Sabermetrics applying to Soccer

[beineke]

Originally posted by maxim-1
When I see Poisson distrabutions I think of being back stage with Bret Micheals.

The Poisson distribution is (hopefully) not too complicated. If the Quakes score 20 goals in 10 games, we know that they average 2 goals per game, but we don't know how consistent they are. Perhaps they score 2 goals every game, or perhaps they alternate between scoring 4 goals and scoring 0. If the former is true, the Quakes will rarely lose; if the latter is true, they'll lose quite often.

When we say "Poisson," we can answer how consistent the Quakes are. The Poisson distribution makes the simple assumption that goals are scored at random time intervals. That is to say, we can ignore all of the complicating factors of tactics and streakiness and fatigue and match-ups and home field advantage. If the Quakes are going to score 20 goals in 900 minutes, we choose 20 random numbers between 1 and 900, and those are the minutes that the Quakes will score.

This gives us an easy way to estimate the probability that the Quakes will get shut out, or score one goal, etc. In most cases, it works pretty well. Then again, the assumption is factually incorrect, so it won't be perfect. Every rose has its thorn.

[superdave]

You know how whenever a game hits the 41st minute or so, and it's 0-0, the announcers make a big deal of one team trying to get a goal to seize momentum?

Here's an interesting question. Breaking the game into 5 minute blocks, what is the record for the team scoring the game's first goal in each block? Obviously, the teams scoring the first goal of the game in the last block would be pretty near perfect. I would expect that scoring first in the first five minutes would be the least important. Anyway, if, in fact, momentum is a big deal, then scoring a goal in the last 5 minutes of the first half would stick out. Teams would have a much better record than teams scoring in the 36th to 40th minutes, or the 46th to the 50th.

Yes, I realize that because of stoppage time, there won't be an even distribution of goals scored. But I don't see that as relevant. I'm interested in the record.

Where's an easy place for me to check this out over the last 1 2/3 MLS seasons? I might work on it at work (during lunch only, of course.)

[kenntomasch]

The league breaks it down by 15-minute intervals, but not five.

I've often wondered about that end-of-half momentum thing, too. Would be interesting to just pore through the boxscores and see goals scored from 40-45 minutes (and first half stoppage) and see if there really is an effect, either on the team's performance in the second half or its changes of winning the game.

[beineke]

Originally posted by superdave

Here's an interesting question. Breaking the game into 5 minute blocks, what is the record for the team scoring the game's first goal in each block? Obviously, the teams scoring the first goal of the game in the last block would be pretty near perfect. I would expect that scoring first in the first five minutes would be the least important. Anyway, if, in fact, momentum is a big deal, then scoring a goal in the last 5 minutes of the first half would stick out. Teams would have a much better record than teams scoring in the 36th to 40th minutes, or the 46th to the 50th.


This does sound like a good idea. I'm not aware of anywhere convenient for MLS data, but you could easily get the Premiership here (as well as several other leagues):
http://www.rsssf.com/tablese/eng03.html

If I may make a suggestion, perhaps you can record a bit of side information for each game, too, in a format like the following:

HOME TEAM,AWAY TEAM,Scoring First,Minute, Final Score
Charlton,Chelsea,Home,7,2-3

The reason I think side information will be useful is that you'll probably want to know whether the first goal is more important depending on the circumstances.

[superdave]

kenn, that's a different angle to take. I'm just looking at first goals, you're looking at any goal.

That would be fun, too. Ask, did a team score in this 5 minute block? What was their record if yes?

Now that I've thought a bit more, I think I'll look at all tiebreaking goals, not just all first goals.

[kenntomasch]

No, I was just thinking about goals scored right before halftime, which are supposed to be momentum-shifters. Not first or tiebreaking goals, just goals.

[FlashMan]

Originally posted by kenntomasch
No, I was just thinking about goals scored right before halftime, which are supposed to be momentum-shifters. Not first or tiebreaking goals, just goals.

I'm coming in at the end of a conversation (and without reviewing it) but that "Analyze This" at mlsnet.com analyzed goals just before or after the healf and as it turns out they don't effect wins or losses any more than any other goal.

Something to that effect. It was a few weeks ago.

[mpruitt]

You know what would be interesting too? To look at games where opposing teams, or even opposite teams score goals in bunches. As you saw for example with the U-17 Brazil game. How many times are teams getting beat by a quick outburst of goals, even if it's just two. I'd think froma coaching perspective after objectively figuring out how often this happens that a little subjective analysis would go greatly towards hoping to understand why a game can sometimes so easily get blown open. It'd also be an interesting question to ask to gauge the consistency of teams, even if there was frequent 10-15 minute intervals where their opponents were knocking in 2 or more goals by subtracting those you might get an interesting idea if a team is living up to its full potential. Just a thought.

P.S. Please please please go to the Suggestions and Announcements forum, and the New Forums section and post if you think it'd be a good idea to have a forum of it's own for this type of conversation. The guys at RSSF.org are great, but they're dealing with historical trivia, and honestly right now we're just bantering around questions and ideas. I think there's a golden opportunity to really get into some substinative analysis and discussion if we could just do it in a more organized format.

[kenntomasch]

Well, there you go. No need to do that one.

Doesn't really surprise me. I've never been completely convinced of the true magnitude of "momentum" in sports.

I think if it actually exists, it's probably strongest in basketball (or, at least, basketball looks like it has more readily-quantifiable shifts in momentum, and players and coaches act as if it really exists).

