« Must have been a Monday | Main | Modal wives and why it is hard to marry well »
Which sport has the most upsets?
Soccer looks random to my untutored eye and perhaps it is:
Eli Ben-Naim, Sidney Redner and Federico Vazquez at the Los Alamos National Laboratory in New Mexico decided to look at unpredictability of results - how often a team with a worse record overcomes an apparently superior one - as the best measure of how exciting a league is. "If there are no upsets, then every game is predictable and hence boring," says Ben-Naim.
The team analysed results from more than 300,000 games over the last century from the US's national hockey, football, baseball and basketball leagues and the top English football league. Rugby and cricket were omitted because they do not have a big following in the US.
Their results showed that the "upset frequency" was highest for soccer, followed by baseball, hockey, basketball and finally American football. But when they looked only at data from the past 10 years, the English football Premiership and baseball swapped places, which suggests that soccer might have become more predictable in recent years.
Here is the story. I have long favored basketball. In any given year, barring major trades or injuries, only three or four teams (if that) have any chance of winning the title. You know who the titans are, and you know who the peons are. Limiting randomness and divvying up the ponds in this fashion boosts suspense and status. The old Celtics-Lakers match-ups were ideal. The league is driven by star teams and players, so let's promote those stars. Chess has the same property, but a few good pitching nights can turn a World Series around.
The implied prediction is that basketball and football will have large bases of casually informed fans, typically relying on mass media. Baseball and soccer will have more fanatics, more trivia contests, and will be more deeply rooted in niche media. You have to know about many players and teams to figure out what is going on, who is likely to win, and why.
Posted by Tyler Cowen on January 7, 2006 at 08:06 AM in Sports | Permalink
TrackBack
TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c66b253ef00e5508357828833
Listed below are links to weblogs that reference Which sport has the most upsets?:
» The most exciting sport from New World Man - where's my thing?
The rules changes in hockey have been fodder for discussion about what makes a sport, or a particular game, "exciting," in terms of what will make fans with lots of other entertainment or leisure choices tune in or go to a game. Hockey had a real probl... [Read More]
Tracked on Jan 7, 2006 12:11:42 PM
» Everybody loves an Underdog, right? - Surprises make soccer the best sport from {high quality noise}
Surprises make soccer the best sport
[The results of the test] showed that the upset frequency was highest for soccer, followed by baseball, hockey, basketball and finally American football.
It seems that the actual excitement of watch... [Read More]
Tracked on Jan 8, 2006 9:47:24 PM
Comments
Soccer is so low scoring that the results
of a match are less likely to reflect the
quality of the teams playing and more likely to
turn on a lucky shot or a momentary defensive
mistake.
Posted by: george at Jan 7, 2006 8:46:04 AM
I haven't read the paper, and I don't know anything about soccer either.
But for the other sports, upset frequency seems to be inversely related to
the importance of each individual game to the teams' championship chances.
The Saints over the Patriots in September is many times more exciting than
the Devil Rays over the Yanks in May.
Posted by: Pat L at Jan 7, 2006 8:51:42 AM
The old quip about baseball seems analogous also: Every team wins one-third of their games and loses one-third. It's the remaining one-third that makes all the difference.
It's all about the sample size.
Posted by: KipEsquire at Jan 7, 2006 10:08:54 AM
It's funny that they take unpredictability as the best measure of excitement, as that would imply that coin flipping is the most exciting sport of all.
As a first approximation, I would suggest that the number of upsets is closely related to the number of scores per game, since fewer scores per game implies that the elements of chance or luck that create or prevent a single score will have a larger impact on the outcome. Baseball may have more upsets than this analysis suggests because of important game-to-game differences in the line-up (4-5 different starting pitchers). I'm not sure why football has fewer upsets than basketball. Some potential factors:
1. Football generally requires several good plays for a score. This means that skill has a larger impact on a team's probability of scoring on a given possession, so the importance of luck is smaller than it would be in most other sports with the same number of scores per game.
2. Basketball has a larger home team advantage. This makes upsets more likely when the underdog is at home and less likely on the road, with the net effect probably being an increase in upsets.
3. Because football requires more players at specialized skill positions, there is a larger disparity between good players and bad. This leads to more variance in the ability of football teams, which obviously makes upsets less likely.
4. Psychological factors (probably involving something like motivation, intensity, or focus) result in a disproportionate number of basketball games being close at the end, which effectively reduces the number of scores per game.
I think that 3 is probably most important.
