked at all?