Friday, March 22, 2013

2014 FIFA World Cup seeding based on generated results (22 March 2013)

As promised, here's a look at the 2014 World Cup seeding for the final draw and European play-offs based on results generated using the Elo Ratings.

I've generated the results for the 491 matches between now and 15 October 2013. I'm well aware actual results could be wildly different. Another thing to take into account: teams will most likely scheduled friendlies for the remaining available FIFA dates and this will affect their final ranking.

The outcome could be a bit suprising for some.

Groups D, I and J in Africa and group G in UEFA seem set for a photo finish, while Uzbekistan look destined to make their World Cup debut.

Here's the breakdown by confederation:
AFC:

Korea Republic and Japan as group winners, Uzbekistan and Australia as runners-up. Play-off between Iran and Jordan to decide who will meet Uruguay.

CAF:

Group winners: South Africa (A), Tunisia (B), Côte d'Ivoire (C), Zambia / Ghana (D), Burkina Faso (E), Nigeria (F), Egypt (G), Mali (H), Libya/Cameroon (I), Senegal/Uganda (J).

The draw mechanics for the next round are not known yet.

CONCACAF:

Direct qualification: Mexico, USA, Honduras. Costa Rica to meet New Zealand in the play-offs.

CONMEBOL:

Direct qualification: Argentina, Colombia, Ecuador, Chile. Uruguay to meet Iran/Jordan in the play-offs.

OFC:

New Zealand to meet Costa Rica in the play-offs.

UEFA:

Direct qualification: Croatia, Italy, Germany, Netherlands, Switzerland, Russia, Bosnia-Herzegovina / Greece, England, Spain.

This is how it looks for the UEFA play-off (if the seeding is done using the October FIFA ranking):

Pot A: Portugal (7), Greece (13) / Bosnia-Herzegovina (16), Belgium (14), Sweden (16)
Pot B: Romania (22), France (24), Norway (25), Poland (37)

Czech Republic out as worst runners-up.

Greece / Bosnia-Herzegovina are both displayed because according to the generated results, they will both finish with 26 points and it will go down to goal difference. Bosnia-Herzegovina (GD +13) currently have the upper hand, Greece's GD being only +4.

This is the top 30 of the October 2013 ranking based those generated results:


1 Spain 1649
2 Croatia 1599
3 Argentina 1538
4 Colombia 1386
5 Netherlands 1349
6 Germany 1290
7 Portugal 1243
8 Mexico 1233
9 England 1227
10 Brazil 1190
11 Italy 1041
12 Switzerland 1025
13 Greece 1002
14 Belgium 996
15 Ecuador 992
16 Bosnia-Herzegovina 986
16 Sweden 986
18 Uruguay 973
19 Côte d'Ivoire 971
20 Russia 942
21 Czech Republic 928
22 USA 923
22 Romania 923
24 France 921
25 Norway 912
26 Ghana 885
27 Japan 875
28 Denmark 846
29 Mali 808
30 Chile 806


The seeded teams would be: Brazil, Spain, Croatia, Argentina, Colombia, Netherlands, Germany and one of Portugal or Mexico.

Based on the generated results, Portugal will be involved in the play-offs. If they don't make it, Mexico would be seeded.



About me:

Christian, husband, father x 3, programmer, Romanian. Started the blog in March 2007. Quit in April 2018. You can find me on LinkedIn.

48 comments:

  1. So if Spain drop a couple of unexpected points and Croatia perform as expected, we could enter the WC 2014 draw as World No.1. Nice dream. :)

    ReplyDelete
    Replies
    1. Like Spain 1 Finland 1? :)

      Spain need to beat France, or they're probably going in the playoffs.

      Delete
    2. I let you dream ELO ratings will be the reality ;-) ELO ratings expect Croatia to beat Belgium but I don't agree with that.

      Delete
  2. OK. This will now be fun to watch. For some time I've thought that Mexico would be a seeded team sometime relatively soon. Looks like there's a reasonable chance of it in Brazil.

    ReplyDelete
    Replies
    1. I wonder if England are now 7th thanks to Honduras and Israel? I'm assuming those ties were not predicted.

      Delete
    2. Lorric,

      Portugal, Mexico and England are locked in an awfully tight battle for 7th place by the looks of it.
      Right now the predicted points for the October 2013 ranking are:
      1 Spain 1698
      2 Croatia 1576
      3 Argentina 1500
      4 Colombia 1358
      5 Germany 1304
      6 Netherlands 1282
      7 Portugal 1212
      8 Mexico 1209
      9 England 1191
      10 Brazil 1152
      11 Italy 1096
      12 Bosnia-Herzegovina 1041
      13 Ecuador 1004
      14 Belgium 970
      15 Cote d'Ivoire 969
      15 Greece 969
      17 Denmark 957
      18 USA 955
      19 Switzerland 951
      20 Russia 933

      Certainly there were a lot of surprises yesterday. Elo's prediction success rate so far this month is just 57% !

