Hell if we advance... that means we took points off of 2 solid teams and possibly even Germany. Who ever is up next could be a bit scared to face us, especially since Belgium is coming from such an easy group.
3 clicks using ELO. 1rst click USA doesn't advance from the group. 2nd click USA doesn't advance from the group. 3rd click USA loses to Brazil in the final - I'll take it!
Thanks to @pirozhok for putting this together in a more usable, distributable form than my spreadsheet. As it's 1/2 author you'd think I could rig it, but no. Took til the 8th run (2 Elo + 3 FIFA +3SPI) before the SPI finally gets the USA out of the group (on 3 points!). But then we run through Belgium, Switzerland, and Italy. Lose to the host in the final though. Enjoy!
That's really neat prozhok (& dlokteff)! Do you mind if I add a link to this from my WC14 rankings game?
i did about 7 quick ones, usa made it out of the group 3 times, never got past belgium in the second round
That's hardly the "perfect simulation" when England wins third place... heck for that to happen Rooney would probably have scored a goal.
SPI and FIFA ratings have been updated. There was a number conversion bug in SPI conversion, has been fixed, so all numbers are good now.
This is encouraging... I clicked the ELO maybe 50 times, and Ghana was the runner up only twice, and never was the winner.
The US never wins the World Cup in all of my simulations. But interesting takeaways-- I found that the rarest result of the group is that the USA gets last place in the group of death. The US advances about 1/3rd of the time, where we usually do well against the Round of 16 opponent--Korea, Russia or Belgium. (Man, a USA/Russia round of 16 match would be dramatic, wouldn't it?) We tend to win those half the time, but in my models either we crash in the round of 8 (Usually against Argentina) or we go on to the third place match (Which happened 3 times). I should have capped it, but one of my favorite results was the US not advancing and only getting 4 points to finish third, but we beat Ghana 5-0 and drew Portugal 2-2. Even though we wouldn't have advanced, I would love to see that opening game.
More encouraging news. The percentages look good against Ghana note so good against Portugal. GHANA MATCH: ELO: Ghana win: 17% US win: 45% Draw: 35% SPI: Ghana win: 23% US win: 24% Draw: 53% FIFA: Ghana win: 13% US win: 50% Draw: 37% PORTUGAL MATCH - ELO: Portugal win: 35% US win: 22% Draw: 43% SPI: Portugal win: 28% US win: 29% Draw: 43% FIFA: Portugal win: 40% US win: 20% Draw: 35%
In the FIFA version, US got out of the group 1/7 times as runner-up and of 1 crashed out in the quarter-finals against France. In the SPI version, US got out of the group 2/7 times, one as runners up losing to France in the quarter-finals and the other had them first place in the group but loses to Korea in the first round. Korea wins the World Cup. In the ELO version, US got out of the group 1/7 times as runner-up and of 1 lost to Argentina in the quarter-finals.
There are some really wild swings in the algorithm. Maybe you could post your algorithm -- we might be able to help. I know rankings are a tough one to base predictions off of, but it seems the randomness is a bit too wild right now. Here's a sample of one of our draws: DE-PT: 3-0|DE-GH: 3-0|DE-US: 3-0|PT-GH: 0-3|PT-US: 0-3|GH-US: 3-0 If we beat Portugal 3-0, would we really lose to Ghana 3-0? I would say create a variable for form factor (+/- logarithmic up to %25), and then one for your current ranking system. I'm assuming your current per-game randomizer the majority of match predictor currently? The results are always only a W/L/D -- so 3-0, 1-1, or 0-3 -- I think it makes more sense to predict goals scored based on the ranking difference -- then there would be more instances of nations going through on goal difference. Blah blah blah bobloblaw's law blog. Maybe I'll take a stab at this. Seems fun. (Edit -- I didn't realize the scores were points, not score)
Of course any ideas which can improve the tool, are welcome. Game score predictor would be awesome for example. This line you are quoting shows results, not scores, GH-US: 3-0 means GH wins, US loses: DE-PT: 3-0|DE-GH: 3-0|DE-US: 3-0|PT-GH: 0-3|PT-US: 0-3|GH-US: 3-0 Match predictor is based on main simulator model, it runs simulator 100 times, and gets the average to display probabilities. Regarding the main algorithm - that's the question for Dr. Lokteff who built the model
Right now we are breaking all the ties (2-way, 3-way and 4-way) randomly, if you can suggest algorithm for goals scored predictor based on rankings - let's discuss it, i'm open for anything which can improve tool accuracy.
Added one more host, you can use any of these 3 links now (3rd one is acting up lately): http://aspspider.ws/pirozhok1/sim.aspx http://aspspider.info/pirozhok/sim.aspx http://pirozhok.somee.com/sim.aspx
Working on adding goals scored feature for simulator, will be available later this week or early next week.
I am working on tie-breaking procedures (using scores) for simulator, and my brain is already fried today, so i have to ask stupid question about WC tie-breakers. This is quote from FIFA WC regulations: Now if e) gives us GD=0, how can i use f) to break a tie? If teams are tied at GD between themselves, GF between them won't make any difference, it will be =0.
I seem to remember in that instance, the advancing team will be decided by a coin toss. This is how Korea were eliminated and Canada went on to win the 2000 Gold Cup. So in that circumstance, it's random 50/50.