Wednesday, September 5, 2018

2022 FIFA World Cup seeding (UEFA): 10000 simulations (5 September 2018)

In the FIFA International Match Calendar 2018-2024 the preliminary draw for the World Cup 2022 in Qatar is scheduled for July 2019. The final draw is scheduled for April 2022.
The July 2019 ranking will probably be used to determine the seeding for the World Cup 2022 European qualification draw.

The upcoming UEFA Nations League will for a large part establish this ranking. Only the first part of the EURO 2020 qualification (2 MD's in March 2019 and 2 MD's in June 2019) is yet unknown (and several friendlies of course), but the July 2019 ranking for UEFA-teams can be simulated already taking into account all UNL-matches (6 MD's in each group and the Final Four in June 2019) and all scheduled friendlies.


UEFA has 13 berths in Qatar so, although the qualification format is not yet known in detail, a fair assumption is that the 55 UEFA teams will be divided in 8 groups of 6 and 1 group of 7 teams. The 9 group winners qualify and the 8 best runners-up are to play 4 play-offs for the remaining 4 spots, all conform the qualification format for Russia 2018. The 2022 qualification draw will in that case use 6 pots, 5 pots of 9 teams and 1 pot of 10 teams.

I've made 10.000 simulations of the complete UNL (group phase plus Final Four) and scheduled friendlies with UEFA-teams involved and calculated the resulting ranking on June 9th 2019 after the Final Four as a proxy for the July ranking (as far as UEFA teams are concerned)
The results of all scheduled matches are predicted using Club-Elo's prediction formula's for goals scored in a match depending on elo home win expectancy.

Below the probabilities for each team to end in each pot are presented. The influence of the new ranking calculation method is immediately clear: teams are much less divided over different pots in this simulation compared to simulation results for previous World Cups. There are more teams with high (> 90%) probabilities for a certain pot, even at this very early stage.

Especially pots 5 and 6 (where UEFA teams have bigger gaps in ranking points) seem already carved in stone. Pot 3 is relatively open. For pots 2 and 4 only a few teams below the dotted line are in a position to have a serious shot.
For pot 1 a big fight is to be expected between Sweden and Germany for the final spot: Germany's chances for pot 1 and 2 are almost equal.

Attention for the position of Russia, a WC quarter-finalist and a solid pot 4 (!) team at the moment. They (and the teams that will draw Russia in their qualification group) are sort of screwed by the poor position Russia had in the June 2018 ranking because of the lack of serious competition in the build-up to their World Cup.


pot 1

France 100,00%
Belgium 100,00%
Croatia 100,00%
England 99,97%
Portugal 98,41%
Switzerland 98,32%
Spain 93,25%
Denmark 89,60%
Sweden 50,50%
----------------------------------------- ------------
Germany 46,57%
Netherlands 8,71%
Italy 8,01%
Poland 4,20%
Wales 2,46%


pot 2

Wales 97,52%
Poland 95,80%
Italy 91,99%
Netherlands 91,29%
Austria 89,27%
Slovakia 73,88%
Romania 67,76%
Northern Ireland 61,75%
Republic of Ireland 54,21%
----------------------------------------- ------------
Germany 53,43%
Sweden 49,50%
Ukraine 19,34%
Serbia 16,83%
Denmark 10,40%
Iceland 9,68%
Spain 6,75%
Bosnia-Herzegovina 3,72%
Turkey 3,15%
Switzerland 1,68%
Portugal 1,59%
Scotland 0,29%
Montenegro 0,10%
England 0,03%
Czech Republic 0,03%
Greece 0,01%


pot 3

Turkey 91,12%
Bosnia-Herzegovina 90,49%
Iceland 90,32%
Scotland 87,66%
Serbia 81,39%
Ukraine 79,99%
Greece 64,63%
Montenegro 57,67%
Czech Republic 49,58%
----------------------------------------- ------------
Republic of Ireland 45,79%
Northern Ireland 38,25%
Romania 32,21%
Slovakia 26,11%
Bulgaria 20,19%
Russia 12,34%
Norway 11,52%
Austria 10,73%
Hungary 7,99%
Slovenia 1,93%
Albania 0,07%
Wales 0,02%


