It does. It's from yesterday (10 of June) under the heading "2026 FIFA World Cup™: FIFA Council designates bids for final voting by the FIFA..." as it wasn't the only news item from the council meeting.
If you don't understand ELO systems or its associated math well enough to understand that the issue that tweet pointed to isn't a methodological or substantive one (like the ones we have been discussing here) but purely one of poor writing of the PR departement then I'm not sure it's any use to continue this discussion.
I don't think that's wrong though, is it? Teams that qualified for the WC must have gotten better results playing the same opponents as teams from their qualifying group that didn't qualify. So they earned the higher ranking. There is a bit of a snowball effect where teams that do better, get higher seeds and easier groups, who then therefore are likely to continue doing better. But that will happen whenever you have seeding. Its not the method of seeding that causes that (see for e.g. the UEFA Champions League which ranks teams purely based on results).
If FIFA are keeping the current Rankings points as the starting points (probably so that historical numbers have meaning) we will soon end up with teams on negative rankings points. eg. If Tonga lose an OFC Nations Cup qualifier to PNG they will then be on -10 points. Not that its a big issue stats wise, it just looks bad for the countries involved. Teams that rarely ever get positive results will find themselves permanently in the negative. If they gave all nations a base starting point of 200 points at least the ultra weak nations would stay on the positive side of things for many years. The match importance weightings also seem too low, changes in the ranks will take too long to match reality.
You are right, in that my alarm at the new system was purely down to the extreme swings that the published formula would have effected. I was assuming that was by design - one of the stated aims is to emphasise more important matches. It turns out just to be an error in the formula. Seems pretty substantive to me. Luckily somebody spotted it. As I said at the outset, I'm not an expert on ranking systems. So I was having to do the calculations to find out what the outcome would actually look like. What's the basis of the 600 in the formula?
"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" It's not ranking points here - it's positions. J
Given that the formula looked very similar to the one used in elorankings.net I don't think many people would have been confused by what is a problem with formatting from word to pdf. The 600 is just a number that they probably chose from a list of options - I'm not sure why elo use 400 in their version. The most likely reason is that the multiple for matches are on average less, so the "die off" in this function needs to be lowered to balance - but that's just a guess. J
It was just an error by the PR departement, I might not have much faith in FIFA but this I'm sure they would have gotten right where it mattered. FIFA have used an ELO rating for the Women's game for years and they have a similar term that is correctly used. Why they picked 600 I have no idea, I expect they had some good reason. What the factor does I'm not sure I can explain that well but it has to do with how much the starting ratings difference between the teams affect the final rating. A term of 600 at FIFA should mean that there are smaller point swings than if the term was 400 (as it is on eloratings.net). But I'm not 100% I got that right.
Yea but you won't have a raking position if you don't have ranking points to sort/differentiate teams. And the Elo system needs starting points to work. So either way you are going to have some built in methodological issues with the rankings.
