More about the simulations in the second and third paragraphs of the first 'Road to Brazil' post.
Most improved since the 12 April simulations:
32.25% - Italy
30.49% - Belgium
19.55% - Chile
19.26% - Colombia
13.1% - Uruguay
Most declined:
-49.15% - Croatia
-23.69% - Mexico
-23.11% - Ecuador
-6.58% - Argentina
-5.64% - Russia
Chances of being seeded over the 10000 simulations:
100% - Brazil
99.12% - Spain
97.41% - Germany
87.31% - Argentina
76.5% - Italy
72.36% - Colombia
61.93% - Netherlands
47.52% - Belgium
33.49% - Croatia
25.02% - Portugal
23.98% - Uruguay
23.65% - England
22.74% - Chile
14.69% - Switzerland
7.44% - Ecuador
5.8% - Bosnia-Herzegovina
0.56% - Peru
0.31% - Czech Republic
0.12% - Mexico
0.02% - Albania
0.01% - Russia
0.01% - USA
0.01% - Greece
Chances of being seeded if the team has qualified:
100% - Brazil
100% - Spain
98.11% - Germany
87.31% - Argentina
79.65% - Italy
72.59% - Colombia
61.99% - Netherlands
55.08% - Uruguay
51.8% - Belgium
49.94% - Croatia
37.78% - Portugal
29.18% - England
24.61% - Chile
15.5% - Switzerland
7.73% - Ecuador
6.58% - Bosnia-Herzegovina
2.89% - Peru
2.18% - Czech Republic
0.37% - Albania
0.14% - Mexico
0.02% - Greece
0.01% - Russia
0.01% - USA
Bit of a long shot, but could be pretty cool if all CONMEBOL teams (5 or 6 that qualify) get seeded! Brazil, Argentina, Columbia, Uruguay, Chile (and maybe Ecuador) -- home continent advantage :-)
ReplyDeleteI wonder what the probability is that this happens ;-)
All 6 to be seeded has never happened during the 10000 simulations.
Delete5 to be seeded (Brazil, Argentina, Columbia, Uruguay, Chile): 0.75%
Since Uruguay's ranking is going to improve (due to Conf. Cup results), there is a big chance of having 4 south americans and 4 europeans seeded (if Uruguay manages to qualify of course :) )
DeleteThanks, Edgar.
DeleteI think it would be somewhat neat if FIFA did decide to seed five CONMEBOL teams -- which would likely require a change to their seeding criteria. If they did, that would leave three UEFA teams, with Spain, Germany, and Italy looking as the three most probable (objectively and perhaps subjectively). Again, if FIFA did, then the 4th best UEFA team (likely Netherlands) wouldn't get a seed, and would have an argument (objectively and perhaps subjectively) that they deserve it more than Chile. If Chile isn't seeded, then they would be in a group with one of the UEFA seeds (Spain, Germany, Italy, Netherlands). Does FIFA want CONMEBOL teams to have more seeds and go deeper into the tournament? Does FIFA want the host to go deep by seeding lots of UEFA teams so Brazil can avoid these in the group stage?
It will be interesting to see how many CONMEBOL vs UEFA seeds there will be (the last S. American host was back in 1978). I'm hoping FIFA likes the seeds based purely on the ranking criteria, especially (as Daniel Burnier points out), there could be four CONMEBOL seeds based on the rankings. Edgar, if you had some extra time ;-) what is the probability (post-Confederations Cup) that four CONMEBOL teams are seeded (based on rankings).
THANKS!
If you look at Ed's post of 22 june 8:11AM, it is pretty likely there will be 5 European teams seeded and 3 CONMEBOLs.
DeleteIf so, i guess these are the groups, taken into account the best predicted teams (Edgar's 21 june 2014 FIFA WC Q chances) of each continent qualify.
why: if you have 5 seeded European teams, you end up with 8 remaining EU-teams, which would form 1 group (in order to avoid 3 EU-teams in 1 group). The remaining groups can be mixed either AF/NA or AF/SA and NA/OC/AS or SA/OC/AS.
(AS=Asia, OC=Oceania, NA= North America, SA=South America)
Seeded group:
SA Brazil
EU Germany
EU Spain
EU Italy
SA Argentina
SA Colombia
EU Belgium/Croatia
EU Netherlands
Second group:
EU Switzerland
EU Russia
EU Bosnia-Herzegovina
EU England
EU Croatia/Belgium
EU Portugal
EU Greece
EU France
Third group:
USA NA
Costa Rica NA
Mexico NA
New Zealand OC
Japan AS
Korea AS
Iran AS
Australia AS
Fourth group:
Cöte D'Ivoire AF
Nigeria AF
Ghana AF
Algeria AF
Egypt AF
Ecuador SA
Chile SA
Uruguay SA
Yes that is a good point... if there are 4 SA teams seeded we will have 8 europeans in pot 2 and 1 in pot 3... this will make the drae more complicated (I mean, more restrictions, etc...) I believe FIFA will want to avoid that.