pot 4

Albania 99,93%
Finland 99,83%
Slovenia 98,07%
Hungary 92,01%
FYR Macedonia 88,76%
Norway 88,48%
Russia 87,66%
Bulgaria 79,81%
Czech Republic 50,39%
----------------------------------------- ------------
Montenegro 42,23%
Greece 35,36%
Scotland 12,05%
Belarus 11,01%
Bosnia-Herzegovina 5,79%
Turkey 5,73%
Serbia 1,78%
Ukraine 0,67%
Cyprus 0,27%
Luxembourg 0,11%
Romania 0,03%
Slovakia 0,01%
Israel 0,01%
Estonia 0,01%


pot 5

Faroe Islands 100,00%
Israel 99,99%
Estonia 99,99%
Georgia 99,97%
Luxembourg 99,89%
Cyprus 99,73%
Armenia 99,65%
Azerbaijan 92,01%
Belarus 88,99%
----------------------------------------- ------------
FYR Macedonia 11,24%
Kazakhstan 8,30%
Finland 0,17%
Lithuania 0,05%
Latvia 0,02%


pot 6

Andorra 100,00%
Moldova 100,00%
San Marino 100,00%
Kosovo 100,00%
Malta 100,00%
Liechtenstein 100,00%
Gibraltar 100,00%
Latvia 99,98%
Lithuania 99,95%
Kazakhstan 91,70%
----------------------------------------- ------------
Azerbaijan 7,99%
Armenia 0,35%
Georgia 0,03%



About me:

Software engineer, happily unmarried and non-religious. You won't find me on Twitter or other so called social media. Dutchman, joined the blog in March 2018.

21 comments:

  1. Lets hope that Denmarc can resolve their current issues and not be fined or even excluded.
    Dewi

    ReplyDelete
  2. It's highly unlikely that there'll be a group of 7 teams. That would require 14 qualification dates (same as a group of 8), as opposed to the 10 a group of 6 teams demand. Most likely UEFA will keep the Nations League and have 10 groups of 5 and 6 teams (5 each), similar to the qualification for 2020. 10 group winners will qualify directly and they'll try to persuade FIFA to allow the nations league to determine play-off teams, in a way still unknown (for instance, the six second placed teams that have higher ranking in the nations league; i've also read somewhere that UEFA was thinking of involving in the play-offs only teams from leagues A and B).
    If, with the world cup starting only in december, they found that they could have 14 dates, then I think it would be more likely that they'd have 6 groups of 8 teams and one of 7 - similar to the qualification for euro-2008.

    ReplyDelete
    Replies
    1. I agree. It's highly unlikely that there will be a group that requires four more dates than all the other. I predict ten groups, but I'd love to see seven groups instead - with top two from each group qualifying minus the worst second place team and no play-offs.

      Delete
  3. Why don't you use a simulation with fifa's new elo system? It is much simpler than club elo's. I can help you with it.

    ReplyDelete
    Replies
    1. Well, obviously the quality of the predictions using FIFA's elo ratings is not as good as with classic elo's predictions. At the moment classic elo has a prediction success rate of 52,2% starting from June 4th (which is quite low, I've seen prediction-success rates of close to 60% some years ago). FIFA only reaches 30,9% at the moment.
      Explainable of course because FIFA started with a very 'compressed' rating-list. Teams are at the moment very close in terms of ratings so FIFA predicts much more draws then they will in the future when the ratings will have settled with more 'normal' gaps.

      And Club Elo is not very hard to apply. It's only a probability-distribution with one parameter: elo's home team win expectancy. So thank you for the offer.

