World Cup Simulation Results [Rs]

Discussion in 'USA Men: News & Analysis' started by voros, Jun 1, 2006.

  1. superdave

    superdave Member+

    Jul 14, 1999
    VB, VA
    Club:
    DC United
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    Tim, stories pop up from time to time of tribal affiliation affecting roster selection.
     
  2. KenC

    KenC Member+

    Jun 11, 2003
    Re: World Cup Simulation Results

    Very cool and interesting. Not knowing your relative rankings of Germany and the US, but does this imply that if the US were hosting the WC, they'd have a pretty decent chance of winning the tourney? Germany is rated 11th best and have the 3rd best chance of winning. How big a jump would the US have if it were hosting?
     
  3. giffenbone

    giffenbone Member

    Jan 22, 2006
    Raleigh, NC
    Re: World Cup Simulation Results

    good question.

    voros could you possibly give us that info?
     
  4. HoustonSoccer

    HoustonSoccer New Member

    Mar 11, 2006
    Houston, Texas
    Club:
    Houston Dynamo
    Re: World Cup Simulation Results

    Just curious what are the parameters you are using in your formula. And 10,000 simulations what is that number actually mean?

    Did your program predict that Greece was going to win the Euro 2004 contest?
     
  5. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    A simulation is a computer model of the entire World Cup tournament, IE the final scores from all of the 64 games. A re-did this 10,000 times to get ballpark estimates of teams' chances for various results.

    And of course the system didn't pick Greece to win Euro 2004, any system that did would have been wrong, despite the result.

    Checking an old file it looks like the system predicted Greece with a 0.63% chance of winning the tournament. The team with the highest chances of reaching the final was Portugal and they did. The team with the second highest chances of reaching the semifinals (besides Portugal) were the Dutch, and they did as well.

    The end result being that the system corrected itself some afterward, but not nearly as much as other systems like ELO, and that proved very wise since Greece's subsequent WC qualifying campaign was not inspiring. By the way, those are my two biggest gripes with ELO: overrating of weak schedule teams, and too large of a shift in ratings based on a small number of results.

    Anyway the big advantage to computer systems as opposed to aggregating the subjective opinions of experts is that a human being tends to think in percentages of 0% and 100%. To a human, when something happens they tend to assume it _had_ to happen, and when it doesn't happen, they tend to assume there was no chance (though obviously the extent these assumptions are made depends on the event). A computer more accurately deals with possibilities outside of the actual results that happened.
     
  6. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    Gimme a bit.

    Remember that the Germans advantage is not just home field, but also receiving a seed due to hosting. I'd have to re-draw the teams somehow. Maybe I could just switch us with Mexico to be simplistic and still match the way teams are drawn.
     
  7. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    When people say this I'm always stunned that I'm as poor as I am. :)

    Soccernethost's post is 99% correct. The only thing he got wrong is that I no longer work for any MLB team, for the time being anyway. Such is baseball.
     
  8. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    Two things happened to Africa this time around. One, the system says that the majority of the best African teams didn't qualify. Of the top 5 African teams according to the system, only two qualified: Ivory Coast and Tunisia. Also the team the system says is best, Cameroon, just missed qualification.

    Two, Africa's best team in the tournament got blindsided by drawing the Dutch and the Argies in the same group. They can still make it out, but I'd like their chances much better in any other African team's draw except Ghana's.
     
  9. bigbrooklynlou

    bigbrooklynlou New Member

    Oct 7, 2004
    Brooklyn, NY
    Re: World Cup Simulation Results

    Poor in money but rich in admiration. Remember "Its a Wonderful Life". Just because you dont own the town doesnt mean the town doesnt care.

    Send your resume to Omar Minaya and the Mets. Put in your cover letter "I, Voros Mccracken, would never have recommended the trading of Kazmir". My Metsies could use someone with common sense. :D
     
  10. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    They're doing fine right now, though I suspect the season will probably get a little tougher as it goes on.
     
  11. Soccernethost

    Soccernethost New Member

    Apr 16, 1999
    Re: World Cup Simulation Results

    Sorry to hear that V.

    Which still doesn't mean I can actually DO a rating system. Though I may have a go at one for fun if I can pick up a few extra bits of SAS know how.
     
  12. sidefootsitter

    sidefootsitter Member+

    Oct 14, 2004
    Re: World Cup Simulation Results

    Lemme ask you this ... is it possible to drop the US rating to ... eh, 20-22nd and raise Ghana's to somewhere around 40th and recalculate the group?
     