There is also one factor that points against football being more stable than basketball: the number of games per season. An "upset" with an 0-1 team beating a 1-0 team is really not much of an upset, since the teams' records after one game are not very diagnostic of their true abilities. After more games, the record will be a better indicator of true ability, so upsets will be a bigger deal. In sports like football with fewer games per season, a larger proportion of the games will take place before the records tell us much about the ability of the teams (e.g. 1/16 of games occur when each team has just played one game). I'm not sure if the researchers corrected for this in their study. Even if they used teams' end-of-the-season records to determine what was an upset, rather than their records at the time, it would still be necessary to correct for the fact that a team's record after a sample of 16 games is a worse measure of their true ability than their record after 82 games.
Posted by: Blar at Jan 7, 2006 12:05:24 PM
Tyler, you missed a trick here; the clear implication is that all the regimes of draft picks, salary caps, etc etc that are used in the NBA and NFL to keep the sport interesting and the results equal, don't actually work and these sports would be better off following the model of soccer which, AFAICT is the only major world sport to operate a genuinely free market in players.
Posted by: dsquared at Jan 7, 2006 1:54:56 PM
(by the way, these guys are completely reinventing the wheel as is usual for econophysicists. There is a big old literature on the economics of sport, including a big debate over whether "upsets" are a valid measure of the interestingness of the league.
Posted by: dsquared at Jan 7, 2006 1:56:17 PM
"the clear implication is that all the regimes of draft picks, salary caps, etc etc ... don't actually work
This was the insinuation that I picked up on, and although I'm concentrating on soccer the link is below
http://thefilter.blogs.com/thefilter/2006/01/unintended_cons.html
Posted by: AJE at Jan 7, 2006 2:08:59 PM
What do football and basketball have in common, and for that matter what do soccer, baseball, and hockey have in common? The former pair involves high scores with high frequency; the latter trio involves low scores. One would naturally expect that in higher-scoring games, the laws of large numbers would begin to kick in, and the superior team in a given pairing would win with much greater frequency.
Another analogy would be a Vegas one. Say that I were to bet a dollar on red at a roulette wheel. One bet later, I'm up $1 with a probability of about .47 and down $1 with a probability of .53, for an expected loss of six cents. A thousand bets later, I'll almost certainly be down about six cents per dollar bet. Similarly, the percent difference in score between teams in a basketball game will be a pretty good approximation of the difference in the teams' actual abilities that day, since the number of scoring opportunities is so large. On the other hand, even teams with "Sox" in the name win the World Series on occasion. One home run or an error will often determine who wins a game.
Posted by: Chris R at Jan 7, 2006 2:52:19 PM
Daniel, that implication isn't clear at all. There are all sorts of differences between the NFL and the English premier league that could influence the number of upsets, with the difference in the sports played being the most obvious and probably also the most important. This study tells us very little about how the number of upsets in the NFL is influenced by its policies on the salary cap, draft, etc. It tells us even less about the other ways in which these policies are supposed to improve the league, like by creating more year-to-year variability in which teams are successful. I'm sure that there are some papers out there in the big ol' literature on the economics of sport that are actually designed to assess the impact of these kinds of policies (like by comparing a league before and after it instituted a policy). If you want to argue that having fewer regulations would benefit the NFL (or the NBA), you really should be citing one of them.
Posted by: Blar at Jan 7, 2006 3:01:47 PM
They misdefined "upset". It's not just a lesser team beating a better one; it should also include a better team winning but by too narrow a margin. The point spread defines the expected victor and their excepted margin of victory -- it needs to be used.
For line with a money line, they should have compared the incidence of wins by the lunfavored team to the money line's implied likelihood of winning.
Betting lines and spreads are simply pricing and as such are great sources of information . I'd say "Use 'em since ya got 'em..."
There are two other interesting points:
1) There are huge differences in the style of play between the English Premier League, the Spanish La Liga, the Italian Serie A and Germany's Bundesliga. A better study would look to see if the likelihood of upset is league dependent.
2) There does not appear to be a "clumping" of the discrete play style games -- baseball and US football, and the continous flow games -- hockey, soccer, and basketball.
'd love to see the other sports included. Some Americans appreciate both Rugby and Cricket....
The Vagabond
Posted by: Jim at Jan 7, 2006 3:03:57 PM
The Law of Large number is the key - the low scoring rate in football means that lucky fluke scores gives underdogs their day.
But certainty of laws is probably important too:- soccer abounds in subjective judgements that can determine a match – from penalties to off side.
But comparing like with like, the best comparator is rugby union vs rugby league. The scoring rate is similar in both but the laws in Union are far more subjective and the results far more unpredictable.
I’m not an expert on yank sports but I’d hazard that Basketball has the most subjective rules and baseball the least. The fact that baseball is less predictable seems to me evidence that large numbers is the most important factor.
Posted by: GILES at Jan 7, 2006 3:54:50 PM
No mention of Curling? ;-)
Posted by: Christopher Meisenzahl at Jan 7, 2006 6:06:44 PM
The comment about not using the lines is halfway correct.
