Latest updates

Check the Important info page for latest updates! (8 May 2018)
Twitter

Monday, June 11, 2018

FIFA ranking: July 2018 probable ranking (new method)

FIFA has adopted a new calculation method for their ranking. It will be used for the very next ranking, the first ranking after the World Cup, to be published on July 19th.

I have made a probable July 2018 ranking, applying the new calculation method for all matches played from June 4th onwards. See the probable results for the predicted results used. 


These are the rules/assumptions for this calculation:
- as start rating for each team I use the unrounded June FIFA-points;
- all matches from June 4th onwards count for this ranking;
- as there is not an International Match Calendar window in June the friendlies get an I-factor of only 5;
- the matches in the World Cup groups and Round of 16 get an I-factor of 50;
- the matches from the Quarter finals onwards get an I-factor of 60;
- the losing teams in all knock-out matches in the World Cup preserve their points from before the loss.

Biggest points exchange occurs in the predicted World Cup final Brazil - Germany (1:0): 34.18 points.
The preparatory friendlies deliver at maximum 3.24 points to exchange.

Where normally the July ranking after a World Cup is quite shaken-up (compare for instance with the old probable July ranking I published last Saturday) this looks like an oasis of peace and tranquility :)


Probable July 2018 rank - Team - Probable July 2018 points - +/- Ranking - +/- Points


