The John De Lange Sophisticated Prediction Contest - WC 2010
in [World Football]
|
From: jvazquez on 12 Jun 2010 21:09 On 12 jun, 17:51, Futbolmetrix <futbolmet...(a)yahoo.com> wrote: > That's because unfortunately there was a major bug in the spreadsheet, > which almost undermines the whole sophisticatedness of the SPC. There > should have been two additional restrictions: the sum of QF exit > probabilities in groups A-D and E-H should have been equal to 2, and > the sum of (semi + LF + WF) in each half of the draw should also have > been equal to 2. Check, Check. JV
From: Philipp Lucas on 13 Jun 2010 01:49 Futbolmetrix <futbolmetrix(a)yahoo.com> wrote: >That's because unfortunately there was a major bug in the spreadsheet, >which almost undermines the whole sophisticatedness of the SPC. There >should have been two additional restrictions: the sum of QF exit >probabilities in groups A-D and E-H should have been equal to 2, and >the sum of (semi + LF + WF) in each half of the draw should also have >been equal to 2. So, when it comes to see my predictions crumbling in the face of reality, I can blame it all onto missing error checking in the Excel sheet which lured me into making wrong predictions? Great! (My predictions are unbalanced in both cases.) -- Philipp Lucas phlucas(a)f-m.fm
From: milivella on 13 Jun 2010 09:19 juanvazquez: > How do you translate say ELO into the sheet? Do you have a "system"? > Can ELO (or any other simulation) be translated with different sets of > numbers? I hope that someone who knows (Daniele!) will reply you, but my guess is that Elo (or any similar system) can predict the outcome of any match. I.e. it can give you a table with the probability of each different outcome, something like England vs. USA 0-0 10% 1-0 20% 0-1 5% 1-1 15% etc. Then you can run (e.g. in Excel) thousand virtual WCs, computing the outcome in each match in the following way: the computer picks a random number between 1 and 100 and finds the outcome related to that number (in reality it's a bit less anthropomorphic than this!), e.g. England vs. USA if the random number is 3 -> then the outcome is 0-0 14 1-0 27 1-0 Now, if you run 10,000 simulations and Spain wins in 752 of them, you just assign ,0752 to Spain as winner in the sophisticated prediction contest! In general, I've seen that math is more empiric and inductive than most people think: you don't need (and probably can't have anyway) a _deductive_ system that takes Spain's Elo rating and convert it into chances to win the WC, since you can just run a lot of virtual WCs and _see_ how many times Spain win. (Or I'm completely wrong, and in this case I hope you've appreciated my piece of fiction.) -- Cheers milivella
From: Myk Cameron on 14 Jun 2010 07:06 "milivella" <milivella(a)gmail.com> wrote in message news:c2b9e0e7-1345-4dbe-a058-0b519a3c0b94(a)k39g2000yqb.googlegroups.com... juanvazquez: > How do you translate say ELO into the sheet? Do you have a "system"? > Can ELO (or any other simulation) be translated with different sets of > numbers? I hope that someone who knows (Daniele!) will reply you, but my guess is that Elo (or any similar system) can predict the outcome of any match. I.e. it can give you a table with the probability of each different outcome, something like England vs. USA 0-0 10% 1-0 20% 0-1 5% 1-1 15% etc. Then you can run (e.g. in Excel) thousand virtual WCs, computing the outcome in each match in the following way: the computer picks a random number between 1 and 100 and finds the outcome related to that number (in reality it's a bit less anthropomorphic than this!), e.g. England vs. USA if the random number is 3 -> then the outcome is 0-0 14 1-0 27 1-0 Now, if you run 10,000 simulations and Spain wins in 752 of them, you just assign ,0752 to Spain as winner in the sophisticated prediction contest! In general, I've seen that math is more empiric and inductive than most people think: you don't need (and probably can't have anyway) a _deductive_ system that takes Spain's Elo rating and convert it into chances to win the WC, since you can just run a lot of virtual WCs and _see_ how many times Spain win. (Or I'm completely wrong, and in this case I hope you've appreciated my piece of fiction.) -- Cheers milivella **** I didn't run simulations to convert the AQB ratings into probabilities of WC victory. I have a spreadsheet (across multiple sheets) that summarises the joint probability of every combined set of possible outcomes in the first round, and works out the probabilities that a team finishes first or second. I also have sheets that work out the probabilities of victory of every possible matchup from that point on, allowing an estimate of the probability constructed from the joint probabilities of every possible outcome. Hope that makes sense - it's kind of late here! Myk
From: milivella on 15 Jun 2010 17:35
Myk Cameron: > "milivella" <milive...(a)gmail.com> wrote in message > > news:c2b9e0e7-1345-4dbe-a058-0b519a3c0b94(a)k39g2000yqb.googlegroups.com... > juanvazquez: > > > How do you translate say ELO into the sheet? Do you have a "system"? > > Can ELO (or any other simulation) be translated with different sets of > > numbers? > > I hope that someone who knows (Daniele!) will reply you, but my guess > is that Elo (or any similar system) can predict the outcome of any > match. I.e. it can give you a table with the probability of each > different outcome, something like > England vs. USA > 0-0 10% > 1-0 20% > 0-1 5% > 1-1 15% > etc. > > Then you can run (e.g. in Excel) thousand virtual WCs, computing the > outcome in each match in the following way: the computer picks a > random number between 1 and 100 and finds the outcome related to that > number (in reality it's a bit less anthropomorphic than this!), e.g. > England vs. USA > if the random number is 3 -> then the outcome is 0-0 > 14 1-0 > 27 1-0 > > Now, if you run 10,000 simulations and Spain wins in 752 of them, you > just assign ,0752 to Spain as winner in the sophisticated prediction > contest! > > In general, I've seen that math is more empiric and inductive than > most people think: you don't need (and probably can't have anyway) a > _deductive_ system that takes Spain's Elo rating and convert it into > chances to win the WC, since you can just run a lot of virtual WCs and > _see_ how many times Spain win. > > (Or I'm completely wrong, and in this case I hope you've appreciated > my piece of fiction.) > > -- > Cheers > milivella > > **** > > I didn't run simulations to convert the AQB ratings into probabilities of WC > victory. I have a spreadsheet (across multiple sheets) that summarises the > joint probability of every combined set of possible outcomes in the first > round, and works out the probabilities that a team finishes first or second. > I also have sheets that work out the probabilities of victory of every > possible matchup from that point on, allowing an estimate of the probability > constructed from the joint probabilities of every possible outcome. OK, so mine account was really fictive! Anyway, thanks for the explanation. Have you personally designed the spreadsheet? I hope that you'll be able to explain how is possible to have predictions from the ratings: I see that the following page http://www.image.co.nz/aqb/about_predictions.html is expected to be filled. :) -- Cheers milivella |