Re: World Cup Simulation Results [voros] "When the Dutch play the French in France, France's advantage tends to be the same as when the Russians or Australians or whomever plays in France." Would at least suggestthat the effects of distance traveled are overrated.
Re: World Cup Simulation Results Could at least one (but preferably all) of you people posting %s to win the whole thing and advance just show how you are "simulating" the matches? And not even all of them, I'd just like to see an example of how you do the simulation for a single match. Like for example, maybe post how you (or your system) are determining the most likely outcome of, say, USA vs Italy (or any other game if you'd prefer to show something else, like Holland vs Argentina or even Tunisia vs Saudi Arabia). Could anyone do this, please? Thanks!
Re: World Cup Simulation Results I'd imagine that he comes up with a "true overall winning percentage" that approximates the actual ability level of a team relative to all other teams, and then uses the log5 method, a means of estimating the chance of one team beating another team given their winning percentages, a method originally developed for baseball.
Re: World Cup Simulation Results I was speaking more along the lines of individual games and also the odds to advance.
Re: World Cup Simulation Results For mine, the rating system is such that it predicts the expected scoreline of a match between two teams. The poisson distribution is then used in the simulation to determine the goals scored of every simulated match. Runs scored in baseball do not match the poisson distribution well, but in soccer it's, for the most part, pretty close.
Re: World Cup Simulation Results I don't think so because the circumstances under which the travel was undertaken change significantly from game to game. Sometimes a team has a day to make the trip, sometimes it takes a week. Furthermore the travel effects change from player to player depending on where they may be playing their club ball. Again I'm not doubting the potential effect, I'm saying that because of the differences in the various situations, I can't accurately capture any effect mathematically.
Re: World Cup Simulation Results There's definite home field adjustments and they're part of the system. For example a team that plays an abnormal percentage of its games at home (like us) suffers in the rankings accordingly. But for the actual sim, only Germany was given those advantages. I have the ability to adjust for half-home field type of advantages, but what I don't have is the evidence that I ought to, nor do I have any reasonable guideline as to when I should and shouldn't apply such advantages. When Mexico plays another team in the U.S., should I apply that advantage? It turns the home field status of every potential matchup into guesswork, and then when you consider the system as it currently exists holds almost 6,700 matches, it's impractical as well.
Re: World Cup Simulation Results The thing is that such an assessment falls under the category of "multiple endpoints". If you assume each of the eight top favored teams in each group had a 75% chance of advancing, the chances that three or more would not advance is a healthy 32%. That those three would specifically be Argentina, France and Portugal wasn't known until after the fact. So you can't separate their chances out from the chances of Spain, Brazil, Germany, Italy and Japan, all of whom did advance. It is far more likely that one of the group favorites will _not_ advance to the second round than it is that they will all advance. We just don't know which one it might be right now.
Re: World Cup Simulation Results I see you are catching up to earlier posts. I would appreciate your input on my group points question (9-6-3-0; 9-3-3-3, etc.). Is this something you can and would be willing to tackle? Or can you help me get started, at least, or perhaps direct me to someone who can? Many thanks!
Re: World Cup Simulation Results Your analytical grasp and your clarity in expressing yourself is truly first-rate. If I could rep you again I would. But I can't. Can I offer you a job?
Re: World Cup Simulation Results I have performed this calculation for Group E (using a similar method to Voros's). Results are here: https://www.bigsoccer.com/forum/showpost.php?p=8396369&postcount=31
Re: World Cup Simulation Results You made assumptions about how it works, I didn't. All rankings use a mathematical system, so that tells us nothing. You might have to seed the simulation with the relative strengths of the teams, otherwise you start from scratch and there are not enough games between teams within a short time period to start from scratch. The teams turnover players every year, as well, making things more difficult, as teams are clearly far from static quanitities. So that why I pondered if/how he seeded the teams, and whether he used a ranking system such as FIFA's (or ELO or the myriad others) to do so. Additionally one can do linear regression against rankings and your own formuals in a game-by-game basis and correct for bias. That he uses a mathematical system he devised and based on game outcomes does not preclude the use of any sort of seeding data or real-time comparator. I've written a similar program for the NCAA basketball tournament, and I incorporated the RPI and coaches polls for both seedings and as comparators.