1 Germany 1565 0 7
2 Brazil 1526 0 95
3 Belgium 1289 0 -9
4 Argentina 1278 1 37
5 Portugal 1266 -1 -8
6 France 1251 1 53
7 Switzerland 1204 -1 5
8 Spain 1185 2 59
9 Poland 1163 -1 -20
10 Chile 1136 -1 1
11 England 1116 1 65
12 Peru 1109 -1 -16
13 Colombia 1054 3 68
14 Uruguay 1050 0 32
14 Denmark 1050 -2 -1
16 Mexico 984 -1 -5
17 Netherlands 981 0 0
18 Wales 953 0 0
19 Italy 951 0 0
20 Croatia 944 0 -1
21 Iceland 910 1 2
22 USA 874 3 1
23 Sweden 870 1 -10
24 Tunisia 868 -3 -42
25 Costa Rica 859 -2 -25
26 Austria 844 0 -1
27 Northern Ireland 803 2 0
28 Slovakia 801 0 -3
29 Senegal 796 -2 -42
30 Romania 783 0 1
31 Republic of Ireland 777 0 0
32 Paraguay 772 0 -1
33 Venezuela 755 0 0
34 Serbia 746 0 -5
35 Ukraine 733 0 0
36 Congo DR 706 2 0
37 Turkey 705 1 -1
38 Iran 695 -1 -13
39 Australia 694 -3 -24
40 Bosnia-Herzegovina 693 0 0
41 Morocco 679 0 -7
42 Scotland 660 0 0
43 Montenegro 652 0 0
44 Greece 650 0 0
45 Czech Republic 643 1 2
46 Ghana 625 1 1
47 Bulgaria 614 2 0
47 Cameroon 614 2 0
49 Hungary 609 2 -3
50 Burkina Faso 604 2 0
51 Egypt 599 -6 -50
52 Norway 598 1 2
53 Nigeria 589 -5 -29
54 Jamaica 588 0 0
55 Panama 573 0 2
56 Slovenia 558 0 0
57 Korea Republic 550 0 6
58 Albania 538 0 0
59 Bolivia 535 0 -1
60 Ecuador 524 0 0
60 Japan 524 1 3
62 Russia 509 8 52
63 Honduras 503 -1 0
64 Finland 487 -1 1
65 Mali 484 -1 0
66 Cape Verde Islands 478 -1 0
67 Algeria 474 -1 0
68 Cote d'Ivoire 463 0 0
68 Guinea 463 0 0
70 FYR Macedonia 451 1 0
71 Saudi Arabia 443 -4 -22
72 El Salvador 442 0 0
73 Syria 440 0 0
74 South Africa 431 0 3
75 China PR 423 0 0
76 Zambia 419 0 -1
77 United Arab Emirates 418 0 0
78 Belarus 416 0 -1
79 Canada 411 0 0
79 Lebanon 411 0 0
81 Curacao 408 0 0
82 Uganda 403 0 0
83 Congo 393 0 0
84 Luxembourg 392 1 2
85 Oman 391 -1 0
86 Gabon 390 -1 0
87 Cyprus 384 0 0
88 Benin 379 0 0
89 Iraq 377 0 0
90 Faroe Islands 376 0 0
91 Trinidad and Tobago 370 0 0
92 Kyrgyzstan 363 0 0
93 Israel 362 0 0
94 Estonia 358 0 -1
95 Uzbekistan 354 0 0
96 India 351 1 1
97 Georgia 349 -1 -3
98 Qatar 348 0 0
99 Palestine 347 0 0
100 Armenia 337 0 -1
101 Libya 336 0 0
102 Vietnam 334 0 0
103 Niger 323 0 0
104 Haiti 322 0 0
105 Azerbaijan 321 0 0
106 Mauritania 304 1 0
107 Korea DPR 303 1 0
107 Central African Republic 303 1 0
107 Madagascar 303 -1 -5
110 Jordan 296 0 0
111 Sierra Leone 292 0 0
112 Bahrain 289 1 0
112 Kenya 289 0 -2
114 Mozambique 282 0 0
115 Philippines 280 0 0
116 Kazakhstan 276 1 3
117 Namibia 275 -1 -2
118 Zimbabwe 267 0 2
119 Tajikistan 264 0 0
120 New Zealand 261 0 5
121 Guinea-Bissau 255 0 0
122 Thailand 248 0 0
123 Malawi 247 0 0
124 Chinese Taipei 242 -1 -5
125 Togo 241 0 0
126 Antigua and Barbuda 239 0 0
127 Lithuania 238 -1 -1
128 Sudan 232 0 0
129 Turkmenistan 229 1 0
129 Andorra 229 1 0
131 Latvia 228 -2 -2
131 Nicaragua 228 1 0
133 Yemen 227 0 0
134 Swaziland 220 0 -2
135 St. Kitts and Nevis 218 0 0
136 Rwanda 216 0 0
137 Angola 209 0 0
138 Botswana 206 1 1
138 Myanmar 206 0 0
140 Tanzania 200 0 0
141 Kosovo 197 0 0
142 Hong Kong 195 0 0
143 Equatorial Guinea 190 0 0
143 Solomon Islands 190 0 0
145 Afghanistan 188 0 0
146 Guatemala 180 0 0
147 Lesotho 179 0 4
148 Burundi 174 0 0
149 Comoros 172 0 0
150 Maldives 171 0 0
151 Ethiopia 166 0 0
152 Dominican Republic 162 0 0
153 Suriname 154 0 0
154 New Caledonia 150 0 0
155 Mauritius 140 0 0
156 South Sudan 132 0 0
157 Tahiti 130 0 0
158 Liberia 129 0 0
159 Kuwait 128 0 0
160 Barbados 127 0 0
161 Nepal 118 0 0
162 Vanuatu 117 0 0
163 Belize 115 0 0
164 Indonesia 111 0 0
165 Fiji 104 0 0
166 Papua New Guinea 103 0 0
166 Cambodia 103 0 0
168 Grenada 102 0 0
169 Singapore 99 0 0
170 St. Lucia 95 0 0
171 Malaysia 93 0 0
172 Gambia 92 0 0
173 Moldova 89 2 1
173 St. Vincent and the Grenadines 89 0 0
173 Puerto Rico 89 0 0
176 Chad 88 -1 0
177 Dominica 86 0 0
178 Bermuda 82 0 0
178 Laos 82 0 0
180 Liechtenstein 80 0 0
181 Cuba 75 0 0
182 Guyana 73 0 0
183 Bhutan 68 0 0
184 Malta 65 0 0
185 Macau 60 0 0
186 Sao Tomé e Príncipe 51 0 0
186 Mongolia 51 0 0
188 Seychelles 48 0 0
188 Aruba 48 0 0
190 Guam 45 0 0
190 Timor-Leste 45 0 0
192 Cook Islands 38 0 0
192 American Samoa 38 0 0
194 Bangladesh 35 0 0
195 Gibraltar 34 0 0
195 Brunei Darussalam 34 0 0
197 Samoa 32 0 0
197 Djibouti 32 0 0
199 US Virgin Islands 18 0 0
200 Sri Lanka 17 0 0
201 Pakistan 10 0 0
202 Cayman Islands 9 0 0
203 San Marino 8 0 0
204 British Virgin Islands 4 0 0
204 Montserrat 4 0 0
206 Eritrea 0 0 0
206 Bahamas 0 0 0
206 Somalia 0 0 0
206 Tonga 0 0 0
206 Turks and Caicos Islands 0 0 0
206 Anguilla 0 0 0