      Delete
  4. I suspect the qualifying might be run in 11 groups of 5, actually. 11 group winners auto-qualify. UNL League A and B group winners (but not C or D) play-off for the last two berths. Sucks for any league C or D team that finishes 2nd in their group in regular qualifiers behind a powerhouse but whaddayado. ¯\_(ツ)_/¯

    10 groups is possible but there's no neat system. You either have six teams in two-legged play-offs or twelve teams in single-legged semifinal-and-final play-offs. I find the latter preferable. In particular, I dislike the runner-up comparisons which seem more random than anything. Eliminating just one "worst" runner-up directly is borderline acceptable; the "worst" runner-up usually had a so-so record in their group. But tossing out the four "worst" runner-ups? Please no.

    Nine groups or fewer isn't possible if every team has to be included. Ideally, I'd like to see League D as a pre-qualifier leading to eight or nine groups of six (the bottom seven has to be eliminated in the first case, just one team in the second). The play-off berths would then be determined based on regular qualifiers alone. This would be unprecedented but everything featuring the Nations League is.

    ReplyDelete
    Replies
    1. If you have to integrate the Nations League into qualifying but don't want to give automatic spots to League C/D winners, I suggest the following:

      Run the Nations League groups in Sept/Oct/Nov. Then go with 14 groups (13 of 4, 1 of 3), play games in March/June/Sept/Oct. 14 group winners advance to a qualifying playoff, top runner-up to a consolation playoff round.

      Then run the 4 Nations League finals without those 15 teams in November, using same rules they will this time around. Also play a qualifying playoff between 14 group winners (7 qualify). Take the 7 qualifying losers, the 4 Nations League winners, and the 1 best runner-up and play another playoff round in March. 6 winners qualify to make 13 total.

      They won't do it (your 11 groups of 5 is more likely) but it's what I would propose.

      Delete
  5. If I understand correctly, the Euro quakifyinh matches aren't included in the simulations.
    Is that correct?
    Would you be able to include them in the near future (simulating the Euro Q draw and the scheduling as well of course)

    ReplyDelete
    Replies
    1. That's right Amir, I can only include matches in my simulations where both teams are known or both teams can be found using results of earlier matches.
      So as soon as the draw for the EURO 2020 qualifying has been made (December 2nd this year in Dublin) and the match schedule is published I will include these matches into my simulations.

      I am able to simulate the draw itself but I don't want to fill an assumed match-schedule with the draw-results, just to be able to include these simulated matches in my simulations.
      I think that (in that case) the simulation results will become dependent on too many simulated variables to be very insightful anymore.

      Delete
    2. It's your choice of course, but in term of variance, the old FIFA ranking simulation had much bigger variance than simulating the schedule with the current system.

      Delete
    3. The current simulation you made doesn't give to much information about the real precentages in my opinion, it is more a reflection of the current rankings...

      Delete
    4. That's right and you know why. The used I-factors for different types of matches in combination with the 600 in the Expected Win-formulae results in an inert ranking, even with these small starting gaps between team ratings.

      But you tell me: what are the real percentages ? What I definitely don't want to do is assume a schedule for a set of matches with unknown pairings and fill them with teams from a simulated draw. That is just too much simulated variables for my taste.

      Delete
  6. Hello Ed. So you´re using club elo to predict the results and FIFA´s elo to rank the teams? Because in the end, that´s what you´ll see in July 2019, FIFA teams ranked with FIFA´s new elo system.

    ReplyDelete
    Replies
    1. Yes exactly, I have a schedule of future matches, predict the result of each scheduled match using Club Elo and treat them as played matches to update the FIFA Elo points and thus the FIFA ranking. The ranking after all scheduled matches are predicted is the July 2019 ranking and the assumption is that that ranking will be used to seed the teams for the draw.

      The prediction of results for scheduled matches is based on a probability distribution so if you do that enough times you get a overview of the consequences on the resulting ranking and thus seeding probabilities.

      That's the whole trick.