  13. Soccernethost

    Soccernethost New Member

    Apr 16, 1999
    Re: World Cup Simulation Results

    lol...yeah Voros...

    Can you rig your mathematical rating system to so that it spits out pre-determined results?

    While you're at it...I'd like to see one where the US is the favorite to win the Cup.

    Sorry, SFS, but this couldn't be more in keeping with your online persona. If the observable facts don't match your subjective view of the world, just change the facts!
     
  14. sidefootsitter

    sidefootsitter Member+

    Oct 14, 2004
    Re: World Cup Simulation Results

    Uh, Dave, the changing of calculations was to reflect the systematic error in Elo rating as noticed by Voros (and me as well long time ago, which is why I brought up the old Brigham Young football rankings on these boards -- BY dominated a weaker conference, going 10/11-0, and received the #1 national ranking despite not beating anyone worth of note).

    But nice of you to launch another warrantless attack. Doing anything otherwise would be so out of character.
     
  15. Soccernethost

    Soccernethost New Member

    Apr 16, 1999
    Re: World Cup Simulation Results

    Leaving aside I have no idea who Dave is...

    Voros' system is his own. Not ELOs.
     
  16. redtyre

    redtyre New Member

    Apr 18, 2004
    Re: World Cup Simulation Results

    I did not want to start this own thread for such a topic, I believe at the Hour, 1 pm est, 12 cst, 11 mst, etc. Univision will conclude their predictions from last week, last week, groups A-D were done, I think, here shortly, they will review group E etc. They have a panel of six who comment on the predicaments.

    Last week, I figured a slam dunk would predict Argentina to advance, but not so; it was close to even between Argentina or Serbia in advancing.

    So, assuming this show concludes at this hour, it will be interesting.


    I'm trying to tape it though, cause it's 2 hours long to predict the 4 groups and they have a lot of extra stuff, like showing the stadiums and stuff; so just letting you be informed.
     
  17. sidefootsitter

    sidefootsitter Member+

    Oct 14, 2004
    Re: World Cup Simulation Results

    I was going to start a thread ... but ...

    their panel was Jesus Bracamonte, Jose Luis Chilavert, Fernando "El Presidente" Fiore, Jorge Perez Navarro and two other dudes.

    The US was 3:3 to advance. (3 votes for, 3 against)

    Czechia was 5:1.

    Italy was 4:2.

    Ghana was 0:6.
     
  18. HoustonSoccer

    HoustonSoccer New Member

    Mar 11, 2006
    Houston, Texas
    Club:
    Houston Dynamo
    Re: World Cup Simulation Results

    What I would like to see a more realistic simulation of each team’s potential, since past team scores may not represent the make up of the current team. Teams go through multiple changes from each WC to other. So using results more than two years old may be flawed. Most teams that are playing in the current WC do not accurately reflect even their qualifying games in the last two years.

    We know in American Football, they use extensive scouting of the current roster and new recruits. We need to find a way of using all such factors that have an influence on the game.

    I recall very clearly, the near dominating performance of Argentina prior to WC 2002, yet bowed out after the first round. Most folks failed to notice the weaknesses in their defense that was always there.

    I think some of the concepts in your simulation pickup the importance of defense. As the old saying goes - "Offence wins games, and defense wins championships." I believe most folks overlooked Greece's defense in Euro 2004.
    So, that leads to questioning Brazil's potential in this WC. As great an attacking team they have - are their defenders equal to their attackers. England potentially faces the same problem.

    And watching the WC 2002 finals, Brazil's keeper made some fantastic saves that kept them in the game. Despite Ronaldo’s winning goals, the Keepers heroics should not be overlooked. And then Germany would not be in the finals if not for the missed the handball save, by the referees, on the Goal line against USA. And the 2002 US team was far weaker then the current US WC team.

    Officiating plays a very big role in soccer games. Simulations in order to be accurate should include a provision for officiating, but then we never know who is going to officiate a game. And we have numerous examples of bad calls in WC games.

    I believe USA has one the best defensive teams in the tournament and that gives them much higher chance then your simulation has so far predicted. And let us not forget USA always played Brazil tough on defense even in 1994.
     
  19. Soccernethost

    Soccernethost New Member

    Apr 16, 1999
    Re: World Cup Simulation Results

    All sounds very subjective to me, Houston. Based on faulty observation and subject to the same sort of "get the answer I want" sort of thinking as SFS so ably uses.

    Kind of like blaming Argentina's defense for their problems in 2002. They gave up two goals in three games, 1,1,0.
     
  20. sidefootsitter

    sidefootsitter Member+

    Oct 14, 2004
    Re: World Cup Simulation Results

    Way to bring me into this, nitwit.
     