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.

52 comments:

  1. So now teams who don't play don't receive a change in their points? What made the points change with the old formula? Thanks

    ReplyDelete
    Replies
    1. Devaluation of old results: when past results fell into the next timeframe (with a lower weight) timeframe averages changed and thus the FIFA points changed, even if a team did not play a match in the last month.

      Delete
    2. I think the question is referred to the new system, and there will be no more timeframe averages with the new system. So the answer seems to be no, teams who don't play do not change their points (as the ELOrating system). However we need further details to know if teams inactive for several years will be maintained in the ranking forever or, after some year of inactivity they will be deleted. And to know what will be the method used to calculate the starting rating for future new associated nations.

      Delete
    3. You're right, I only answered the second part of the question.
      Teams who don't play a match stay on the same points-total. That's inherent to the used elo-like system.

      Delete
    4. OK thanks; also what was the timeframe prior to the ratings change? Did it take into account the last 2 years? The last 3 years? Thanks again.

      Delete
    5. Matches played in the last 4 years were taken into account, with descending weights: 1st year 1.0 weight, 2nd year 0.5, 3rd year 0.3 and 4th year 0.2 weight.

      Delete
    6. Thanks again, could you please answer the question I asked on the AFC Asian Cup predictions?

      Delete
  2. Under the new system as it appears so far, do you think it's possible that teams could end up on negative ranking points?

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. When a team is already that low on points it will lose very little or no points at all when losing the next match. But yes, in theory teams could end below zero with this system.
      In the classic elo system the initial value for Palau for instance was 500 while this team now has 402 points.

      Now we start with 6 teams on 0 points so this situation might occur (very) soon. I suppose FIFA applies an extra constraint for that situation, though. They probably don't want to assign negative points to teams, I guess :)

      Delete
  3. The K-factor is really small even for the world cup. It is strange that FIFA went from "last year means everything" to this slow adjusting rankings.

    ReplyDelete
  4. Looks unfair if starting "points before game" will be taken from the last rankings calculated with old method
    New method means that points earning and movements in the rankings will be minor if compared to old system.
    That's why recalculation "from zero" is necessary. Maybe from the point of time when previous method was introduced in 2006 or even earlier

    ReplyDelete
  5. With the classic elo-rating system they use a scientific guess as starting value for new teams. After some 30 matches the elo-rating for the new team by design converges to a new steady state.
    So something like that is to be expected for the currently used June rankings as starting values. After some 30 matches the rating will have been converged to a new steady state. Only the figure of 30 matches will be different as FIFA applies not a strict zero-sum approach like elo does.

    ReplyDelete
  6. If the new system allows only very slow growth, then no more Poland, Switzerland, Colombia and maybe even Belgium in the World Cup seeds, right?

    ReplyDelete
    Replies
    1. There will be 16 seeds starting from 2026 anyway.
      Since the rankings are a lot better now, a seeded team in 2022 would need to be a lot more consistent, not just have a good run towards the end of qualifying.

      Delete
    2. Thanks. And also, what is

      We = 1 / 10(-dr/600) + 1)

      supposed to mean? There's one closing bracket too many, and multiple interpretations are possible. It's in the FIFA document linked to above (not sure how to link to it myself).