      Let's wait and see what Tuesday's matches will bring.

      Delete
    3. Why did England drop points? They won their match. Shouldn't it still be 1227 and 7th?

      Delete
    4. And why do Spain have 49 more points when they dropped points to Finland?

      Delete
    5. You forget that Edgar probably applies a slightly different method of predicting results based on the ELO win-probability (Pw).

      I apply the following rules:
      For friendlies:
      if Pw > 0.634 than home-team wins
      elseif Pw >= 0.366 than draw
      else away-team wins
      For non-friendlies (importance-factor > 1):
      if Pw > 0.609 than home-team wins
      elseif Pw >= 0.391 than draw
      else away-team wins
      Both numbers 0.634 and 0.609 are derived from the percentage of draws in the last 4 years for each type of match, 26,8% in friendlies and 21,9% in non-friendlies.

      Furthermore, at the moment, I do not adjust the ELO-rating for a team according to each predicted future result. I am busy implementing that right now, though.

      Maybe Edgar is willing to shed some light on his method of establishing predictions based on ELO?

      Anyway, you can see that the overall trend is the same. The exact points differ slightly.

      Delete
    6. Ah.

      Well, the best thing that can happen is simply for Portugal to fail to make it. I'd like to say the same for Mexico, but there's no way, surely, even with their stuttering start. But I think they'll drop some more points on the road.

      It's annoying this can happen to England, even with victories predicted for all their matches, and a quarter final appearance and qualifying group and tournament group wins at the Euros.

      But as was said about a certain rabbit in the forest, we've got a Montenegro problem first. I expect the saying means you have to go and kill the rabbit before you can eat it.

      Delete
    7. Yeah, big match for England this Tuesday. I guess anything less than a win will be a major upset for the English fans.

      Delete
    8. @Lorric
      You got it right. It goes something like: Thou shalt not prepare a skewer when a rabbit is still in the forest. :)

      Delete
    9. Well we couldn't beat them in the last campaign. I want a win. I don't want them sitting up there 2pts clear, let alone 5 with a loss, but I could take a draw, which would leave it still in our hands, then beat them at home and we'd be fine. I don't know whether they use H2H or GD, but our GD is massive, so we should be okay if it's GD.

      Ukraine and Poland need to be watched too. Both are yet to play San Marino Ithink, so the 5pts they're on now is basically 11pts, same as England. Montnegro have 13, and the latter two have both played San Marino twice, so it's up for grabs. Ukraine and Poland also have a game in hand. So this is still a 4 way dance.

      Delete
    10. Confirmed, both haven't played San Marino yet. So realistically it's a 3 way tie for 2nd with Montenegro 2pts clear.

      Delete
    11. Lorric,

      it's about GD first. England atm +18 vs. Montenegro +11, both have slaughtered San Marino the best they could :)

      Poland and Ukraine both have dropped three more points compared to England. If they both win home and away against San Marino, then they both have 11 points from 6 matches. England has 11 points from 5, Montenegro has 13 from 5.

      Delete
    12. Thanks for that. You're being vey helpful today, aren't you :)

      Delete
    13. Nothing better to do, today,
      and it's always nice to help :) Otherwise, people got to have some patience before Edgar replies.

      I promise this will be my last reply today (or a very interesting question must pop up somewhere on this site....).

      Delete
  3. Croatia, Colombia and Mexico in the playoffs... expect some FIFA manipulation :) I can't see them allowing that to happen.

    ReplyDelete
  4. Nothing to do with the seedings, but it could well be a very nice summer for Belgium. If my predictions are correct (which they are usually not entirely), the Belgians could claim a place in the top 10 for the first time ever. That event would take place in the rankings of July with a possible 9th position. In August and September they would be in or near the top 10 also, before dropping back a bit again in October and thus missing out on a seeded spot in the World Cup.

    If they make it to the World Cup, the Red Devils are a force to be reckoned with. I for one don't want them in the same group as my country (Holland).

    ReplyDelete
  5. So has it been confirmed if the Europe & Africa play-off draws will be seeded yet? Or are they going to do the usual trick and see who is in them first? Could be very interesting in Europe if Spain, Portugal, England, Croatia & Sweden are all there.

    ReplyDelete
    Replies
    1. Yep, throw in Greece, Turkey and Denmark and an "open" draw could throw up some very tasty "local" encounters.

      Delete
  6. Just great !

    The Netherlands goes to Asia in June and plays two friendlies against Indonesia and China (for what undoubtedly will be a substantial fee). With that their predicted position in the October ranking drops from 6 to 9 (Lorric, even behind your England) !
    Bye bye, seeded position.

