Re: World Cup Simulation Results The next most popular question besides "who wins?" is probably "who advances?" Here are the chances of advancement (1st or 2nd place) by group: Group A Germany = 91.93% Poland = 52.81% Ecuador = 32.99 Costa Rica = 22.27% Group B England = 76.61% Sweden = 73.77% Paraguay = 46.36% Trinidad & Tobago = 3.26% Group C Netherlands = 75.73% Argentina = 68.33% Ivory Coast = 28.76% Serbia and Montenegro = 27.18% Group D Portugal = 81.13% Mexico = 68.59% Iran = 32.48% Angola = 17.8% Group E Czech Republic = 74.1% Italy = 67.55% U.S.A. = 46.68% Ghana = 11.67% Group F Brazil = 86.12% Croatia = 47.6% Japan = 33.18% Australia = 33.1% Group G France = 91.59% Switzerland = 51.59% South Korea = 41.6% Togo = 14.86% Group H Spain = 86.92% Ukraine = 49.61% Tunisia = 49.47% Saudi Arabia = 14% Here is one possible result (taken from one sim) that would have the U.S.A. advancing: Italy 1 - 0 Ghana Czech 1 - 1 U.S.A. Italy 2 - 0 U.S.A. Czech 3 - 0 Ghana Italy 4 - 1 Czech U.S.A. 3 - 0 Ghana U.S.A. advances with 4 points on goal differential over the Czechs. Germany beats out France for having the highest chance of advancing. T&T has the lowest chance. The stage is set for Africa to not have a very good cup as their best team (Ivory Coast) is saddled with a near impossible draw. As such Tunisia appears to have the best chance of advancing of the five African teams. There are five teams that the system ranks below the U.S. that the sims say have a better chance of advancing due to the bad draw the U.S.A. got: Croatia, Poland, Ukraine, Switzerland and Tunisia. None of the teams the system says is better than the U.S. has a worse chance of advancing, none are close. Officially speaking, this was indeed a tough draw for us. Finally I should note that the system may very well underrate favorites slightly in individual matches, meaning that the top teams percentages maybe should be a touch higher and the bottom team's percentages maybe a touch lower (the middle teams should on the money mostly). I'm torn on this question and I won't bother you with the details on the whys and wherefores. Results are an imperfect way to assess team strengths due to the many factors beyond team strength that can go into any individual result. It is my opinion, however, that they are the least imperfect of the known various methods you could use to answer the question.
Re: World Cup Simulation Results Very cool analysis, Voros. I know that the model has been validated in a different sport and at Euro 2004, but just for s**** and giggles I wonder how well it would predict the results of the last few world cups?
Re: World Cup Simulation Results "You must spread some Reputation around before giving it to voros again." As a Math/Statistics major this stuff was really interesting. I've been playing around with a much less sophisticated system for American Football for awhile now. Any insights you could give into the actual nuts and bolts of the program would be greatly appreciated.
Re: World Cup Simulation Results The problem is you'd be asking me to do a lot of work that I haven't done yet. I'd have to gather up a bunch of old results, re-run the rankings circa May 1, 2002 and 1998, re fiddle the simulations. Have to rename and re-calibrate various countries due to the fall of the Soviet Union, etc. I could do it, I don't think I could clear the time to get it done before the cup starts. I could tell you quite simply that at the very least both Brazil and France would have been given decent chances to win each cup. In France's case, you can see how far upward home field has dragged what appears to be one of the weaker German cup teams. You saw what it did for South Korea as well. Consider the numbers a good ballpark estimate but not any better than that. Certainly not worth risking money over just yet. Like I said, when you say something should happen 12% of the time, you need many trials to be able to judge how close to accurate such an estimate was.
Re: World Cup Simulation Results Covered, and I have exactly the same sentiment. I've been fascinated by this thread! Just out of curiosity, has anyone out there (or can anyone ) run a similar set of probabilities based on the ELO rankings? I'm curious how their probabilities would compare to Voros's.
Re: World Cup Simulation Results Looking through old files, it looks like I have the results necessary to do WC2002 already done. Maybe I'll take the time to run the numbers and see what comes out. I think WC1998 is out though for now. The results go back 8 years to get the necessary sample size for some of the teams, but for any team with a halfway normal schedule, results older then 5 years don't mean much at all, and games from last week obviously count more than games from three years ago.
Re: World Cup Simulation Results If France is at 12 to 1 for example, can you bet the field meaning bet 1200 (+ vig) to win 100?
Re: World Cup Simulation Results Again, great thread! Re 2002, it's hard to imagine that an objective program would predict Argentina, France and Portugal failing to advance out of their groups. What I assume it would show is that even a strong favorite to advance (e.g., Brazil, at any WC) is nowhere near a 100% 'lock' to advance. Speaking of France, I don't think I overstated their chances. They should advance, and having advanced, with their defense they will be very difficult to beat. Some team should start practicing their PKs in anticipation. I'm going to comment separately on your groups, but the percentages in our group certainly surprise me. Even as a clear third choice, the US has almost a 50% chance of advancing. Not quite as bad as many may have thought.
Re: World Cup Simulation Results That stuck out for me, too. I know the sim is supposed to trump most subjectivity about the chances, but I can't believe the US is a coin flip to get out of this group (nor that Ghana only has a bit better than a 1 in 10 chance). That seems like too much of a spread for me between those two.
