The Statistical Problem With Soccer

http://www.technologyreview.com/blog/arxiv/24182/

by Anon84 • 17 years ago

36 37 17 years ago

newsdog3• 17 years ago

All sports are like this - they try and eliminate it with multi game playoffs, but you can't do that with intense games like this.

jimbokun3• 17 years ago

The NBA, NHL, and Major League Baseball do exactly that. They eliminate with multi-game playoffs.

Or do you mean soccer's intensity is like American Football, in that it is just too physically demanding to play that many games?

newsdog3• 17 years ago

Exactly - also the nba nhl etc with multi game playoffs LESSEN the prob of a lesser team betting a better one - but it never goes away.

lionhearted• 17 years ago

I used to play a lot of cards. Like, a whole lot. Every now and then a card-playing friend of mine would complain about their luck, and I always answered the same way:

"People don't gamble on Chess."

So very quickly:

> Their approach is to think of a soccer game as an experiment to determine which of two teams is the best. The question then is this: what is the probability that the outcome of the experiment truly represents the relative abilities of the two teams. And the answer, unfortunately, is not very probable.

This is not a bug; it is a feature. Spectator sports have drama and theatrics and exist first as entertainment, and only second as a contest of skill.

In Major League Baseball, the New York Yankees are far and away the best team this year. It's not even close. Almost all their position players are between top 3 and top 10 at their position, they have a very solid top of the rotation and mid rotation, and good to amazing relief pitching.

They'll compete with eight other teams to win baseball's championship, the World Series. I'd put their chances around 22% of winning it all. But they're far and away, clearly, the best team in baseball.

But that's why people watch. People don't gamble on Chess.

locopati• 17 years ago

Furthermore, the contest is not skill-on-paper or skill-over-a-season, the contest is skill-right-now. The NE Patriots won 18 straight American football games (the only team to have an undefeated 16-game regular season), yet they still had to try to win the Super Bowl. They came within a couple of minutes of doing so until an improbable catch gave the Giants the chance to win the game. As ESPN's Chris Berman once said, 'that's why they play the game.'

jibiki• 17 years ago

People do gamble on chess... but I guess "people mostly prefer gambling on games like poker rather than gambling on chess" would be too long for a proverb.

chickamade2• 17 years ago

It might be true the best team only has 28% chance of winning the World Cup. Though, we should know that it leaves only 72% chance for the rest of the 31 teams to win, and they are not uniform. So maybe 28% for the best team, 20% for the 2nd best, 12% for the 3rd best, etc., which sounds fair to me.

There are 32 teams playing the World Cup, if one of them is a 1:3.5 dog, they have a pretty DAMN GOOD chance. The betting odd is usually sth like 1:4 or 1:5, I think.

More importantly, the team which wins is the better team, by definition. It is a question whether they can perform consistently in a series of matches with the correct strategy each time, whether they can adapt to changes, whether they are focused at the important moments. Certainly there will be luck and drama, but the team which plays better (for most of the match duration) wins much more often than that number 28% tells you.

Look at professional poker. You can for sure say that there is more luck involved, but times and again the familiar faces show up at the final tables.

chickamade2• 17 years ago

Actually I made a mistake, 28% would be 1:2.5 dog in betting lingo, and that's a HUGE when there are 31 other competitors.

jimbokun3• 17 years ago

I think this is at the root of what really bugs (some) U.S. Americans about soccer, but they are unable to express it.

Sports popular in the U.S. almost seem designed to allow for statistical analysis. Baseball is the ultimate game for a statistician, as you have data for every single pitch thrown and the outcome. American Football breaks the game down into series of discrete events (downs) each of which has a numerical outcomes (yards gained or lost). Basketball generates 10s of events every game with a definite outcome that can generally be assigned to a single player (shot made, missed, turnover, free throw, rebound, etc.).

All of this gives fans ways to analyze and rank individual players and teams in a fairly objective way. Thus, the immense popularity of fantasy leagues.

Soccer is inherently more subjective, as there are far fewer events with an objective outcome. With scoring a rare event, and no obvious ways to assign numerical values to all the things that happen between scoring chances, subjective analysis is all that's available.