They are not trying to show to what frequency a team covered. They are just purely looking at upsets. For an upset the unfavored team must WIN.
Although the metric they used to decide who was "favored" should have been the vegas betting line. Record is a poor indicator of the favored team, for instance in basketball recently, teams in the East playing teams in the West may have a better record, but still not be favored.
Also with baseball and pitching, if a poor team has one ace pitcher, they may be favored on nights even when they play a team with a superior record.
I would have to say this flaw really makes any inferences that you can try to wring out of this pretty useless.
Posted by: Zac Barry at Jan 7, 2006 6:19:37 PM
"What do football and basketball have in common, and for that matter what do soccer, baseball, and hockey have in common? The former pair involves high scores with high frequency; the latter trio involves low scores."
Scoring in football isn't as high as the raw numbers suggest. A goal in soccer and a run in baseball count as just one point/run, while in football a field goal is three points and a touchdown, in most cases, seven. A football game therefore often ends in a high numerical score yet involves a relatively low number of scoring events (field goals and touchdowns). A typical baseball game probably has as many total runs as a football games has scoring events.
Basketball, of course, is another thing entirely.
Posted by: Peter at Jan 8, 2006 1:31:00 AM
It would be interesting to compare the two leading individual sports, golf and tennis.
I would guess that upsets are more common in golf than in tennis. For example, Tiger Woods record in the Ryder Cup match play between the US and Europe is only 7-11-2, although most of that is in doubles play rather than singles, where he is 3-1-1. The second best American of recent years, Phil Mickelson is 9-8-3. Davis Love III is 9-12-5. For the last decade or so, the Europeans have not had many true superstars, yet they've done very well against the higher ranked American players.
If you go farther back into the past, the top Americans have better records, such as Arnold Palmer's 22-8-2 and Jack Nicklaus's 17-8-3.
I think a couple of things are different between golf and tennis -- the number of times the ball is touched is fewer in a round of golf than in a five set match, so there's more randomness due to lower sample size. Also, the game of golf is played against the course, the game of tennis against your opponent. In tennis, the better player can overpower the lesser player, but not in golf, except metaphorically.
Posted by: Steve Sailer at Jan 8, 2006 3:24:40 AM
In tennis the bounces are truer. A golfer can hit a ball with a certain degree of precision, but even if he makes a good swing there is randomness in where the ball ends up. The contact with the ball itself probably follows something like a normal distribution. Need to look at the amount of randomness and the number of random events over which the outcome is diversified. Makes sense that basektball is less random with so many events.
Posted by: joshg at Jan 9, 2006 10:26:18 AM
C'mon, this is yet another example of economists taking a decade to describe what is obvious to any informed current-day spectator.
OF COURSE football (soccer) has become more predictable -- it's become a game that is dominated by economics. Just as the NFL has super-market teams that can pay for superstars, so too has the European football leagues. Real Madrid, Chelsea, Bayern-Leverkusen are all leaders in their divisions for team revenue and player expense.
If an economist wanted to do an interesting analysis, then they should probably correlate the amount in revenue and/or player expense against this predictability function. I'd go further and say that looking at the income gap between top and bottom-ranked teams plus the increase in revenues league wide would determine when sports become predictable.
I mean seriously, it took a statistical study to figure this out? No wonder I never listen to economists...
Posted by: TechTrader at Jan 9, 2006 12:26:31 PM
I'm surprised no one has mentioned that the conventional wisdom suggests that ML Baseball has too many teams. If the best players are thinned out by bidding wars between marginally successful teams, then we would expect that there would be more upsets (lower variance from the mean).
In other words, I think this study does not measure "excitingness", but rather the variance between the best and worst teams. After years of Lord Voldemort, err, Dan Snyder's rule, I think it becomes obvious that there is a huge gap between the NFL's best and worst. Does this necessarily make the league less 'exciting'? Heck no - the diffusion of talent in MLB means a lower quality of play, and less excitement.
Also, it is (almost) impossible to watch every game in any league during any given weekend, and even if one does accomplish this feat, repeating it would be painful. Most sports fans tune in for some subset of the most exciting contests of the weekend - this weekend I may watch, say, Skins v. Seattle and Pats v. Broncos - and catch highlight reels of the other games, thereby satisfying 'excitingness'. In other words, the overall league 'excitingness' is a poor measure of the excitement value of a given sport league, as a given fan will only be interested in viewing a subset of the games played.
Posted by: Robert at Jan 9, 2006 1:19:05 PM
Has anyone considered that football (American) and to a lesser extent basketball are contact sports.