I just fed the data in as if this new ranking formula had started after the June 2014 FIFA Ranking. I had to make some assumptions so I used the actual FIFA Ranking points (June 2014) as the starting values for the new ranking (which probably isn't going to be correct) and the not losing points in the knockout stages of a tournament (Penalty shoot-outs I considered draws, so the team that progressed sometimes should have lost points, this may also be wrong, maybe shoot-outs are just considered wins and losses). Anyway this is what the current ranking would look like if the formula had started June 6th 2014. (Apologies for the names of some countries, it's just how I had the records at hand) Position Team Points 1 Spain 1343 2 Germany 1308 3 Brazil 1264 4 Argentina 1204 5 Portugal 1164 6 Belgium 1121 7 Colombia 1113 8 England 1093 9 Switzerland 1076 10 France 1071 11 Italy 1062 12 Uruguay 1044 13 Chile 1038 14 Mexico 1012 15 Netherlands 990 16 United States 959 17 Croatia 922 18 Greece 846 19 Costa Rica 840 20 Ukraine 836 21 Peru 831 22 Denmark 814 23 Ecuador 798 24 Bosnia and Herzegovina 788 25 Russia 782 26 Sweden 777 27 Scotland 760 28 Serbia 746 29 Romania 745 30 Turkey 743 31 Wales 742 32 Panama 731 33 Slovenia 726 34 Poland 720 35 Iran 714 36 Czech Rep. 714 37 Venezuela 711 38 Iceland 710 39 Ghana 706 40 Slovakia 705 41 Egypt 703 42 Austria 699 43 Algeria 699 44 Tunisia 690 45 Paraguay 690 46 Ivory Coast 678 47 Japan 665 48 Nigeria 655 49 Honduras 645 50 Australia 639 51 Senegal 637 52 Ireland 634 53 Hungary 630 54 Cameroon 627 55 South Korea 624 56 Morocco 608 57 Bolivia 573 58 Norway 569 59 Dem. Rep. of Congo 563 60 Burkina Faso 562 61 Cape Verde 562 62 Northern Ireland 560 63 Montenegro 560 64 Armenia 559 65 Albania 557 66 Mali 536 67 Jamaica 530 68 South Africa 521 69 Finland 519 70 Uzbekistan 517 71 Guinea 516 72 Sierra Leone 496 73 Trinidad and Tobago 493 74 Bulgaria 475 75 Zambia 471 76 United Arab Emirates 470 77 Haiti 470 78 Libya 468 79 Saudi Arabia 466 80 Uganda 459 81 El Salvador 451 82 Israel 446 83 Jordan 445 84 China 426 85 Belarus 426 86 Congo 418 87 Oman 415 88 Macedonia 411 89 Canada 396 90 Gabon 390 91 Benin 390 92 Iraq 389 93 Estonia 382 94 New Zealand 372 95 Zimbabwe 368 96 Azerbaijan 362 97 Qatar 358 98 Georgia 353 99 Togo 350 100 Palestine 349 101 Equatorial Guinea 342 102 Syria 329 103 Cuba 327 104 Lithuania 326 105 Central African Republic 325 106 Botswana 316 107 Guatemala 312 108 Kenya 305 109 Angola 299 110 Mozambique 299 111 Lebanon 298 112 Bahrain 298 113 Ethiopia 292 114 Luxembourg 291 115 Niger 284 116 Tanzania 283 117 Namibia 276 118 Liberia 276 119 Rwanda 274 120 Malawi 271 121 Latvia 269 122 Kuwait 269 123 Vietnam 266 124 Kazakhstan 258 125 Moldova 253 126 Cyprus 247 127 North Korea 243 128 Mauritania 238 129 Burundi 235 130 Madagascar 234 131 Sudan 230 132 Kyrgyzstan 230 133 Philippines 230 134 Thailand 227 135 Chad 214 136 Guinea-Bissau 212 137 Turkmenistan 207 138 Antigua and Barbuda 202 139 Saint Kitts and Nevis 201 140 Tajikistan 198 141 Aruba 196 142 Dominican Republic 189 143 Surinam 186 144 Afghanistan 184 145 India 182 146 New Caledonia 180 147 Nicaragua 176 148 Saint Lucia 172 149 Lesotho 158 150 St. Vincent & Grenadines 156 151 Curacao 155 152 Swaziland 152 153 Malta 151 154 Gambia 151 155 Hong Kong 148 156 Indonesia 148 157 Faroe Islands 144 158 Grenada 138 159 Belize 136 160 Tahiti 133 161 Guyana 131 162 Myanmar 128 163 Barbados 128 164 Comoros 110 165 Yemen 104 166 Puerto Rico 104 167 Maldives 91 168 Liechtenstein 89 169 Singapore 89 170 Pakistan 87 171 Montserrat 85 172 Dominica 82 173 Bermuda 82 174 Malaysia 80 175 Solomon Islands 79 176 Sao Tome e Principe 74 177 Taiwan 65 178 Mauritius 63 179 Guam 61 180 Nepal 53 181 Papua New Guinea 48 182 Vanuatu 41 183 South Sudan 39 184 Kosovo 31 185 American Samoa 29 186 Cayman Islands 26 187 Bangladesh 22 188 Sri Lanka 22 189 Cook Islands 18 190 Seychelles 18 191 Bahamas 17 192 Fiji 17 193 Laos 16 194 Brunei 2 195 Eritrea 1 196 Western Samoa -2 197 US Virgin Islands -5 198 British Virgin Islands -7 199 Mongolia -7 200 Andorra -8 201 Somalia -13 202 Gibraltar -20 203 Turks and Caicos -21 204 Cambodia -22 205 Tonga -22 206 East Timor -34 207 San Marino -44 208 Djibouti -45 209 Macao -49 210 Bhutan -58 211 Anguilla -59
The base of 600 in the formula doesn't mean anything mathematically, it is just a scaling factor. It does mean that K factor of 60 is equal to 40 on the original Elo ratings (since the scaling factor of it is 400).