Which ties back into the original article in that freak occurrences and dumb luck can have a much larger impact in soccer. One team can play much better for most of the game, but a freak goal, controversial call or uncharacteristic error can have an outsize impact in a low scoring affair. This is mitigated in the round robin format rounds with accumulated wins and points, but is very much a factor in one-and-done games.

I believe I read before that at one time even the final round was a multi team affair, with the champion determined by cumulative totals against the other teams, but it was done away with because fans wanted a definitive championship match between just two teams.

philwelch• 17 years ago

Dumb luck plays an even bigger role in American football though. In American football a small amount of turnovers can dictate the result of the whole game. It is way more unpredictable than even soccer, which has some 90 minutes of uninterrupted play where each team continuously tries to either hold or gain possession and make an attack.

Plus, soccer has more statistics than you'd realize: proportion of possession, number of corner kicks, number of crosses, number of shots, percentage of shots on target, possession per area of the field, etc.

_3ex7• 17 years ago

When discussing a players career though, typically its goals, and appearences per club.

sfg• 17 years ago

You can get a lot of statistics from a game of cricket, but it does not seem to have taken off in the U.S.

Perceval• 17 years ago

Cricket and baseball are substitute goods. If you have one, you don't really need the other. For example, baseball hasn't taken off in the Commonwealth either.

miracle• 17 years ago

It's called football not soccer.

newsdog3• 17 years ago

Correct. But we dumb North Americans need to make the distinction with the weird violent USA game, which almost everyone in Canada cares about, myself included.

Somehow, they convinced us it's important. I really want to know how they did that.

hughprime3• 17 years ago

I'm sorry, that game is called "Grid Iron". The word "football" is reserved for Rugby League.

(Not to be confused with Rugby Union, which is "rugby" or Australian Rules Football, which is "footy". And definitely not soccer, which is "soccer".)

diego_moita• 17 years ago

This is actually an excelent answer. If you can't dig it, you can't dig what the game is all about.

heyitsnick• 17 years ago

- "In fact, Skinner and Freeman have carried out an extensive statistical analysis of the scores at the last World Cup. They point out that if the outcomes of games were a true reflection of the teams' abilities, then the situation "Team A beats Team B beats Team C which beats Team A" should never occur. They call this an intransitive triplet."

The outcome of the game is never meant to be a "true reflection" of the teams' abilities, but a reflection of their ability against that side, on that day, in those conditions. A round-robin style tournament is used to have a team playing different opponents, a mix of different styles. Central attacking sides. Defense and counter-attack. Wide open matches. Closed central movement. Teams alter their strategy against specific opponents and try to find holes in their game and exploit an edge.

This seems like a problem with the statistical model, not with the sport, because "A > B > C > A" is possible in a team sport, and not just because of a luck factor. One team may have strengths that exploit certain weaknesses in other teams. That are uncountable examples - Team A lacks fitness and is weaker in the last 20 minutes; but this will only be exploited by a team who have a high fitness level. Team B has as weak left wing, but this will only be exploited by a team with a strong right winger. The list goes on.

radu_floricica• 17 years ago

When talking about a single game, yes. But in larger numbers of games you'd still expect to see a bigger difference between better teams and not so good teams, and the article argues the difference is not so big (17% intransitive tuples vs 25% in random pairs).

xiaoma• 17 years ago

That was my thought exactly. Pikemen beat cavalry who beat footmen who beat pikemen.

cromulent• 17 years ago

Soccer is a low scoring game. It's not a "statistical problem", it's a feature.

Treating it as "an experiment to determine which of the teams is the best" is not a good approach. There are a thousand contests in 90 minutes, but probably only two or three of them result in a goal, one of which may be from a penalty anyway.

High scoring games are much better at determining the best team on the day.