DeleteIt can be done. FIFA could make a special pot for the worst placed European team. For the draw of the WorldCup 2006 they did the same thing.
DeleteSame restrictions can be applied on the draw: maximum of 2 European teams in one group. So with 4 Sout-American teams being seeded, the European team in the special pot would have to be placed in one of the groups with the Sout-American seeded team.
Instant "group of death" material, with 1 South-American and 2 European teams...
If that's how it turns out, those pots look fairly balanced. As for a "Group Of Death" how about Germany, England, Mexico, Uruguay? "Group Of Life": Belgium, Greece, New Zealand, Algeria? On that note, it would seem a bit unfair if Belgium are seeded, being as they haven't qualified since 2002 whereas the likes of England & Portugal are regular qualifiers who usually at least reach the Last 16. I also don't see NZ beating the 4th Concacaf team in the play-off as the Concacaf team will be "battle-hardened" whereas NZ's last competitive fixture will have been against a small Oceania island nation. Perhaps they should have taken the Oceania Cup more seriously and qualified for the Confederations Cup to get a taste of higher quality football?
DeleteTalking about "group of death"... considering that Uruguay gets seeded AND considering that France is the worst placed European in the ranking (which is not impossible) and goes to the "extra pot"; we could have a group with Argentina, England and France :-)
Delete@Dorian
DeleteExactly 4 CONMEBOL seeds: 16.9%
In addition to the new simulation results here's the latest update of the top 20 of the probable October ranking based on 'best guess' elo predictions for all matches up until and including October 15th.
ReplyDelete1 Spain 1629
2 Germany 1341
3 Colombia 1333
4 Argentina 1207
5 Italy 1195
6 Netherlands 1173
7 Croatia 1166
8 Belgium 1116
9 Brazil 1088
9 Portugal 1088
11 England 1078
12 Bosnia-Herzegovina 1069
13 Switzerland 1053
14 Chile 998
15 Mexico 993
16 Russia 977
17 Uruguay 966
18 USA 964
19 Ecuador 962
20 Cote d'Ivoire 946
I question? According to the prediction, what would be the result between Crotia - Belgium to be played on October? Could Belgium overcome Croatia in the ranking with another result?
DeleteCroatia can't have as much as 1166 pts with a loss to Belgium (according to my calcs, 1166 pts means Croatia drawing with Serbia and beating Belgium), so yes, Belgium would overcome Croatian in case of a different result.
DeleteFederico, MV,
DeleteElo predicts a win for Croatia against Belgium (and indeed a draw against Serbia), so yes, with another result Belgium will overtake Croatia:
with a draw Belgium will have 1163 pts (7th) and Croatia 1058 pts (12th).
with a win Belgium will have 1258 pts (4th) and Croatia 1004 pts (14th).
I'm off line the next three weeks so mid July we'll see what kind of influence the Confed Cup finals has had.
Back from holiday here's the latest update of the top 20 of the predicted October ranking after the conclusion of the Confederations Cup and the first rounds of matches in the Gold Cup and Cosafa Cup:
Delete1 Spain 1554
2 Germany 1339
3 Colombia 1334
4 Italy 1229
5 Argentina 1205
6 Netherlands 1173
7 Croatia 1165
8 Brazil 1121
9 Belgium 1115
10 Portugal 1088
11 England 1078
12 Bosnia-Herzegovina 1069
13 Switzerland 1052
14 USA 1024
15 Chile 999
16 Russia 978
17 Uruguay 966
18 Ecuador 961
19 Mexico 942
20 Cote d'Ivoire 937
Italy has made a good impression in the Confed Cup and are now firm favourites to take a seeded spot in the WC draw. Mexico on the other hand has slipped away after a rather dismal display at the CC and a far from convincing start of the Gold Cup.
USA climbed considerably thanks to an updated prediction for the Gold Cup final: a straight win against Mexico instead of a PSO win. Cote d'Ivoire feels the impact of their unpredicted CHAN-qualifier loss to Nigeria.
