I fixed most of the bugs i could find with your help, guys. Also Match Predictor now has scores option, please take a look and see if there are any other bugs, before i deploy it to main site: http://aspspider.info/pirozhok/sim.aspx
Ran through a bunch of simulation but I didn't keep much record so I can't be sure but it feels like it is less glitches but also that there is a lot less 3 or 4 way ties. Not sure if anything you've done would have that outcome or if I just need to do more simulations. This one though I think came out wrong. I make it as goals scored would be the decider and if so the wrong team comes out on top. (simulation done 20 minutes before post) And this one too. A GD of -4 is ranked above a GD of -3, should be the other way round. Lastly one that isn't an error but simply fun. They did a round of 1-0 victories against each other and then all drew 1-1 v the 4th team of the group. So a 3 way tie with everyone on exactly equal stats.
Scores option added for Match Predictor, it derives most likely score now. Please note you can go directly to Match Predictor from Schedule page - click on any cell and it would take you to Match Predictor for selected match with Match info prepopulated: http://pirozhok.somee.com/sim.aspx
Main website was acting up today, so i've added another link: http://aspspider.info/pirozhok/sim.aspx Main: http://pirozhok.somee.com/sim.aspx
Knockout stages tie-breakers have been added (AET or PKs wins): http://pirozhok.somee.com/sim.aspx http://aspspider.info/pirozhok/sim.aspx
Didnt count 20-30 would be a guess just kept clicking, Germany/Portugal win the group 70-80% of the time, Ghana hardly ever appears.
Try to run with scores, and also you can try Match Predictor using weighting factors in this case you can override historical data and ratings.
Having a lot of fun with this! One question I have - I seem to often get a lot of exact replicas of the whole tournamnet. How does the random number seeding work?
the danger here for me is, if i run this thing enough looking for results i like, i start to think the us might actually make it out of the group. before i know it, i'm like "c'mon, no way we lose to switzerland in the quarters..."
I'm using VB.NET built-in random number generator which in reality does not produce truly random numbers (there are a lot of articles online discussing this issue), tool uses one seed for each simulation, random floating number between 0 and 1 for each match. I'm running World Cup projections simulation right now - 100,000+ times which will allow to eliminate ambiguities and will provide accurate enough projections. Will publish results later this week (to run each model 100k+ times takes between 27 and 30 hours ).
Simulator is based on ratings and historical data, due to our unique geographical location in CONCACAF US ratings usually peak before WC - thanks to Gold Cup and WCQ results, so yes, simulator model is probably overestimates us.
This would not be the case though for ELO ratings would it? Surely there is no advantage to how many games you play or the quality of the opposition using this model?
Well, if ratings are higher, team has higher probabilities. Please note that simulator uses ratings and historical data (team's WC performance) to calculate probabilities, not just ratings alone.
Interesting. Is the exact mixture of previous history and rankings a secret? Whatever - it's a great little page you've made there! If I could make one (well two!) teeny requests they would be: 1) Show the matches in the group in order (might be easier to do by allocating teams as 1,2,3,4 by their group position then just list 1v2,3v4,1v3,2v4,1v4,2v3) 2) Dates (maybe times too) that the games will take place? I'm being greedy now I know!
Matches in each group are played in different order, so this request will require some time, don't know if it's worth the trouble . Dates and times are on the Schedule page (click Schedule item in top-level menu).
When you say they are played in a different order ... they are all played in the 12,34:13,24:14:23 order in every group - the number of each team though was determined by draw when it was made Dec 6th last year (this was the second ball that was drawn by number). So in theory in you just slotted each team in the right number and applied that order to each group then all would fall through. But, to be honest, I love this tool so much that if it's a lot of trouble then don't risk breaking I don't know if you keep track on the number of generations people have done. Seriously I reckon I'm on 3000+ LOL
The Australia one is interesting. I don't know if you guys ever look at football-rankings.info but he runs a lot of sims on there which are based solely on ELO (with no added history as priozhok has done) - and he finds Australia get out of that group 15.59% percent of the time. So they are punished quite a lot in this model it seems.