  21. sardonic

    sardonic New Member

    May 28, 2006
    San Diego
    Re: World Cup Simulation Results


    I think that the point is that this simulation and other results oriented analytical frameworks are not meant to replace our subjective analysis of the game. Rather, they are meant to give a different perspective that subjectively would be hard to achieve. Whenever looking at something like this, you have to ask yourself, "What is this telling me?" and then "How does it know that?"

    Basically, I think this sim is saying, teams with results like the ones in the WC typically do this well against teams with their results. Ie. teams with the US's results tend to advance around 50% of the time against teams with the CR, Italy and Ghana's results.

    How does it know that? It looks at the team's results over the past 4 years. It takes into account the location of the match. It weighs more recent results more heavily than previous results. It takes into account the quality of the side each team faces.

    I can think of reasons why it would be off for the US, and why it would be accurate. For one thing, it penalizes the US for playing at home a lot, without "knowing" that the US's home games are often like other teams' road games. On the other hand, the US talent pool/leadership has remained relatively stable through the past 4 years, meaning that the results achieved over the time period are probably fairly representative of this group of players and coaches. Having a results-oriented framework doesn't mean that this stuff doesn't matter or isn't important, it just means that those things are very difficult to model in an accurate way -- a model wouldn't be very good at quantifying how those things impact a team's chances.

    Similarly, the model doesn't know that Nedved wasn't around for many of the results in the past two years, but has decided to rejoin the team and is playing well, or doesn't know that Italy is facing a scandal right now, nor should it.

    Basically, I think the model does an excellent job of giving us an idea of how teams can be expected to fare in the Cup given their results, and helps to put their results into perspective by taking into account the quality of opposition in a more comprehensive and objective way than possible mentally. It's a great starting point for analyzing teams' chances in the WC. But it doesn't mean it should try to model everything, because there are many things that can't be modeled. And it certainly doesn't mean that we should check our subjective observations at the door either; just to understand when to use which mode of analysis.
     
  22. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    Re: World Cup Simulation Results

    Well for club teams anyway, I found that three years worth of data is better than two in predicting future results. This suggests several things:

    1. Since National Teams play significantly fewer games per year than club teams, National Teams should go back at least three years for sample size considerations alone.

    2. Player movement on National Teams is obviously far less than on your typical club team. Again suggesting National Team data can go back farther than club team data.

    3. The mathematics of the system breakdown if you only go back two years because you don't have enough overlapping matches.

    4. Even going back as far as eight years, there are strong relationships between the quality of teams then and the quality of teams now. If you look at the top 10 now and the top 10 in 1998, you'll find at least half of those teams are the same teams.
     
  23. Maximum Optimal

    Maximum Optimal Member+

    Jul 10, 2001
    Re: World Cup Simulation Results

    Love you work on both baseball and soccer Voros. Since you have defensive and offensive ratings of each team, I was wondering if you would be kind enough to give the rankings by offensive and defense. Also how has the US's offensive and defensive rating been evolving. My subjective impression is that we do not have the offensive firepower from four years ago, but that our defense is much stronger.
     
  24. HoustonSoccer

    HoustonSoccer New Member

    Mar 11, 2006
    Houston, Texas
    Club:
    Houston Dynamo
    Re: World Cup Simulation Results

    Well put, maybe I am pushing the envelope by extending the analytical approach from its current stance. If not for any thing else but to help us develop our soccer program in this country. And it is a good tool in its current form.

    Look how in-depth we get in Football, Baseball and Basketball. We scout the kids from Jr. High, and an array of statistics are developed, which allows a level of evaluation that is currently missing in soccer.

    We do not evaluate football teams on the basis of three years of game scores, rather the make-up of the current team and the coach. And coaches we know have a strong influence on the outcome.

    Current soccer analysis is quite rudimentary. Any tool we can develop to better our game is not a bad idea.
     
  25. HoustonSoccer

    HoustonSoccer New Member

    Mar 11, 2006
    Houston, Texas
    Club:
    Houston Dynamo
    Re: World Cup Simulation Results

    I was watching the Houston Dynamos against LA Galaxy last night. Both teams are missing key players because of WC duty. And the performance is significantly affected. Yet, if the teams were analyzed solely on the game results it would be inaccurate.

    I am just saying that the analysis if extended beyond just game results would provide a more accurate picture. Similar analyses if performed in the other sports like Basketball, Football or Baseball would not look good.

    Maybe, we can do what we have in College Football - a poll by coaches and another by Sports writers.
     

Share This Page