    ReplyDelete
    Replies
    1. Well, maybe you don't have to worry about England now. Damn bogey team! Lucky to scrape out of there with a draw. England need to wake up. One of the worst 2nd halves of football from an England team I've ever seen in my life, Montenegro choked on some of their chances, they could have scored 3 or 4.

      Delete
  7. Yes, but other nations will play friendlies too. For example, Croatia plays Liechtenstein in August (reasons for this idiotic decision remain unknown). This should even out the rankings in the end.

    ReplyDelete
  8. Here's an update of the predicted October 2013 ranking after the matches of yesterday. Predictions are made using ELO for all matches which are included in the FIFA fixture list at the moment (nogomet: no Croatia-Liechtenstein friendly included yet). New feature in the calculation compared to Saturday: ELO ratings are now adjusted for each predicted result.

    Predicted October 2013 ranking (between parentheses the position last Saturday):

    1 Spain 1617 (1)
    2 Croatia 1577 (2)
    3 Argentina 1418 (3)
    4 Germany 1305 (5)
    5 Colombia 1230 (4)
    6 Mexico 1222 (8)
    7 Portugal 1201 (7)
    8 Netherlands 1177 (6)
    9 Brazil 1163 (10)
    10 Italy 1131 (11)
    11 England 1111 (9)
    12 Bosnia-Herzegovina 1064 (12)
    13 Belgium 1000 (14)
    14 Ecuador 999 (13)
    15 USA 982 (18)
    16 Cote d'Ivoire 975 (15)
    17 Greece 971 (15)
    18 Switzerland 958 (19)
    19 Russia 948 (20)
    20 Sweden 895 (21)

    England seems to lose contact with Portugal and Mexico following their unforeseen draw with Montenegro and is now positioned behind Italy. The Netherlands drops 2 positions because of two newly included friendlies in June.
    Next update June 8th.

    ReplyDelete
    Replies
    1. 11th... At least it won't feel like an injustice to lose the seeding anymore. England just aren't worthy of it right now if they get there.

      Delete
    2. Ed, are you simulating out the Gold Cup as well?

      Delete
    3. Yes, Gold Cup, East Asian Cup, every match that's included in the FIFA-fixture list right now.
      And indeed, for the Confed Cup and the Gold Cup the whole tournament is predicted. Brazil wins the Confed Cup (after PSO against Spain) and Mexico the Gold Cup (after PSO against USA), mark my (oh no, ELO's) words :)

      Delete
  9. Amazing to see how the result of one match can influence your FIFA ranking position. I'm very skeptical about Brazil winning the Confederations Cup though. They've got a new manager that has failed to get the expected results from Elo Ratings (draws against Italy and Russia instead of the expected wins).

    But anyway, brilliant work.

    ReplyDelete
    Replies
    1. Rafael, thanks !

      A few comments:
      - I share your thoughts about Brazil.
      - Elo's prediction success rate has climbed slightly to 59%. That's not what I had expected beforehand. It's the first time that I work with Elo predictions and I thought a success rate of 65-70% should be feasible.
      - keep in mind that each and every result has effect on the result prediction for each match a team plays after that (Elo ratings get adjusted) and also has an effect on the FIFA points a team builds up for each month's ranking. So after an unsuccesful prediction the positions on the next ranking will possibly be different and therefore this can also influence match points a team can get in future matches. So everything depends on everything and that for some 350 matches until October 15th.
      In that light it's almost amazing how little the predicted ranking get influenced by an unsuccesful prediction :)

      Delete
    2. Hi Ed,

      ELO is a probabilistic system, therefore if it says a team has 0.7 Equity, when you have to remember that part of it are draws too, ELO itself says the favorite won't win 70%...
      The best thing would be simulating every match a lot of times according to the probabilities implied by ELO.
      That's what I did to simulate the qualifing in Europe here:
      http://simtheworld.blogspot.com/

      If you could please give me your FIFA Rankings worksheet lease, I could simulate every match and list the chances for every team and position.
      You can send it to:
      amir.junkmail@gmail.com
      (I have to use my secondary mail because I put it here for everyone to see...)

      Regards,
      Amir

      Delete
    3. Hi Amir,

      I saw your simulations for the UEFA qualifying phase. Very interesting, but for my purpose (predicting the ranking points per team) the resulting probabilities for each 'state' are not really useful.
      As you know in the FIFA ranking calculation a team receives match points per match played, so I need one 'fixed' prediction of the result of each match, not a probability distribution per match.

      I know it is not statistically sound, but I use the win expectancy from Elo as an approximation of the chance the home-team will win a match against a specific opponent. That way I can easily generate a 'best guess' about my desired end situation: the predicted points for a team in October.