Re: World Cup Simulation Results Thanks! I'll spend some time looking over your results, and see if I have any comments. I am also interested in a more abstract question about points and probabilities, especially as regards likely tie-breakers (goal-difference, etc.) I'll illustrate, to make things clearer. There are six games in a group of four round-robin. If all six games are ties/draws, all teams have three points [3-3-3-3 = 12], all are equal head to head, all have a zero GD, so we look at total goals scored. That's why 2-2 is better than 1-1 is better than 0-0. If all but one game is tied (i.e. 5 ties), add two to the winner and subtract one from the loser, and you get 5-3-3-2 = 13. Two teams are tied for second, equal head to head, both with a zero GD, so again we look at total goals scored. If two games are tied, ... At the other end, if no games are tied, you would seem to have four basic patterns, 9-6-3-0 (with a clear 1 & 2); 9-3-3-3; 6-6-6-0; and 6-6-3-3 = 18, with the last three requiring an analysis of tie-breakers. I'm hoping someone has (or quickly and easily can) run all the various combinations of results. In the abstract, how good are your chances with 3, 4, 5, 6 points, especially 4 and 5? And what percentage of the total number of results include a 'standings tie' for first or second place that requires the application of a tie-breaking formula? Anyone?
Re: World Cup Simulation Results So what does this all mean? Are we going to win or what? This math crap is too confusing. Can't we just play the games and see who wins? Actually, pretty awesome stuff. But I always lose in Vegas, even playing banker in bacarrat.
Re: World Cup Simulation Results I understand it. Basically, if your team gets 0 pts - 0% chance of advancing 1 pt - 0% chance of advancing 2 pts - 1.23% chance of advancing 3 pts - 7.87% chance of advancing 4 pts - 54.32% chance of advancing 5 pts - 98.77% chance of advancing 6 pts - 97.53% chance of advancing 7 pts - 100% chance of advancing 9 pts - 100% chance of advancing of course this does not take into account the fact that getting 4 points by beating the Italians, tying the Czechs and losing to Ghana is much different than getting 4 points by beating Ghana, losing to the Czechs and tying the Italians just because of the relative strengths of the teams.
Re: World Cup Simulation Results Hope it's ok that I'm posting this... Voros works for a MLB team (or has done consulting for them) and gets a fairly serious mention in Moneyball for developing a baseball-related metric that's proved very useful (IIRC it's that the only control a pitcher has once a player swings the bat is making him miss (i.e. stike him out) and keeping the ball in the park (i.e. prevent home runs) -- any other ball put in play has a mostly random outcome). And that throwaway line he had about winning the predictions contest for Euro 2004 is kind of a big deal. There were a LOT of entries from the world's soccer stat heads. I'll get some of the details wrong, but hopefully close enough to make sense... The basic idea of Voros' system is that each team gets a rating for its offense and defense based on prior results. Those two scores are adjusted to take into account the quality of that team's opposition, and then the opposition's quality is adjusted to reflect the adjusted quality of the team. Repeat repeat repeat until the scores stop moving in a meaningful way. Once you have a stable offense and defense score for every team you can pair any two teams and adjust their scores to reflect that they're playing each other. That gives you a number for both teams equal to their chances of scoring in any given minute (or whatever timeframe you decide is the shortest likely amount of time a team could score twice). To simulate a result between those two teams you use a Poisson distribution to determine how many goals are scored by each team. Run that 10,000 times and you get a very good idea of how two teams will fare against each other. He's got some sort of adjustment in their to take into account home field advantage and also has some way of weighting for the meaninfulness of the competition (friendlies less than non-WC tournament, less than WCQ, less than WC I think). There could be other tweaks, but I think that's the gist of it. At least in very broad terms -- again, I'm sure I've got some of it wrong.
Re: World Cup Simulation Results So Voros is THE Voros Mccracken? The guy that did the work on pitchers and defense stats? Wow. Nice to know that the sabermetric crowd is turning an eye towards soccer - god knows the sport needs it.