      Delete
  7. I´m simulating the points earned by each team and their provisional positions in the ranking, AFTER the matches are played. I am also wondering what´s the factor for CONCACAF Nations League Qualifiers. FIFA did not make explicit the factors for the EAST ASIAN qualifiers and SOUTH ASIAN CHampionship.I guess they will be treated as scheduled friendlies (10 points). They appear as friendlies on the FIFA website. What do you think?

    ReplyDelete
    Replies
    1. Taliholic, a small remark about semantics first: I would call that what you do 'calculating points' instead of 'simulating points' because you don't use predicted results but realised results. But that's just a matter of taste.

      The I-factors for the tournaments you mentioned are not evident from the documentation FIFA provided. My assumption is that the CONCACAF Nations League qualifiers have a weight of 15, but it could also be 25 or 10.
      The smaller regional tournaments were in the past always treated by FIFA as friendlies and I don't think they've changed their attitude on that respect. So I use an I-factor of 10 for the Asian regional tournaments, as far as they are played inside the International Match Window (September 3rd to 11th). Otherwise it's only 5.

      Delete
  8. I feel that the presentation of data is a little misleading. For example, at a first glance, it looks like Ireland has a good chance of making Pot 2. But in reality, Germany or Sweden, whichever doesn't qualify for Pot 1, has a better chance. Ireland would most probably be in Pot 3 and Czech in Pot 4. Maybe, a table with list of Nations as rows and Pot 1 to Pot 6 as columns would better convey the data. Then in that list, the first 7 is the most probable for Pot 1, the next 7 for Pot 2 and so on. What do you think?

    ReplyDelete
    Replies
    1. This is the scenario then.
      Pot 1: Fight between Sweden and Germany. The loser moves to Pot 2.
      Pot 2: Three way fight between Romania, Northern Ireland and Republic of Ireland for two spots. Loser to Pot 3.
      Pot 3: Two from Greece, Montenegro and Czech Republic. Loser to Pot 4.
      Pot 4: Not much of a fight as two of Greece, Montenegro and Czech Republic are already in Pot 3!

