User:Joriki/FIDE discrepancies
These are emails I wrote to FIDE regarding discrepancies on their rating site, in particular between the two types of top 100 lists published by FIDE. I'm posting them here so that I can refer to them in discussions about FIDE ratings and rankings here on Wikipedia. Joriki (talk) 18:42, 24 July 2021 (UTC)
I'm writing to make you aware of some discrepancies between the top 100 lists published at https://ratings.fide.com/toplist.phtml (hereinafter "type 1") and those published at https://ratings.fide.com/top_lists.phtml (and equivalently at https://ratings.fide.com/rankings.phtml) (hereinafter "type 2").
As there are quite a few details in the following, which you may not have the resources to process immediately, let me highlight the most important discrepancy: Vladislav Kovalev is not listed in the type 2 lists, whereas he is (correctly, it seems) listed in various type 1 lists between July 2018 and March 2021.
1)
The two types of lists resolve rating ties differently. Both types break ties between players with the same rating by using the number of games played in the rating period. But whereas the type 1 lists order players with the same rating and the same game count alphabetically, I could not identify any consistent pattern in how the type 2 lists resolve these ties. This difference results in very many discrepancies in the rankings. (To give just one example, for the current rating period of July 2021, the type 1 list (at https://ratings.fide.com/toparc.phtml?cod=649) lists Michael Adams as No. 26 and Andrey Esipenko as No. 27, and the type 2 list (at https://ratings.fide.com/rankings.phtml?rating=standard&period=2021-07-01) has them the other way around.)
2)
In the following cases, the lists contain different ratings, with the ratings in the type 2 lists matching those listed in the individual players' profiles. These errors fall into two categories:
a) different game counts
In several cases, the game counts in the type 1 lists differ from those in the players' profiles, while the type 2 lists are consistent with the rating calculations in the players' profiles:
March 2015: Onischuk, Alexander (2669 vs. 2665). The type 1 list has a game count of 2, whereas the game count in the player's profile for this rating period is 0. The lists for April 2015 agree on a rating of 2665, with a matching game count of 0, so the type 1 list seems to be in error.
March 2017: Kasimdzhanov, Rustam (2699 vs. 2696) The type 1 list has a game count of 0, whereas the player's profile for this rating period shows 1 game, along with the corresponding rating calculation, which is consistent with the type 2 list.
September 2017: Vallejo Pons, Francisco (2717 vs. 2716) The type 1 list has a game count of 0, whereas the player's profile for this rating period shows 7 games, along with the corresponding rating calculations, which are consistent with the type 2 list.
b) rounding errors
April 2015: Navara, David (2744 vs. 2743) Wojtaszek, Radoslaw (2738 vs. 2739) Ponomariov, Ruslan (2714 vs. 2713) These seem to be rounding errors in the type 1 list. The game counts are correct, and the ratings in the type 2 lists are consistent with the rating calculations in the players' profiles.
October 2016: Iturrizaga Bonelli, Eduardo (2658 vs. 2659) This rating and the corresponding game count aren't in the player's profile page, so this could be either type of error.
3)
The following players are listed in the type 1 lists but not in the type 2 lists:
Kovalev, Vladislav He is listed (correctly, it seems) in various type 1 lists between July 2018 and March 2021. (See also https://ratings.fide.com/top_files.phtml?id=13504398.) No reason for his absence in the type 2 lists is apparent.
Bukavshin, Ivan (died January 2016) He is listed in the type 1 lists from August 2015 to July 2016. It's not clear why he's missing in the type 2 lists, or why he disappeared from the type 1 lists in August 2016. His last rated games were played in January 2016, so the regular delisting due to a year of inactivity would have occurred in January 2017. The regulations don't seem to provide for an earlier delisting in case of death. Bukavshin no longer seems to have a profile page, but other FIDE pages still refer to him: https://ratings.fide.com/top_files.phtml?id=4199758, http://ratings.fide.com/title_applications.phtml?details=1&id=4199758&title=GM&pb=31, https://ratings.fide.com/chess_statistics.phtml?event=4199758.
Rausis, Igors (banned December 2019) He is listed in the type 1 lists from August 2018 to February 2020 (except October 2018). It's not clear why he's missing in the type 2 lists, or why he disappeared from the type 1 lists in March 2020. His last rated games were played in August 2019, so the regular delisting due to a year of inactivity would have occurred in August 2020. The decision sanctioning Rausis for cheating (https://www.fide.com/news/246) expressly states that it does not affect his ratings, and the rating lists available at http://ratings.fide.com/download.phtml still list him. Nevertheless, his profile page (https://ratings.fide.com/profile/11600098) no longer contains his ratings and lists him as "Not rated".