Ed,
DeleteNice work! I'd be curious to know the list of all intra-top 20 matches between now and the Oct ranking. You've mentioned a few (Portugal/Netherlands, Portugal/Brazil, Croatia/Belgium, USA/Mexico) -- and it would be interested to see the predicted ELO outcome (2, 1, X) for these as well.
For extra credit ;-) it would be interested to see the MIN and MAX points for the October 2013 ranking (although I suppose not knowing the exact matchups of tournament play like CONCACAF Gold Cup would make MIN and MAX points a bit tricky to calculate...)
Dorian,
Deleteyou're welcome!
Here's the list of the current predictions for intra top 20 matches and all other matches where the top 20 team is predicted not to win (up until and including October 15th):
2013-07-28 USA - Mexico; Continental final; 1
2013-08-13 Bosnia-Herzegovina - USA; Friendly; X
2013-08-14 Belgium - France; Friendly; X
2013-08-14 Ecuador - Spain; Friendly; 2
2013-08-14 Italy - Argentina; Friendly; X
2013-08-14 Japan - Uruguay; Friendly; X
2013-08-14 Mexico - Cote d'Ivoire; Friendly; X
2013-08-14 Portugal - Netherlands; Friendly; X
2013-08-14 Switzerland - Brazil; Friendly; 2
2013-09-06 Costa Rica - USA; World Cup qualifier; X
2013-09-06 Colombia - Ecuador; World Cup qualifier; 1
2013-09-06 Peru - Uruguay; World Cup qualifier; X
2013-09-06 Serbia - Croatia; World Cup qualifier; X
2013-09-09 Brazil - Portugal; Friendly; 1
2013-09-10 USA - Mexico; World Cup qualifier; 1
2013-09-10 Uruguay - Colombia; World Cup qualifier; X
2013-09-10 Norway - Switzerland; World Cup qualifier; X
2013-09-10 Ukraine - England; World Cup qualifier; X
2013-09-18 Brazil - Argentina; Friendly; 1
2013-10-02 Argentina - Brazil; Friendly; X
2013-10-11 Colombia - Chile; World Cup qualifier; 1
2013-10-11 Ecuador - Uruguay; World Cup qualifier; 1
2013-10-11 Croatia - Belgium; World Cup qualifier; 1
2013-10-15 Costa Rica - Mexico; World Cup qualifier; X
2013-10-15 Chile - Ecuador; World Cup qualifier; X
2013-10-15 Uruguay - Argentina; World Cup qualifier; X
MIN-MAX for most top 20 teams in October:
Spain 1334 - 1553
Germany 947 - 1340
Colombia 892 - 1428
Italy 947 - 1252
Argentina 972 - 1368
Netherlands 839 - 1206
Croatia 835 - 1255
Belgium 866 - 1294
Portugal 876 - 1176
England 766 - 1151
Bosnia-Herzegovina 695 - 1107
Switzerland 620 - 1213
USA 706 - 1104
Chile 846 - 1215
Russia 623 - 978
Uruguay 867 - 1248
Ecuador 699 - 1234
Mexico 749 - 1066
I leave Cote d'Ivoire out because their opponent in the 1st leg of the play-off is hard to predict.
The minimum points are calculated as a worst, really worst case scenario, so including losses against Andorra, Liechtenstein, Luxembourg and so :)
Hi Ed,
DeleteMany thanks for this! Great stuff. Very interesting that everybody (except Spain) could knock themselves out of the top 8 (with MIN points). Also, interesting to see all the August 14 matches that may rattle the pots!
Is it possible to say that Portugal will be seeded if they win all their remaining quali-matches (but finish behind Russia)?
ReplyDeleteI mean of course seeded in the WC (after advancing by winning the play-off of UEFA)
DeleteYou can say that, but you can't be sure of it ;).
DeleteThe predictions for Portugal include a draw vs. the Netherlands and a loss versus Brazil. But these 2 are friendlies and 2 wins instead of a loss/draw don't add very much points.
Belgium can also gain points (vs. France & vs. Croatia), while Croatia can only lose points (vs. Belgium f.e.).
So, it's possible, but it wouldn't be a miracle if it didn't happen.
Thank you Jeroen,
DeleteI just realized that the play-off matches don't count for the WC-seeding, so it will not matter if Portugal wins the group or not :)
Jeroen,
DeleteHow many points Portugal would have if they win all the 5 matches remaining, including the two friendlies?
Thank you
Daniel, Mauro,
DeletePortugal is predicted to win their three remaining qualifiers against Northern Ireland, Israel and Luxembourg. As Jeroen said Portugal has still room to improve in their planned friendlies: if they win their two friendlies against the Netherlands and Brazil they will have 1176 pts (6th spot) at the expense of the Netherlands who will drop to 1156 pts (8th spot). So it will be a rather important August friendly for both Portugal and Oranje.