      Delete
    2. Truthy. I agree. In that presentation you can also better see the inertia of the simulated seeding ranking with the new calculation method. Something like this (but with a better lay-out). I will use both presentations next time, thanks for the feed-back:
      France - 100,00% - 0,00% - 0,00% - 0,00% - 0,00% - 0,00%
      Belgium - 100,00% - 0,00% - 0,00% - 0,00% - 0,00% - 0,00%
      Croatia - 100,00% - 0,00% - 0,00% - 0,00% - 0,00% - 0,00%
      England - 99,97% - 0,03% - 0,00% - 0,00% - 0,00% - 0,00%
      Portugal - 98,41% - 1,59% - 0,00% - 0,00% - 0,00% - 0,00%
      Switzerland - 98,32% - 1,68% - 0,00% - 0,00% - 0,00% - 0,00%
      Spain - 93,25% - 6,75% - 0,00% - 0,00% - 0,00% - 0,00%
      Denmark - 89,60% - 10,40% - 0,00% - 0,00% - 0,00% - 0,00%
      Sweden - 50,50% - 49,50% - 0,00% - 0,00% - 0,00% - 0,00%
      Germany - 46,57% - 53,43% - 0,00% - 0,00% - 0,00% - 0,00%
      Netherlands - 8,71% - 91,29% - 0,00% - 0,00% - 0,00% - 0,00%
      Italy - 8,01% - 91,99% - 0,00% - 0,00% - 0,00% - 0,00%
      Poland - 4,20% - 95,80% - 0,00% - 0,00% - 0,00% - 0,00%
      Wales - 2,46% - 97,52% - 0,02% - 0,00% - 0,00% - 0,00%
      Austria - 0,00% - 89,27% - 10,73% - 0,00% - 0,00% - 0,00%
      Slovakia - 0,00% - 73,88% - 26,11% - 0,01% - 0,00% - 0,00%
      Romania - 0,00% - 67,76% - 32,21% - 0,03% - 0,00% - 0,00%
      Northern Ireland - 0,00% - 61,75% - 38,25% - 0,00% - 0,00% - 0,00%
      Republic of Ireland - 0,00% - 54,21% - 45,79% - 0,00% - 0,00% - 0,00%
      Ukraine - 0,00% - 19,34% - 79,99% - 0,67% - 0,00% - 0,00%
      Serbia - 0,00% - 16,83% - 81,39% - 1,78% - 0,00% - 0,00%
      Iceland - 0,00% - 9,68% - 90,32% - 0,00% - 0,00% - 0,00%
      Bosnia-Herzegovina - 0,00% - 3,72% - 90,49% - 5,79% - 0,00% - 0,00%
      Turkey - 0,00% - 3,15% - 91,12% - 5,73% - 0,00% - 0,00%
      Scotland - 0,00% - 0,29% - 87,66% - 12,05% - 0,00% - 0,00%
      Greece - 0,00% - 0,01% - 64,63% - 35,36% - 0,00% - 0,00%
      Montenegro - 0,00% - 0,10% - 57,67% - 42,23% - 0,00% - 0,00%
      Czech Republic - 0,00% - 0,03% - 49,58% - 50,39% - 0,00% - 0,00%
      Bulgaria - 0,00% - 0,00% - 20,19% - 79,81% - 0,00% - 0,00%
      Russia - 0,00% - 0,00% - 12,34% - 87,66% - 0,00% - 0,00%
      Norway - 0,00% - 0,00% - 11,52% - 88,48% - 0,00% - 0,00%
      Hungary - 0,00% - 0,00% - 7,99% - 92,01% - 0,00% - 0,00%
      Slovenia - 0,00% - 0,00% - 1,93% - 98,07% - 0,00% - 0,00%
      Albania - 0,00% - 0,00% - 0,07% - 99,93% - 0,00% - 0,00%
      Finland - 0,00% - 0,00% - 0,00% - 99,83% - 0,17% - 0,00%
      FYR Macedonia - 0,00% - 0,00% - 0,00% - 88,76% - 11,24% - 0,00%
      Belarus - 0,00% - 0,00% - 0,00% - 11,01% - 88,99% - 0,00%
      Cyprus - 0,00% - 0,00% - 0,00% - 0,27% - 99,73% - 0,00%
      Luxembourg - 0,00% - 0,00% - 0,00% - 0,11% - 99,89% - 0,00%
      Israel - 0,00% - 0,00% - 0,00% - 0,01% - 99,99% - 0,00%
      Estonia - 0,00% - 0,00% - 0,00% - 0,01% - 99,99% - 0,00%
      Faroe Islands - 0,00% - 0,00% - 0,00% - 0,00% - 100,00% - 0,00%
      Georgia - 0,00% - 0,00% - 0,00% - 0,00% - 99,97% - 0,03%
      Armenia - 0,00% - 0,00% - 0,00% - 0,00% - 99,65% - 0,35%
      Azerbaijan - 0,00% - 0,00% - 0,00% - 0,00% - 92,01% - 7,99%
      Kazakhstan - 0,00% - 0,00% - 0,00% - 0,00% - 8,30% - 91,70%
      Lithuania - 0,00% - 0,00% - 0,00% - 0,00% - 0,05% - 99,95%
      Latvia - 0,00% - 0,00% - 0,00% - 0,00% - 0,02% - 99,98%
      Andorra - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      Moldova - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      San Marino - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      Kosovo - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      Malta - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      Liechtenstein - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%
      Gibraltar - 0,00% - 0,00% - 0,00% - 0,00% - 0,00% - 100,00%

      Delete
  9. Final decision about 2022 FIFA World Cup Qualifying in UEFA zone will be confirmed by UEFA Executive Committee.

    ReplyDelete
    Replies
    1. When?
      I would give the holder spot back for France in 2022. The Winner of 2019 Nations League. Leaving to 53 natons. 10 grop winners and the best places non qualified in 2021 Nations League.

      Delete