You are half right... The base of 600 in the formula doesn't mean anything mathematically, it is just a scaling factor. So a difference of 400 on the original Elo is equal to a difference of 600 here. It does mean that K factor of 60 is equal to 40 on the original Elo ratings (since the scaling factor of it is 400).
Th starting values doesn't mean anything after enough matches, Elo converges to specific rankings given the K values no matter what are the initial points.
There are ways of dealing with the zero point problem (ratings floor is one). There is however no way to use an ELO system with just a ranking list, so there will be a points list too.
So looking at the old weighting of 4-3-2.5-1 and the new weighting of 60-35-25-10 (just comparing like for like, the new rating have more levels) that means that most every match outside the WC got a somewhat lower weighting, relatively speaking. Takes about 30+ games for the ratings stop being provisional usually right? (is that just a leftover from when it was considered a normal distribution???) So any issues should be gone in 3-5 years then. Just in time for them to tinker a bit more before the next WC.
It really doesn't - one side is related to exponentials, the other is linear. For a initial ranking difference of +5 - a win under 400/40 gains you 19.7 points, under 600/60 you gain 29.7. Even if you changed the 40 and 60 so the +5 case gave the same result, the impact wouldn't be equal for any other starting difference (and you would have to drop from 40 to 39.8ish to do that). That is not all contributing to a general lowering in K values. Remember that K is either the same as before or lower for a one goal victory, but in Elo the K factor is boosted for wins by more than one goal - so in the FIFA system the K factors is always less than or equal to Elo - often significantly less. The higher scaling factor (600>400) makes the We factor closer to 0.5 -> (1-We) (based win score point gain) is more for the higher ranked team - but less for a win by a lower ranked team. So both changes to the scaling actually downplay the impact of surprise victories (while only the K change limits the impact on expected wins), and means draws will see less change in numbers J
Trust me on this, I'm a mathematician... For a initial ranking difference of +5 - a win under 400/40 gains you 19.7 points, under 600/60 you gain 19.7*1.5 for an initial difference of 1.5*19.7. That means that if you take an Elo rating under 400/40, and multiple each rating by 1.5 and then apply N matches under 600/60, it is the same as applying these matches under 400/40 and then multiplying each rating by 1.5. In simplified terms, 600/60 is equivalent to 400/40 since you would get exactly the same order if you start all teams from the same ratings, no matter if you do it under 400/40 or under 600/60. This is not necessary the case for 400/60 for example. The reason for that is that it is true that the scaling is on the component, but it is re-scaling the difference between the teams inside the formula.
"The ranking system is set to be updated following the 2018 World Cup as announced by FIFA on 10 June 2018. The calculation method to be adopted will be closely modeled after the Elo rating system and rankings of its member associations will be updated on a game-by-game basis. The weighting designated for each confederation for ranking purposes will be abolished." So will the new method already be used for results from Russia 2018 games, the June ranking being the final one based on the old formula, or will they start after, the July ranking being the final one based on the old formula. I'd say the latter one would be better. Brazil 2014 results have depreciated in the June ranking while in the July ranking those results would have been replaced by Russia 2018. Seeing Russia have been in a freefall in the rankings because as hosts they could only play friendlies it would be very harsh to not give them a chance to improve their ranking first. The new method isn't very responsive and it would take them a long time to recover.