Re: World Cup Simulation Results Thanks everyone, including 'Dr. Gamera,' of course. Note first, what I (as a non-stats guy) would call an anomaly: moving from five points to six slightly 'decreases' your chances of advancing. Can that be correct, or is there a hidden error, or a rounding error, etc.? What's the explanation? Note also that almost half of the scenarios come down to the tie-breaking formula, either for 1st (critical in our group especially) or second. Next, the basic pattern is very intuitive: 9 and 7 = yes, 0 and 1 = no, of course; 6 and 5, almost certainly yes, 2 and 3 almost certainly no; 4 puts you almost right on the edge in between. And, yes, in the real world, it will very likely matter vs. which team you are earning/dropping points. I think the timing and order in which the games are played is essential information. This is only the first part of what I'm looking for, unfortunately. I'll be back when I've worked out what else I would like to know.
Re: World Cup Simulation Results 5 points means that you've denied all three teams points, and didn't lose to any of them. 6 means you lost to someone. So I think it probably makes sense. Another way to think about it is that 2 wins and a loss means you took 6 of the 9 points awarded from your three games (67%). A win and two ties means you took 5 of the 7 points awarded from your games (71%).
Re: World Cup Simulation Results That makes a lot of sense, thanks. I'm still thinking - I guess worrying would be more accurate. Numbers like 5-5-5-0 keep popping into my head. Or 0-0, 0-0, 1-0 - fail to advance! That's why I hate to keep reading BS posters advising, "Safety first in Games 1 & 2, then beat Ghana and we're through!" [Hey, that rhymes! Can I copyright that?] How about, "Let's score (clap clap clap) one more (clap clap clap)!" [Sorry.] I put Italy in the driver's seat, not only because they are the 'best' team, but because they will know the result of Game 1A (US-CZ) when they play Game 1B (IT-GH); and they'll know the result of 2A (CZ-GH) when they play us in 2B (US-IT). [I see some of you shaking your head and rolling your eyes, but bear with me.] We'll also have a lot of critical information when we meet Italy, and I would like to know in advance how to take advantage of that information. If you start with 729 possible points combinations before Game 1A, you'll have a much narrower range by the time you get to Game 2B (knowing the results of the first three games). Does this (sort of analysis) interest anyone else? If not, I'll go quietly.
Re: World Cup Simulation Results Count the number of African teams that have been in the world cup in the since 1990 inclusive (don't really know the answer there) and then count the number of African teams to make the quarterfinals (Cameroon 1990 and Senegal 2002, right? I think that's it.) And of course, no African team has ever made the semifinals. You'll see it makes sense. Africa is a continent that has tons of potential, but at this point, the smart bet is to not bet on them till they prove it.
Re: World Cup Simulation Results I don't bet, but you're right. Especially with four of five African teams at the finals for the first time, and the best (i.e. highest-ranked) one (IC) in the Group of Death. I see a number of factors working against the African teams, some more 'correctable' than others. (1) Talent drain. The two-edged sword of exporting players to Europe. Great for individual player development; not so good for team cohesion, or at the very least, 'familiarity.' Brazil and Argentina can overcome this problem, but that is not to say it's not a genuine issue elsewhere. (2) Poor management. Just picking up on bits a pieces from other threads, you see it's very common for these teams to be badly managed: (a) as to finances (i) poorer accommodations, training facilities, etc. (ii) players not paid, payment not timely, bonuses not paid. (b) as to player selection (occasional corruption)* (i) a 'rich' player will buy his way onto the team/into the first team. (ii) a President will interfere with player selection. (Imagine President Bush calling up Bruce Arena and telling Bruce whom to pick/play!) (3) Lack of Experience. You get better by playing better teams. I assume that not many teams (to this point) have been willing to leave Europe or South America to play in Africa. Maybe that will change for 2010. Other factors must enter in to the equation. Not being seeded, rightly or wrongly, is a clear disadvantage (not just for African teams). *No, I don't have evidence I can bring into court. These are ideas I've gathered right here on BS from (African?) posters who sound credible. Use your own judgement - doesn't sound impossible, certainly. Just look at the recent scandals in the sport elsewhere (e.g., Italy, Germany, Brazil).