A draw against Brazil (and a win against Oranje) leaves Portugal at 8th spot (1141 pts) right behind the Netherlands.
The battle for the last seeded spots continues, with some new opponents facing each other: the Netherlands, Croatia, Portugal and Belgium.
As usual - thanks for the input, Ed.
ReplyDeleteHere's the latest update of the top 20 of the predicted October ranking after the conclusion of the Gold Cup, the Cosafa Cup, the East Asian Cup and most of the CHAN-qualifiers:
ReplyDelete1 Spain 1552
2 Germany 1341
3 Colombia 1305
4 Croatia 1257
5 Italy 1230
6 Argentina 1206
7 Netherlands 1171
8 Belgium 1116
9 Brazil 1109
10 Portugal 1087
11 England 1082
12 Bosnia-Herzegovina 1063
13 Switzerland 1052
14 Chile 1047
15 USA 1005
16 Russia 975
17 Uruguay 966
18 Cote d'Ivoire 938
19 Greece 928
20 Ecuador 918
Not much happened, most matches with top-20 teams involved went according to the predictions. Most significant change: the prediction for Serbia-Croatia on September 6th changed from a draw to a win for Croatia (and that's because a friendly Colombia-Serbia is inserted on August 14th). This boosts Croatia's predicted ranking, but the clash against the Serbs will not be an easy win for Croatia.
Mexico is definately out of the picture for a seeded spot.
Due to Panama's fantastic results in the Gold Cup they are now predicted to take 4th spot in the CONCACAF hexagonal (instead of Honduras). An intercontinental play-off against New Zealand looms at the horizon with a good chance to qualify for their first ever World Cup.
Ed,
DeleteGreat update! Thank you.
Have you thought about using the ESPN SPI Rankings to forecast outcomes? I wonder how much different your top 20 predicted October results would look.
For these top 20, I put together the table below (and I hope it formats legibly...). The table is sorted by the difference between the SPI Ranking and the Elo Ranking (as of July 18th, so no Gold Cup knockout-round results):
COLUMNS: [Ed's Oct] [Elo] [SPI] [Diff] [Team]
12 22 9 13 Bosnia-Herzegovina
14 15 6 9 Chile
18 15 10 5 Cote d'Ivoire
3 8 5 3 Colombia
6 5 2 3 Argentina
8 18 16 2 Belgium
13 21 19 2 Switzerland
17 13 11 2 Uruguay
9 1 1 0 Brazil
15 20 20 0 USA
20 12 12 0 Ecuador
1 2 3 -1 Spain
2 3 4 -1 Germany
19 23 24 -1 Greece
11 6 8 -2 England
7 4 7 -3 Netherlands
10 9 14 -5 Portugal
5 7 13 -6 Italy
16 10 18 -8 Russia
4 11 30 -19 Croatia
SPI Ranking should boost Bosnia and Chile the most, while hurting Croatia and Russia the most.
Does ESPN publish a way to determine "match odds" for their rankings/ratings?
DeleteWith Elo, "match odds" are determined via their relative ratings, not rankings.
There is this: "One feature of the OFF and DEF ratings as we've designed them is that they can be combined to estimate the win probabilities for any given game between any two given teams." from here:
ReplyDeletehttp://espn.go.com/soccer/worldcup/news/_/id/4447078/GuideToSPI
Dorian,
DeleteI've studied the fundamentals of the SPI-ranking (but not thoroughly I have to admit). I don't see an easy way to use SPI as a predictor for match results. Indeed, they claim they can predict using OFF en DEF, but they don't give an example in the guide how that's done. And even then, the calculation of OFF en DEF is far from easy, because you have to know a good deal about the rosters of the teams in a match too. And that's information I don't register in my match database.
That's why I'm happy with the easiness-of-use of the Elo win expectancy. As Amir already pointed out earlier it's not completely sound to use this win expectancy the way I do, but hey, I want predictions for each match result and I want them automatically reproduced and easily calculated and I want them as a sort of 'best guess'. Elo gives me all that. So to answer your question, no, I don't think about using SPI.
Ed,
DeleteThank you for investigating the SPI approach and the kind comments in reply.
Ed,
DeleteOne other thing I realize about SPI vs Elo: with Elo you can simulate ranking points going forward for many matches. However, with SPI, because the ranking points are determined in part by how the players perform in each match (and which players play), there really is no way to simulate a future ranking for SPI. Meanwhile with Elo, there is!
Thanks again.