Soccer is relatively low contact and baseball is obviously non-contact. Contact sports like football that place a high importance on strength and speed will not have many upsets as any disparity in pure atheltic ability eg speed, strength etc will be exploited.
I don't know much about baseball but soccer clearly has the most upsets. Teams several levels below can and do beat much high ranked team (especially in the FA Cup):
http://www.abc.net.au/sport/content/200601/s1543765.htm
These results would be analagous to a college (or high school) team defeating an NFL team. Something that would never happen.
Posted by: Peter Shaw at Jan 9, 2006 7:43:59 PM
I think measuring the 'excitement' of a sport based on upset frequency is downright stupid. Not to mention that it's pretty self-evident that world football with the least 'scores' is going to be the most variable. This seems almost like an excuse to justify its popularity against scorn from the US, when such a justification is obviously not necessary.
"Just as the NFL has super-market teams that can pay for superstars"
This commenter is not familiar with the NFL system, there is now a hard salary cap.
"Tyler, you missed a trick here; the clear implication is that all the regimes of draft picks, salary caps, etc etc that are used in the NBA and NFL to keep the sport interesting and the results equal, don't actually work and these sports would be better off following the model of soccer which, AFAICT is the only major world sport to operate a genuinely free market in players."
How can you possibly measure the "regimes" across different types of sports? Soccer will always have the most upsets because of the nature of the sport, the NFL's 'failure' to have more upsets than soccer has nothing to with the salary cap. If you want to disprove the salary cap, you could only compare NFL pre-cap and NFL post-cap. It seems fairly obvious from eyeballing the NFL data the the cap has increased upsets (If that's even a 'good thing', see my first comment).
Posted by: BillWallace at Jan 9, 2006 10:09:03 PM
fioricet http://fioricet.atspace.us ; Thanks!
Posted by: fioricet at Jan 26, 2006 7:07:26 PM
Hi. You have a great site, thank you for this info. I want to present fitness sport, please, give a moment to your own health.
Fitness
Equipment fitness
Beauty fitness
Center fitness
Fitness
Fitness Health
La fitness
Fitness Magazine
Fitness Muscle
Fitness program
Posted by: Donah Beggins at May 14, 2006 4:56:38 PM
http://iopap.8jsoshns8.info
http://mpcore.8jsoshns8.info
http://comanp.9a3kakad.net
http://setro.9a3kakad.net
http://ceosa.as3sad3.info
http://oapea.as3sad3.info
http://aieksa.asiasdpe.org
http://oaiosa.asiasdpe.org
http://osopas.asidkoasd.org
http://ueioa.asidkoasd.org
http://mpcore.eioasiap.org
http://mpcore2.eioasiap.org
http://mpcore.i839aka.info
http://mpcore2.i839aka.info
http://mpcore.iaosde.org
http://mpcore2.iaosde.org
http://mpcore.io9s9akk.com
http://mpcore2.io9s9akk.com
http://mpcore.kao90as.info
http://mpcore2.kao90as.info
http://mpcore.kasio9ksa.info
http://mpcore2.kasio9ksa.info
http://mpcore.kaso0ke.com
http://mpcore2.kaso0ke.com
http://mpcore.ki93kas.info
http://mpcore2.ki93kas.info
http://mpcore.ns7837.net
http://mpcore2.ns7837.net
http://mpcore.oas09j7a.info
http://mpcore2.oas09j7a.info
http://mpcore.oas930as.com
http://mpcore2.oas930as.com
http://mpcore.oaspoaspo.org
http://mpcore2.oaspoaspo.org
http://mpcore.sh84ssa3.net
http://mpcore2.sh84ssa3.net
http://mpcore.si9390s.com
http://mpcore2.si9390s.com
http://mpcore.sla903sa.info
http://mpcore2.sla903sa.info
Posted by: oopdpa at Jul 8, 2006 12:06:28 PM
serveerster striptease ^^^ schoolmeisje creampie ^^^ pornostars roulotte ^^^ grosse scatto ^^^ chaud pere ^^^ penaud mediocre ^^^ mama trente six ^^^ frais fils dix huit ^^^ piu carino amatoriali fotti ^^^ grazioso allievo amore ^^^ skygg servitris gruppen ^^^ skrattretande skolflicka ligga ^^^ mousey giovane figa fotti ^^^ strappare fighette fotti ^^^ himmelsk 3kant ^^^ pa kontoret piccolo ^^^ overjordisk hentai ^^^ overjordisk datter ^^^ agapitos efivos agapi ^^^ paralogos mora agapi ^^^ papa trente six ^^^ maman trente et un ^^^
Posted by: levan at Sep 6, 2006 2:13:20 AM
Posted by: at Oct 14, 2008 12:14:05 AM
