Re: World Cup Simulation Results There are at least two things to look at: 1. Statistical models (e.g., Voros) 2. Historical models (e.g., prior WC results) The latter are important, at least in part, I think, because the teams playing their third and last game often know what result they need to achieve. The results have to be viewed in that context. [Of course we 'knew' we needed a draw v. Poland in 2002 and still lost.] Germany only had seven points: 2002 Group E GERMANY 7 REP.IRELAND 5 Cameroon 4 Saudi Arabia 0 In 2002, South Africa (B), Costa Rica (C), Cameroon (E), and Argentina (F) failed to advance with 4 points. Additionally, with very slight adjustment to the final round games in each group, several other teams might have gone out with four points. Not a large sample, but four points is much too close (to elimination) for comfort.
Re: World Cup Simulation Results 1. Um, but with very slight adjustments I'll bet zero teams with 4 points would have missed out. 2. Yeah, there's an understatement (not a large sample.) Essentially, I think, there's 3 ways a team can have 4 points. 1. 7-5-4-0. SOL. 2. 7-4-3-3. You're in. 3. 9-4-4-0. Goal Difference. Yes, there's 4-4-4-4, but that's gotta be very, very, very rare.
Re: World Cup Simulation Results There's also 4. 5-4-4-3 Which is what would have happened had South Korea and Portugal drew. What is interesting however, is that the tiebreakers have changed this time around, and using the new tiebreakers we would have actually advanced against Portugal under that scenario since we beat them head to head.
Re: World Cup Simulation Results In the US's group in 2002, had Portugal and SK tied, the group would have been 5-4-4-3 and the US would have been out on GD.
Re: World Cup Simulation Results Damn, Voros beat me by about 12 seconds on that. As an aside, can someone confirm that the new HTH tiebreakers are in effect in group play? I thought those were just in qualifying and GD, etc. was still in use in the Cup? Using a HTH result seems ridiculous in a three-game preliminary round -- it's way too much of an advantage for the team that wins that match and doesn't necessarily mean that team is the "better" team.
Re: World Cup Simulation Results I checked FIFA's site as I was making the sim model and it seemed to confirm that HTH was indeed the 1st tiebreaker now.
Re: World Cup Simulation Results The statistical models are all historical models. The best statistical models are the best historical models. If you want to add something to the statistical models, you have to provide something they can't incorporate, which is reasoning based on very recent events that haven't or may not have yet made it into any statistical models in a meaningful way. For example, it might be appropriate to discount the strength that Voros's model gives to the French and the Czechs based on an argument or maybe just a suspicion that their key players (e.g., Zidane, Nedved) are now just a little too old to make their teams quite as good as the results over the last several years would suggest they are. I'm not saying that would be right, but it seems like a reasonable possibility.
Re: World Cup Simulation Results Which may have been an anomaly, one of those things that happen in a short tournament where one can score three goals in 30 minutes, which affects the five game total. (Heck, Germany scored eight in their first game, then only six more over their next six). In any case, going back to 1998. Group A : Morocco gets 4 and is eliminated (6-5-4-1 distribution) Group C - Denmark is through with 4 (9-4-2-1) Group D - Spain is out with 4 (6-5-4-1) Of three teams with 4 points, two were knocked out.
Re: World Cup Simulation Results One would think that HTH would be a much better indicator of the "better" team then common results vs two other opponents.
Re: World Cup Simulation Results First of all, 1994, Group E MEXICO 3 1 1 1 3-3 4 REP. IRL. 3 1 1 1 2-2 4 ITALY 3 1 1 1 2-2 4 Norway 3 1 1 1 1-1 4 So, yes it's rare, but not so very rare or unthinkable. You have also oversimplified the patterns, but that's ok. You can change your results to fit your theory, if you want to. In addition to the four 4-point teams actually eliminated, South Africa (B), Costa Rica (C), Cameroon (E), and Argentina (F) I've got: 2002 Round three games --- GROUP A France v Senegal 0-1 Uruguay v Denmark 1-2 France v Uruguay 0-0 Denmark v Senegal 1-1 Denmark v France 2-0 Senegal v Uruguay 3-3 DENMARK 3 2 1 0 5-2 7 SENEGAL 3 1 2 0 5-4 5 Uruguay 3 0 2 1 4-5 2 France 3 0 1 2 0-3 1 France beats Denmark and/or Uruguay beats (not ties) Senegal and two (or all four!) teams finish with four points. We have already looked at Group D - Portugal could easily have tied S Korea and tied us at 4 points (w/ a better GD). [Yes, it's now HTH.] GROUP F England v Sweden 1-1 Argentina v Nigeria 1-0 Sweden v Nigeria 2-1 Argentina v England 0-1 Sweden v Argentina 1-1 Nigeria v England 0-0 SWEDEN 3 1 2 0 4-3 5 ENGLAND 3 1 2 0 2-1 5 Argentina 3 1 1 1 2-2 4 Nigeria 3 0 1 2 1-3 1 If Nigeria happened to beat England, then you would have had England and Argentina at 4 points, with one of them to be eliminated. Yes, still just one 4-point team eliminated. GROUP G Croatia v Mexico 0-1 Italy v Ecuador 2-0 Italy v Croatia 1-2 Mexico v Ecuador 2-1 Mexico v Italy 1-1 Ecuador v Croatia 1-0 MEXICO 3 2 1 0 4-2 7 ITALY 3 1 1 1 4-3 4 Croatia 3 1 0 2 2-3 3 Ecuador 3 1 0 2 2-4 3 If Croatia had tied Ecuador, Croatia and Italy would have tied at 4. Had Croatia beaten Ecuador, Italy would have been eliminated with 4. GROUP H Japan v Belgium 2-2 Russia v Tunisia 2-0 Japan v Russia 1-0 Tunisia v Belgium 1-1 Tunisia v Japan 0-2 Belgium v Russia 3-2 JAPAN 3 2 1 0 5-2 7 BELGIUM 3 1 2 0 6-5 5 Russia 3 1 0 2 4-4 3 Tunisia 3 0 1 2 1-5 1 If Tunisia had beaten Japan, both would have finished with 4, with one eliminated (presumably). If Russia had tied Belgium, the Russians would also have made it to 4. [Of all these hypothetical results, only Tunisia turning the tables on Japan seems like a real stretch, although the French would have had to have played much better than they actually did.] This tells me a lot more than, "Well, duh, those last games are important. They're all important." It tells me that 4 points is far from a guarantee, and will often leave you depending on help (in the form of a helpful result) from someone else. I don't understand the love affair with four points, whatever the statistical model shows. I think if you lose Game 1, you have to go for the win in both Games 2 and 3.