Re: World Cup Simulation Results Agreed. I think the good news is that if Ghana pulls a 'surprise,' it will be against Italy or CZE. We're not good enough yet to take any opponent for granted. Ghana may beat us, but they won't surprise us.
Re: World Cup Simulation Results Great work Voros! Have you examined the probability of seeded/non-seeded team reaching each stage of the playoff round? Comparing that to actual results in the modern era might be a clue about whether your model if giving too big a chance to the long shot teams.
Re: World Cup Simulation Results It amazes me how many around here are just blindly chalking that game up as three points. We may beat them, but it sure won't be easy if our history against opponents with similar styles is any indication. I can't see the US being potent enough on attack (or stingy enough on D) to win a GD battle for second with ITA or CZE if we have four points, meaning we likely need at least two points heading into Game 3 AND a win over Ghana to make it through. It won't be easy, but I'm still damn excited.
Re: World Cup Simulation Results Voros, if your numbers are *anywhere* near right they're worth betting and betting big. Because of the fact that Brazil, England, Argentina and Italy are dramatically overrated in the bookmakers odds; and France, Czech Rep, Spain, the Netherlands, the US and Portugal are dramatically underrated, you could make a bet with an expected value of about +75% of your bet. For example: Country.....Best Odds.....Wager...Net Payout (including losses) Holland......15.6.............$95.........$1177 Spain........17.1..............$85.........$1138.5 France.......14................$100.......$1100 Czechs.......33................$45.........$1130 USA...........100..............$15..........$1115 Portugal......25...............$60..........$1160 According to your odds, there's about a 45% chance one of those teams wins the World Cup. So you'd have a 45% chance of winning $1100 or so, and a 55% chance of losing $400. i.e. it's an unbelievably good bet.
Re: World Cup Simulation Results I agree - there's a lot of wishful thinking out there. And remember, after you look at goal difference, you look at goals scored, making 3-2, better than 2-1, better than 1-0; and 2-2 better than 1-1, better than 0-0. Your first paragraph illustrates a common 'logical' failing (of others) here on BS. Many people seem to confuse confuse two concepts, which I will simplify as follows: (1) We need to beat Ghana; and (therefore) (2) We will beat Ghana. Now you and I seem to be able to grasp the difference, and the fact that the first - the wish - does not automatically lead to the second - the happy result. My fear, expressed at length in other posts, is that our game with Ghana may be almost meaningless, assuming that Italy and CZ have positioned themselves to qualify with a tie in the other game 3. I'll be very pleased if we meet Ghana with our fate in our own hands, but I'll be pretty anxious if we need to beat them by more than one goal or score a certain number of goals. Our team is not built to run up the score. My hope is that we properly respect the experience of the Czechs in the first half - maybe the first hour - holding our own, of course - then make them look old and tired, big and slow in the second half/last half-hour. If we play CZ to a 0-0 tie, a result that looks pretty good 'on paper,' and we later fail to qualify (for any reason, including goal difference/goals scored), we will be getting just what we deserve. Let's go for it in Game 1, particularly if we're tied after 60 minutes.
Re: World Cup Simulation Results Since the draw, I've been saying the Czechs are very beatable, but I'd be fine with a draw, as it would allow us to try to steal a point versus Italy and basically be in with a win over Ghana (by any margin). I think it's very unlikely the US can beat Italy (especially before Italy has clinched passage through the group). I do think we can snag a 0-0 or 1-1 if things break well for us. Either way, a L and a D in the first two games is probably lights out. We're not going to beat Ghana by three or four goals and get through on four points and Italy and the Czechs are too savvy/cynical to allow what happened to Portugal against SK in 2002. If both need a draw to advance, that game will be a draw. Heck, half the starters in that game play for a club team that specializes in preordained results.
Re: World Cup Simulation Results I don't see how this is possible. The cumulative chance of winning for the bottom 11 (approximately the bottom third) teams is less than 1.5%. Even if you grossly oversetimate their chances, it's still not enough to significantly affect the win percentages of the top teams.
Re: World Cup Simulation Results I read the post, and he didn't expand on that point, and I made no assumptions.
Re: World Cup Simulation Results I think the point is not only are favorites (the Portugals and Spains of 2002) not "locks" it looks like, on average, the top favorite in a group is roughly 80-85% to advance. That means that on average, one would expect between 1 and 2 clear favorites to not advance.
Re: World Cup Simulation Results You're right, and it's pretty common for group favorites not to advance. If you look simply at probability, with two clear co-favorites each at 70% to advance, the probability of both advancing is less than 50%. *Spain did advance in 2002, actually, but France and Argentina, two strong favorites, did not.
Re: World Cup Simulation Results Ok, "someone didn't understand the post." [voros] "I keep track of international team strengths from time to time with a mathematical system I developed for use in a different sport. It does work just as well (if not better) in soccer and it did well enough to win rec.sport.soccer's Euro 2004 prediction contest.
Re: World Cup Simulation Results I also think it's important to note that any analytical framework has its drawbacks, and that I think in particular this system underrates European teams' chances and overrates non-European teams' chances in this particular World Cup, just because the Cup is being played in Europe and teams from Europe will probably enjoy some degree of home field advantage (in terms of likely greater fan support, less travel and more familiar surroundings, in general). I know that in baseball home teams generally enjoy at 55/45 advantage in games, and something close to that probably holds in most sports. Voros, for future versions of this system perhaps you could take into account the continent that games are played in and factor that in. I'd imagine travel and other similar concerns are a much greater factor in international soccer than college baseball.