Makes we wonder why you are on a message board and not out solving crimes. But then whenever I see a post starting with "trust me on this, I'm a ..." I am reminded of the responses to Marilyn vos Savant's article on the Monty Hall problem - where literally (wait ... yes ... literally) hundreds of [...]s told her she was wrong (based mainly on their being so wonderfully qualified) Took a while to work out what you were trying to say here - but then I saw that you had made a mistake (highlighted) that meant it did make literally no sense. This is, of course, a fairly drastic simplification of the change that is possible. Indeed, given the whole ranking system relies on differences rather than levels, any linear translation of old ranking to new rankings via a functional form of R' = k + 1.5 x R shares these properties. But, of course, I'm not a mathematician. J
You right, there is no inherit reason for you to trust me. Eventually we seem to agree that 600/60 is a linear translation of 400/40, and given initial values that went through the same linear translation, the rankings are the same. That's what I meant earlier. I do work as an engineer, but I like football, and ranking systems in general. I'm allowed to have hobbies, right? BTW, I meant above 1.5*5, which is 7.5.
I would expect the July Rankings to use the new formula. Wouldn't it be strange to announce the new formula before the World Cup and then have World Cup games use the old formula? The host is seeded, so why would it matter to Russia when the formula changes? The draw was done last year, so when the formula changes doesn't matter unless one of the confederations is about to do seeding for something. World Cup 2006 ended on July 9. Rankings were released on July 12 using the old formula. I don't know how many World Cup games started counting then and how many started counting in August. Sometimes FIFA skips a month and does not making rankings during a World Cup. In August 2006, the first month after the last formula change, Brazil remained Number 1, and their points almost doubled from 827 to 1,630. In the Women's FIFA Rankings, every team has at least 358 points. I think FIFA will do something so that teams never have negative ratings.
The next ranking is scheduled for 19 July, after the end of Russia 2018, and fifa.com's ranking tool still allows you to calculate your July ranking (minus results from KO stages of course). That doesn't have to mean anything, it probably doesn't, but neither does the date when it was decided that the ranking will be overhauled. FIFA could have stipulated from which day onwards the new formula will be used. I'd like more clarity but for me it's also fine to assume the new method will already be in use. However that means the new method will be used retroactively, as games were played before the decision was taken, on 10 June, and that weren't included in the June ranking (from 4 June onwards). There's a big difference between the old and the new method. In case the old method were used you'd see lots of movement in the ranking as soon as the World Cup was decided. When the new method is used your ranking will move more like a continental shelf (these are World Cup matches, not low-yield friendlies, and they wouldn't have a massive impact on your ranking any longer): http://www.football-rankings.info/2018/06/fifa-ranking-july-2018-probable-ranking_11.html The Qatar 2022 Preliminary Draw is next year in July. Not much time to improve your ranking and many teams can only play low-yield games. Russia picked an awful time to host (their ranking plummeted to an all-time low after playing friendlies), France as well, to a lesser extent, having had the honour of hosting the EUROs. OTOH the likes of Wales, Romania, Poland, Switzerland, ... gamed the rankings and now stand to profit again. FIFA's overhaul of their rankings came about because of reports that teams were gaming the rankings and now FIFA will reward those teams (their inflated points total doesn't reflect their comparative strength and due to the new method it will take a lot of time before it erodes). At least FIFA could allow the likes of Russia to recover from their all-time low ranking and start with the new method once World Cup results are well-represented (which they aren't in the June ranking).
Why does it matter if games from June 4 to June 9 are going to count using the old method once or not at all? I'm assuming that regardless of what method the July rankings use, that they won't be used for a qualifying draw by any confederation. I disagree with http://www.football-rankings.info/ about the weighting for friendlies. Even if FIFA doesn't call it a "window," I'm assuming that all clubs must have released their players for those games. Obviously none of the top clubs play this month. I thought the purpose of weighting friendlies 5 or 10 is to weight them 5 when clubs do not have to release players, but I could be wrong.
These calculations seem to be based on K values of 5 to 60. But on the FIFA website, the document that is now linked, uses K values of 0.5 to 6! The internal PDF date is now June 14, 2018 instead of June 10, 2018. See https://resources.fifa.com/image/up...orld-ranking.pdf?cloudid=fzltr4s8tz3v3vy0aqo1 referenced in the June 10 FIFA announcement at https://www.fifa.com/about-fifa/new...nates-bids-for-final-voting-by-the-fifa-.html I'd assume the newer is wrong, or else it would take 30-40 years to stabilize, rather than 3-4 years! (Or perhaps that's in someone's benefit! )