      Delete
    3. That's a typo by FIFA. Right in front of the 10 should be another (.

      Delete
    4. It's not just another (.
      The formula for Elo is:
      We = 1 / (10^(-dr/600) + 1)

      The 600 is just a scaling factor, in the original Elo it's 400.

      Delete
  7. So, the top ranked teams now gain points even when they lose matches, right?

    ReplyDelete
    Replies
    1. No, you never gain points by losing

      Delete
    2. Seeing they won't lose points in KO matches if they lose, it basically is the same as gaining points.

      Delete
    3. Imagine Spain beaten by Iran in the last 16, they wouldn't lose a single point...

      Delete
  8. This comment has been removed by the author.

    ReplyDelete
  9. Ed, is this completely the same as the World Football Elo Ratings that has already existed?

    ReplyDelete
    Replies
    1. Albert,
      no, this is not the same as the 'classic' elo ratings. See here for a brief description of the differences.

      Delete
  10. A question: here you wrote "all matches played from 4th june onward", but in the probable results page there are results from 11 June. What is the correct date to start with the new calculation?

    ReplyDelete
    Replies
    1. Jos, the match results for the matches from June 4th until June 10th were known at the time of writing, but of course the new formula is applied on them as well.
      In the probable results page I presented the predictions for the scheduled matches from June 11th onwards. These are also calculated with the new formula.

      So the correct date is June 4th. Which is of course an assumption. We don't know yet what FIFA exactly decides to do.

      Delete
    2. oh, yes of course. Stupid question the mine. Thanks!

      Delete
  11. Another question, how do you consider results for matches ended in extra-times (i.e. Zambia-Zimbabwe)? A draw for the losing team?

    ReplyDelete
    Replies
    1. By default: when there is extra time played the result after extra time is the result of the match. So your example was a win for Zimbabwe.
      When it is a single knock-out match a decision has to be reached, so when it is still equal after extra time penalties are taken. But for this elo-like calculation the result is in this case a draw.

      Delete
    2. Is that confirmed by FIFA? The current system (and the previous one IIRC) treated a PSO loss as a draw, but a PSO win as halfway between a win and a draw.

      I wish they used one notable feature of the women's ranking, that the margin of victory matters. If Belgium beats Andorra 1-0, that's an underperformance, and their rating should drop a little (and Andorra's should go up).

      Delete
    3. No, that's not confirmed by FIFA. But in their document it says there's only a loss, a draw or a win. Not something like a pso win.
      Besides, the classic elo rating works exactly in the way I described with extra time en penalty shootout situations.
      Because FIFA doesn't bother about such details in their description, the only way to know for sure how they will handle such situations is by proof, i.e. the July 2018 ranking.

      Delete
    4. OK. I think it's entirely possible they do something stupid like consider a PSO result to be a full win/loss.
      If they do consider a game that goes to PSO a draw, hopefully the stronger team would be protected from losing points in late stage knockout games, as they should not be treated more harshly for a draw than they are a loss!

      Delete
  12. Still did't get it clearly:
    If a team plays more than one game between two ranking releases, will "points before game" for the second (and maybe following) game in a frame be recalculated immediatelly after previous one? Or previous rankings will be used for all games in a frame?

    ReplyDelete
  13. The elo-ratings of the two teams are updated immediately after a match is played. This updated rating is starting value for the next match a team plays, even if that match is played before the next ranking is published.
    The ranking is published as the descending order of the ratings as valid on the publishing date of that ranking.

    ReplyDelete
    Replies
    1. Ed, thank you for explanation.
      This makes prediction and preview algorythms more complicated, right? You need to take into account previous games of the team and their rivals.
      For example how do you estimate maximum of points that one team (A) can gather during group stage on World Cup?
      For third game in group you may assume that the opponent (B) won two previous games, but it means that the second game's opponent (C) lost their game to B? Or what will be optimal assumption?
      Could you please clarify?