Re: World Cup Simulation Results Just saw the 1998 numbers, where I was headed next. Morocco and Spain out with four; Spain through with four - not a good ratio! 1994 has to be evaluated differently, with 6 x 4 = 24 teams and 4 of 6 3rd place teams advancing (including US & Italy w/ 4 points). Interestingly, two of the six groups finished 6-6-6-0.
Re: World Cup Simulation Results As I read the official competition regulations, the first tie-breaker is NOT head-to-head. First tie-breaker is goal difference, then goals scored, and then head-to-head. http://www.fifa.com/documents/static/regulations/FWC06_regulations_EN.pdf, Aritcle 31, paragraph 5.
Re: World Cup Simulation Results Not if the context is judging which team in a four-team group should advance in first or second place based on its results over three games. Why shouldn't performance against the other two teams count as much if that's the evaluation metric?
Re: World Cup Simulation Results Maybe this has been linked already, but I've already seen two simulations with different winners; so....it's just that; http://www.psxextreme.com/scripts/news2/new.asp?newID=3768
Re: World Cup Simulation Results Using computers and mathematical formulas is not what I am contesting. It is the addition of parameters that address team strength. Not aware of what factors you use in your formula - I am contesting that just by win-loss or goal differences it does not address a team's potential. I agree with you on friendlies - they should not be counted at all. Friendlies seldom reflect team strength.
Re: World Cup Simulation Results You should say what you mean -- that it's pointless to try to analyze the strength of teams between confederations (because there are too few non-friendlies to get anything useful if you throw them out). I think Voros already gave some pretty good anecdotal evidence to the contrary.
Re: World Cup Simulation Results Can you show me some evidence that the results of friendlies are not at all related to team strength? HINT: just say "no." Because they are related. Not as strongly as the results of meaningful competitions, but they are related. Also...just what portion of a team's "strength" do you find is NOT reflected in its results?
Re: World Cup Simulation Results No, he's right. Jamaica and New Zealand are perfect recent examples, where two soccer powerhouses got blown out in friendlies by no-names England and Brazil. Happens all the time like that, for example the Americans can never beat Brazil in friendlies, now how stupid is that? LOL.
Re: World Cup Simulation Results voros- What are the offense and defense rankings for the other teams in Group E? I'd expect the Czechs to have a high offense and lower defensive rank. thx
Re: World Cup Simulation Results Here's the simplest web page ever with the offense and defense rankings for all 205 official FIFA teams (as of May 31): http://members.cox.net/voros/ratings.html For a while I had added Micronesia to this list because, amazingly, Guam had never scored on anybody else (which fouled the system). But on March 9, 2005, with a scene reminiscent of the "Miracle on Ice", Guam lost to Mongolia by a score of 4-1, thereby making the system work again. They haven't scorted since, but I have seven more years before I have to worry about it again.
Re: World Cup Simulation Results Also remember a lot of other factors besides personnel go into whether a team is unbalanced in favor of the attack or defense: style of play, average field and weather conditions, officiating, etc. So though the U.S.A. has a better defensive rating than some top teams (like the Czechs), don't read into that we necessarily have better defenders than those teams.
Re: World Cup Simulation Results John you just short-circuited my sarcasm detector. When I examined this question as I made the rankings, I ran the rankings once using only competitive matches, than ran them again using only friendlies (and I counted the Confed Cup and other minor tournaments as friendlies). There was a fairly high correlation between the two sets of rankings, about 0.8 'r-squared'. So there is a clear relationship between friendly results and competitive results. So that, plus the massive sample size advantages being able to use friendlies gives you, makes me favor using them at a lesser weight than other matches.
Re: World Cup Simulation Results You must spread some Reputation around before giving it to voros again.
Re: World Cup Simulation Results Voros, you say you give the friendlies a "lesser weight" - how do you determine how much of a weight to give them?