I believe that one of the major reasons for soccers massive worldwide spectator popularity is the low scoreline. You watch the game even if your team is obviously outclassed, as you can still win, and you can still win with just a few minutes left in most cases.

ovi2562• 17 years ago

I can't remember how many games I stopped watching 5 minutes before the end, because I was disappointed, only to later discover that the team I supported scored. Even better are score, counter-score, then score again.

frossie• 17 years ago

Damn right it's a feature. This is why you shouldn't let physicists come up with sports rules

The point of the format is to give a team a chance to win even if they are not the best team. That's what keeps people interested - the chance that their team will win, even thought it is the 2nd, 3rd or 10th best team. When Greece won Euro 2004 against all odds, the psychic energy was just incredible. When Germany wins, it's like "Well, there we go, no surprise, let's go have a beer".

chimariko• 17 years ago

The flaw of Gerald Skinner's approach is that he makes an assumption that there is some kind of absolute abstract quality of a soccer team, whereas the quality of a soccer team is defined by their results in a tournament.

ars• 17 years ago

The game would be pretty boring if the best team always won.

It's a game, it's meant for fun, it's not a science experiment.

By having it semi-random, you spread out then ups and downs.

Someone did an analysis of video games (posted on slashdot, but I can't find it), where they showed that the best video games don't have solid excitement all the way, but rather ups and downs.

The same it true here. If your team always won you would be bored pretty quickly.

mechanical_fish• 17 years ago

The game would be pretty boring if the best team always won.

Indeed, as many sports folks have pointed out, one of the most popular annual sporting events in the USA is the NCAA college basketball tournament. A single-elimination event which is pretty much designed to improve the odds of having the "best" team get accidentally disqualified by an improbable fluke. The winner must win six consecutive games, while under enormous pressure, without making a mistake.

As it happens, the distribution of talent in college basketball still makes the seedings work fairly well. Just as World Cup soccer seedings work decently well, though it too becomes a single-elimination tournament after the first round. But these tournaments could obviously be much better designed if the objective was science.

chasingsparks• 17 years ago

Link to the abstract (w/ PDF download):

http://arxiv.org/abs/0909.4555

oliveoil• 17 years ago

I think that it is exactly because there is so much variance in the outcome of soccer matches that the sport is so much more popular than any other sport (here in Europe at least).

dschobel• 17 years ago

My biggest objection to this piece is the quantification of skill along one axis, "the good scale".

A beat B, B beat C, A loses to C should tell you that there most be something sophisticated going that your model isn't capturing.

American Football fans will surely remember when Vince Young was starting his career with the Titans and all the commentators had this standard line "He just wins games!" because his team was winning despite his terrible numbers.

This analysis doesn't strike me as any more substantial than that ridiculous phrase.

coliveira• 17 years ago

One obvious problem with the approach taken is that soccer teams are not static. They change from game to game. They can start playing badly and improve during the competition, which could easily explain the observed behavior.

Another problem is that soccer really has a level of randomness that is considered a feature of the sport. We don't really expect that the best team (in technical terms) always win. In Brazil, people commonly say that "soccer is a little box full of surprises".

ovi2562• 17 years ago

Around Europe, we say : 'The ball is round and the game lasts 90 minutes. The rest is all theory'.

This means, more or less, that there are so many unknowns that trying to analyze the game is bullshit - just enjoy it. Even the quality of the grass can change the game, or the humidity. But of course, sport journalists make a living out of over-analyzing it.

Danny72• 17 years ago

Although the World Cup is the most well known (and watched) soccer tournament, it isn't that representative of soccer.

1. It's mainly a knockout tournament which obviously increases the variance

2. International sides aren't a cohesive team like domestic club sides only playing a handful of times per year leading up to the tournaments.

Compare this to the domestic leagues where the league format reduces the variance and players spend most of the year working with who they're playing with on the pitch. In the EPL, the #1 ranked side going into the season probably wins the league 75% of the time. It's a much lower number though for the domestic cups because of the knockout format.

Deron• 17 years ago

Italy 2006, Brazil 2002, France 1998, Brazil 1994, Germany 1990, Argentina 1986....

The end results do seem to favor the best teams.

The only people who don't recognize that are England fans, who routinely overestimate their team anyway.

The Statistical Problem With Soccer

Related Stories