      Delete
    2. Vitalii, prediction of scheduled matches is not that hard as there is always a schedule of matches. So you know the teams in the match to predict and you can find the last match where the teams played in and the adapted team-rating as a result of that match. So the start value for each team in the match to predict can be determined. And that's the only dependency with other matches. The rest of the formula is straightforward to calculate.
      Calculating the minimum and maximum points becomes indeed a lot more complicated though. Maybe only a brute force approach can give the correct answer to this question. I haven't researched this problem yet.

      Delete
    3. I'm not entirely sure that's the case actually. It is possible that FIFA will use the last ranking as a base for the calculation, even though it is hinted otherwise on the article they published.

      Calculating the maximum and minimum points is indeed a tough question is general. In practice there aren't so many matches and assigning wins for all opponents of team X for the matches before the match with X, and wins for all of X matches would give the maximum number of points in most cases.

      Delete
  14. A consideration after better reading the Fifa statement: "One of the main advantages of SUM is that it allows for a smooth transition from the current ranking formula to the new one without displacement of teams in the existing ranking table. The current FIFA / Coca-Cola World Ranking will be replaced seamlessly by the new SUM formula without gain or loss of member association rank positions."

    They do not speak about points but only positions, so I think maybe there will be some kind of conversion from actual points into new different ratings

    ReplyDelete
  15. Some (big) upsets occured already in the WorldCup. Biggest points-swaps so far:
    1. Germany-Mexico: 44,98 pts. from Germany to Mexico
    2. Costa Rica-Serbia: 31,12 pts. from Costa Rica to Serbia
    3. Peru-Denmark: 28,37 pts. from Peru to Denmark
    4. Russia-Saudi Arabia: 25,36 pts. from SA to Russia
    5. Morocco-IR Iran: 24,09 pts. from Morocco to IR Iran

    But I'm still coming to grips with the new calculation method, so I could be wrong on those results. Ed? ;-)

    ReplyDelete
    Replies
    1. Looking good, Tobcoach ! I have 23,94 pts swapped for Morocco-Iran, using the unrounded June-ranking as start values and previous friendlies from June 4th onwards. The rest is spot on.

      As if I know how FIFA is going to fill in all the details :)

      Delete
    2. Glad to see I'm doing something right... ;-)

      For the Morocco-Iran game: you have a starting difference in points between Morocco and Iran of about 22,1 pts? For me it's 19,04 pts. and I have the following friendly results calculated:

      Starting points Iran: 708,35 (FIFA Ranking June)
      Friendly 08/06/18 vs Lithuania: +0,71 pts.
      Value before Morocco game: 709,06

      Starting points Morocco: 685,86
      Friendly 04/06/18 vs Slovakia: +3,06 pts.
      Friendly 09/06/18 vs Estonia: +1,10 pts.
      Value before Iran game: 690,02 pts.

      Difference before game: 19,04 pts.

      I'm guessing you didn't include the Slovakia game of Morocco in the calculations? That would make the difference between Iran and Morocco 22,10 points before their game.

      Delete
    3. You are right, Tobcoach. I had skipped the match MAR-SVK for some unknown reason. Thanks for the feedback !

      Delete
    4. Japan +42.74, new #2 in upsets.

      Delete
    5. It's raining upsets, Senegal +39.5!

      Delete
  16. These calcuations are all based on K's of 5 to 60, in FIFA's June 10 version of the background document.

    In the June 14 they (hopefully erroneously) use 0.5 to 6.

    See https://resources.fifa.com/image/upload/revision-of-the-fifa-coca-cola-world-ranking.pdf?cloudid=fzltr4s8tz3v3vy0aqo1 that is NOW referenced in https://www.fifa.com/about-fifa/news/y=2018/m=6/news=2026-fifa-world-cuptm-fifa-council-designates-bids-for-final-voting-by-the-fifa-.html

    ReplyDelete
    Replies
    1. I seriously and genuinly hope that's a typo in the FIFA documents. If they use 0.5 to 6 for the match importance factor, they might as well lock the ranking as of June 2018.