[Flyer Fan]

I have all of the Crew goals scored and allowed - for all competitions - broken down into the following intervals: 1-5, 6-10, 11-20, 21-30, 31-40, 41-45, 46-50, 51-60, 61-70, 71-80, 81-85, 86-90 and 90+ (overtime matches). They're all referenced by Crew match. If anyone wants to have fun analyzing that, let me know (it's in Excel format right now; hoping to move it into a database during the winter).

[mpruitt]

http://soccernet.espn.go.com/feature?id=274389&cc=5901

This looks interesting. A Gamecast feature offered by ESPN which I hadn't seen before, someone mentioned it on the Aresenal board. It's prettyd detailed. Not as detailed as one would like but still pretty good.

[superdave]

Originally posted by kenntomasch
I think if it actually exists, it's probably strongest in basketball (or, at least, basketball looks like it has more readily-quantifiable shifts in momentum, and players and coaches act as if it really exists).
I once designed a roll-the-dice game for college basketball. I learned from all the games I had played, and if I do say so myself, my game kicked ass.

Anyway, from playing that game, I think alot of the momentum in basketball comes from the fact that it's easier to score after you get a defensive rebound or a steal, because the other team can't set up their defense. So then you score, so you can set up your defense, so it's hard for the other team to score. Which makes it easier for you to score. Which makes it harder for them to score. Which makes it...you get the point.

Then throw in free substitutions, and while my gut instinct is that there's some psychological component, it's minor.

[microbrew]

I don't really buy into the concept of momentum, or hot streaks or cold streaks, or players being on fire. I mean, you can pick out such streaks in a coin flip simulation.

Since a soccer game seems to be well modeled by a poisson process, the assumption is that each goal is an individual and independent event. So, when the next goal is scored- it's irregardless of the time, or how many goals have been scored.
Um, who'd thought I would be trying to explain iid to Bigsoccer. Of course, I haven't found any studies looking into if soccer goals are actually independent and identically distributed.

http://mathworld.wolfram.com/PoissonProcess.html

[Nutmeg]

Originally posted by microbrew
I don't really buy into the concept of momentum, or hot streaks or cold streaks, or players being on fire. I mean, you can pick out such streaks in a coin flip simulation.

Since a soccer game seems to be well modeled by a poisson process, the assumption is that each goal is an individual and independent event. So, when the next goal is scored- it's irregardless of the time, or how many goals have been scored.
Um, who'd thought I would be trying to explain iid to Bigsoccer. Of course, I haven't found any studies looking into if soccer goals are actually independent and identically distributed.

http://mathworld.wolfram.com/PoissonProcess.html

Manure. See 2003 FIFA U17 World Cup; Cameroon v Portugal for plenty of evidence to the contrary.

[kenntomasch]

In other news, on the basis of one game from May of 2002, I have determined that Mike Cameron is the greatest home run hitter of all time.

[superdave]

And on the basis of two games in 1938, I can prove Johnny Vandermeer is the greatest pitcher of alltime.

And on the basis of one post, I can prove jri is the greatest poster on bigsoccer. ;)

[voros]

Originally posted by superdave
I once designed a roll-the-dice game for college basketball. I learned from all the games I had played, and if I do say so myself, my game kicked ass.
Thomas Gilovich studied the "hot hand" in basketball way back when (late 70s?) and pretty much came to the conclusion that if it exist, it was inseparable from what would be expected by random chance alone.

Gilovich's "How We Know What Isn't So" should be required reading for every high school student in this country. Fascinating book about the reasons why our reasoning often leads us astray.

[kenntomasch]

Originally posted by voros
Gilovich's "How We Know What Isn't So" should be required reading for every high school student in this country. Fascinating book about the reasons why our reasoning often leads us astray.

Is part of it because the human mind strives to make a connection between discrete events?

[beineke]

Originally posted by microbrew

Um, who'd thought I would be trying to explain iid to Bigsoccer. Of course, I haven't found any studies looking into if soccer goals are actually independent and identically distributed.


Goals most certainly are not iid, but that's not the point. What matters is that pretending they're iid gets you a pretty good approximation to reality.

One other note: it's true that psychological perceptions of streakiness are no good. Voros mentioned a study about baseball, and Tversky did something similar with free throw shooting in basketball. At the same time, tennis does show some degree of streakiness ... the player who won the first set has a substantially higher chance of winning the second. So we shouldn't write off the existence of streakiness, just because our intuition is a poor judge of it.

[kenntomasch]

Originally posted by beineke
At the same time, tennis does show some degree of streakiness ... the player who won the first set has a substantially higher chance of winning the second.

Or would that be because the player who wins the first set has a higher chance of being the better player?

One problem I might see with that right off is that so many matches in tennis tournaments are between players with a wider disparity in ability - in a tournament with 128 players, you're going to have 96 first and second-round matches and only 31 matches in the third, fourth, quarters, semis, and finals, where the differences between the players' general level of playing ability is (theoretically) going to be lessened.

That's like the announcers every January who tell us how important it is to get that bye week in the NFL Playoffs because the week of rest is really important and teams coming off the playoff bye week usually win the next week. Well, they earn that bye week by proving they're a superior team. I'd expect a 13-3 team to beat a 9-7 team at home, whether they'd had a week off or not.

Not that I'm dumping on the importance of a week's rest after a 17-week season in a sport as physical as football, but it seems to me the primary reason most teams who earn the first-round playoff bye win the next week is because they're better teams, not primarily because they had a week off.

And I'd tend to believe that if you win the first set in tennis, you'd have a good chance of winning the second just by virtue of being a better player, and because of the aforementioned disproportionate (theoretically) amount of matches between players of disparate abilities.

I'd like to see the research, though.