      As it is football, and certainly not an exact science, for me that's good enough. In my experience the whole concept of probabilities and statistics in general is too difficult for most people to interpret correctly.
      The constant changing of planned matches, new fixtures etc. is another argument why I want an easy to reproduce calculating method. Now with some simple formulas in Excel (and a few tricks to avoid circular references) I can accomplish that.

      So Amir, thank you for your offer, but I will stick with this approach.

      Delete
  10. So France is almost 100% of certain not-seeded in play-off round....damm, tough.
    What are the possibilities of France and Sweden exchange positions? what France have to do?

    Also, Ed or Edgar, if Africa goes by the same way of Europe, who would be the 5 seeds? Probable Côte d'Ivoire, Ghana, Nigeria, Mali and Tunisia right? Also tough, with Egypt not being seeded.

    ReplyDelete
  11. This comment has been removed by the author.

    ReplyDelete
  12. Gabriel,

    France is predicted 24th in October, right behind Denmark at 22nd spot. They are ranked 14th among UEFA countries.

    I see a fair chance that one or two (Denmark, Portugal ?) of the 13 standing in front of them, won't make at least the play-offs, so I would not call that an almost 100% certain unseeded prospect for France, but more a 50-50% chance of being seeded/unseeded in the play-offs.
    France has a predicted 877 points, only 18 behind Sweden and 6 behind Denmark. Anything is still possible for France, I would say.

    About the play-offs in Africa: that's a lot harder to predict with any conviction because a lot of groups are still very close. At the moment only Tunisia leads their group comfortably.

    All I can give you now is the predicted ranking among CAF-countries in September, the ranking which probably will be used to seed the play-offs:
    1 Cote d'Ivoire 989
    2 Ghana 867
    3 Mali 809
    4 Nigeria 742
    5 Tunisia 732
    6 Burkina Faso 674
    7 Algeria 661
    8 South Africa 659
    9 Cape Verde 607
    10 Cameroon 602
    11 Libya 601
    12 Egypt 598

    Of these twelve only Burkina Faso and Cape Verde seem already out of contention for a play-off spot. Knowing the structure of the African qualification process it was evident from the start that some heavy-weight casualties were to be expected.

    ReplyDelete
  13. Thanks Ed, i believe France shot of being seeded is to make this 18 points deficit with Sweden, or Sweden dont go to the playoffs, since Irealnd and Austria are in dispute too.

    I can't see Portugal oof the play-offs, they still decides the game against Israel home, and Israel still have to go to Russia, while Portugal host them.

    well we can't count Russia on either, if they lose to Portugal and grab a draw in Belfast (if it happens in june) they will be 1 point behind Portugal in june, 1 game to go, and the pressure on them, deciding away the 2 last games.

    ReplyDelete
  14. How do you make these predictions? Do you use an excel table or just take note?

    ReplyDelete
    Replies
    1. Bruno,

      I use Excel for the whole calculation and prediction. Don't know exactly what Edgar uses.

      Delete
  15. Edgar,

    I notice you include the matches of October 15th in the October ranking. The official deadline is October 13th (which I use in my calculations). Do you have any information that the deadline for the October ranking is adjusted to include these Tuesday-matches ?

    ReplyDelete
  16. @nogomet

    Don't sell the fur of the bear that's still in the forest (that's a Romanian proverb).

    @sourav MITRA

    October 2013.

    @Ed

    I derive the draw thresholds from all games in my database regardless of the match type (friendly or competitive). I also update the ratings after each generated result. I wrote an application in C# and I use a MSSQL database.

    Regarding the deadline for October (and September). I expect FIFA do move the deadline to 15 October (and 10 September) to include the qualifiers. They have done this twice before.

    Thanks as usual for all the help!

    ReplyDelete
    Replies
    1. OK, thanks !

      That is a good explanation why I had still quite some differences compared to your predictions. Larger differences also then could be explained from just a slightly different method to establish predictions based on the Elo-probability.

      To avoid confusion I will from now on use the same deadlines for the October and September ranking as you do.

      And of course you are welcome, although I have the feeling that I might be causing more confusion at the moment than be really helpful.

      Delete
    2. So it's rabbits for Croatia, bears for Romania, and chickens for England. Anyone else got any more? :)

      Delete
    3. Birds? "A bird in the hand is worth two in the bush."

      Furrier (USA)

      Delete
    4. Bosnia expands the bird theme: "better a sparrow in hand than a pigeon on the branch". This could apply to our obsession with playoffs, so were going for the pigeon this time!

      Delete
    5. You're getting a bit off. The point of the aforementioned proverbs is to say that one shouldn't count on things too much in advance. The one you're suggesting says that it's better to settle for less because there is a risk of losing everything. Not quite the same. ;)

      Delete