      Delete
    2. We all gave for sure that Fifa will use actual points, but as I said before it is not sure, reading only the text in the pdf document and leaving out the speculations appeared on websites. Constants should be valid if another scale for ratings will be used mantaining the same positions of the June ranking, and I am now finally convinced that this will be the case. There are two supporting elements for my theory: 1) they use a different scale in the formula (600 instead of 400) that causes smaller movements respect the formula normally used in the ELO website. Using actual points, with this formula and constants, positions will be pretty locked 2) if they will use actual points, a lot of teams will have in a short while negative points. I don't think FIFA would appreciate a ranking that looks like this.
      They wrote that simulations are done so I think they are fully aware of all that.
      We have to wait details, for this and other issues that I resume:
      - not losing points for defeat in knock out rounds in final tournaments. It is not specified "major tournaments", so theoretically this definition can involve all tournaments in neutral ground, included CECAFA Cup, EAFF Championship finals and so on, as sample. The fact there will be a different I constant for final stages from QF in the major tournaments, is a separate aspect not necessarily coincident;
      -extra time and penalty shoot out. In the actual system, a win is awarded 2 points, less than a normal win. If they want to mantain the actual philosophy, they should use 0.75 for the winner and 0.5 for the loser, but of course they should change something;
      -What starting rating will be assigned to future new associated federations?;
      -There will be a minimum required number of matches played to be included in the ranking like the women's ranking?
      - OFC nations cup will have finally the same importance factor of the others continental championships? And if so, the I=40 factor will be used from semifinals (since ther are not quarter finals)? I think probably they will continue to consider it as World Cup qualification (I=25);

      Delete
    3. The low k values (0.5 to 6.0) and the large scaling constant (600) can't both be true at the same time. The k values imply that the initial point gaps between elite and average teams would be a few dozen points, perhaps, and the first and the last team would be separated by circa one hundred points. But in that case you'd get almost the same points whoever you beat. If Germany plays San Marino, their Wₑ ≈ 0.6 (for comparison, it's Wₑ ≈ 0.999 at eloratings.net).

      I'm not entirely sure what would happen in the long run. Even with such asinine values, the ranking would eventually converge into *something*. Even San Marino would, defeat by defeat, trail further and further behind all the other UEFA teams until there would be a balance where the occasional draw they get is enough to offset the 50 defeats before. But my hunch is that it'd converge into a 20-year average strength rather than anything resembling current strength.

      If the initial gaps are what they are now, the ranking will never change. For example, Spain (current # 10) would need to beat Germany (current # 1) 65 times in a row to overtake them with the highest possible match weight (5.0; I'm not using 6.0 because the losing side doesn't lose any points in those games). This would be 148 times in a row for Armenia (current # 100) and 182 times in a row for Tonga (currently last). If we use the qualifier weighting of 2.5 instead, these numbers will be 129 (Spain), 296 (Armenia) and 364 (Tonga).

      What's somewhat overlooked in the comments of this blog is that even the k values 5 to 60 are (way) too low for the current gaps. A 45-point swing in a World Cup game may look impressive, but most teams only get to play qualifiers where the highest possible gain is 25 points and a normal one between 10 and 15. For example, you need to gain ca. 200 points to climb 10 places within UEFA (approximately one WCq pot). A current lower-middle-echelon team (think Albania, Finland, Israel, etc.) would need to win pretty much all their Nations League games + EURO qualifiers, ie. all their competitive fixtures for two years, to do that.

      Delete
  17. I hate this new method...

    ReplyDelete
  18. On an unrelated note, I just realized that the Oceania Nations Cup will remain the only major tournament with the classical, straightforward format where the number of participants is a power of 2 (8, 16, 32, etc) and half of them are eliminated after each round. This will no longer be the case with the World Cup, the European, Asian and African championships after their expansions, while the Confederations Cup might well be abolished.

    ReplyDelete
    Replies
    1. Actually, I just found out that the CONCACAF Gold Cup will be expanded from 12 to 16 teams. But still, the traditional formula is no longer the norm.

      Delete