4)
Some more details:
a)
The type 1 lists use inconsistent first names for some players. For instance, the type 2 lists always use "Ivanchuk, Vasyl", whereas the type 1 lists use "Ivanchuk, Vassily" up to April 2020 and "Ivanchuk, Vasyl" from then on. Other affected players are:
- Shankland, Samuel L / Shankland, Samuel / Shankland, Sam
- Harikrishna, P. / Harikrishna, Pentala
- Vovk, Andrey / Vovk, Andriy
- Bologan, Viktor / Bologan, Victor
b)
The type 1 list for September 2009 (https://ratings.fide.com/toparc.phtml?cod=149) lists Svidler, Peter before Ponomariov, Ruslan. They played the same number of games, so by the convention that the type 1 lists otherwise follow consistently, Ponomariov should be listed alphabetically before Svidler.
c)
At https://ratings.fide.com/rankings.phtml, in the selection list for the rating period, the selection item for August 2018 is missing the space between month and year.
d)
The parameters "list=men", "list=men_rapid", "list=men_blitz" (e.g. in the URLs linked to at https://ratings.fide.com/toplist.phtml) are misleading, as they actually refer to the open lists that include all players (currently the female player Hou Yifan).
That concludes the compilation. As I use these lists a lot to maintain data on Wikipedia, I would much appreciate if you could let me know which of the listings you consider authoritative in each case.
P.S.: I just noticed that on https://ratings.fide.com/top_lists.phtml and https://ratings.fide.com themselves, if you don't select a rating period from the dropdown menu, you actually get the current type 1 list (the one at https://ratings.fide.com/top.phtml?list=men) in the style of the type 2 lists, not the type 2 list that you get by selecting "RATINGS: LATEST" in the dropdown menu. That is, if you go to https://ratings.fide.com and click on "Top", you get one list, and if you then select "CHECK MORE RANKING LISTS ..." and then "GET RANK LIST" (leaving "RATINGS: LATEST" selected), you get a different list in the same style; e.g. currently Michael Adams and Andrey Esipenko switch places. So this is not just a problem of different lists appearing in different parts of the website; the inconsistency is directly visible on the front page of the ratings site.
P.P.S.: Further to my previous emails, I just noticed that a further inconsistency is introduced by the "World Rank (Active)" entry on the players' profile page. For example, currently on Illya Nyzhnyk's profile (https://ratings.fide.com/profile/14118084) this entry says "51", but Nyzhnyk's rank both on the type 1 list (https://ratings.fide.com/toparc.phtml?cod=649) and on the type 2 list (https://ratings.fide.com/rankings.phtml?continent=0&country=&rating=standard&gender=&age1=0&age2=0&period=2021-07-01&period2=1) is 50. He has the same rating as three other players but a higher game count, so the rank in the lists seems to be consistent with the usual tie breaking system and the rank in the profile appears to be in error.
I am again writing to point out errors and inconsistencies on your site.
At https://ratings.fide.com/top_federations.phtml, there is a list of federations ordered by the average rating of their top ten players. These averages seem to be correct. However, the list is only available up to June 2021, the data for July and August 2021 is missing; and it is only available for all players, not for women separately.
There is a different list at https://ratings.fide.com/topfed.phtml. It doesn't specify a date, but the individual player ratings displayed when a federation is selected seem to be the current ones for August 2021. This list also allows women to be selected separately. (Note that the label "Men" for the radio button at the top is incorrect; in fact all players are included in the list when this is selected.)
However, this list seems to be broken. The averages are not the correct averages of the top ten players that are displayed when the federation is selected. For instance, for August 2021, the average rating of the top 10 active Russian players (https://ratings.fide.com/topfed.phtml?ina=1&country=RUS) is 2732,9, and the average including inactive players (https://ratings.fide.com/topfed.phtml?ina=2&country=RUS) is 2748,4, whereas the list displays 2739 for Russia, independent of whether "Active only" is selected or not. Even if the average were meant to be for a different month, it should make a significant difference whether inactive players are included.
- I got a response from FIDE to this one. They write that https://ratings.fide.com/topfed.phtml and https://ratings.fide.com/topstat.phtml (which I'd written about to them earlier) are no longer supported, and they've now forwarded them to the more modern-style pages https://ratings.fide.com/top_federations.phtml and https://ratings.fide.com/rankings.phtml, respectively. They also added the missing July 2021 and August 2021 options at https://ratings.fide.com/top_federations.phtml. Joriki (talk) 04:03, 3 August 2021 (UTC)
I am writing to point out some ambiguities in the announcement of qualification paths for the Candidates Tournament 2022 (https://fide.com/news/1097).
For qualifying event C, both contenders in the 2021 match are mentioned, and their places are passed on to the next non-qualifying player.
By contrast, for qualifying event D, only "players [who] happen to be already qualified through events A, B or C" are mentioned.
There are two problems with this:
- The 2021 match is scheduled to take place after the Grand Swiss Tournament, so no player will have qualified through event A when event D finishes.
- Even if this rule were retroactively applied to the runner-up of the 2021 match once he is known, it would still not apply to the winner of the 2021 match, who is not a "player qualified through event A". Thus, it is not clear what happens if one of the two top finishers in the Grand Swiss wins the 2021 match.
Neither of these problems would arise if the provision for event D instead mentioned "the two contenders in the 2021 match", like the provision for event C does.
Another ambiguity arises due to the possibility of players giving up their qualification spot in order to take part in the Grand Prix: There is no provision for filling such a spot.