FIDE rating magic

02.01.2015 /   Andrejic, Vladica (2268)

My little analysis of the FIDE rating and another proof of FIDE incompetence. Happy New Year!

Few years ago FIDE invented Rapid and Blitz rating. After my first considerations I wrote Rogulj's Blitzkrieg: A Long-lasting April Fool's Joke or a New Rating FIDE-ology? FIDE didn't like Rogulj's 2920 on the first place of the April 2013 Blitz rating list, so they retroactively change their calculations, but we notice some inaccuracy in texts Correct Rating is a Rare Thing and FIDE Rating Bug Still Biting.

Recently, I have noticed some details in the first announcement (16.11.2011) for Rapid and Blitz rating lists: "If any player plays 35 games or more in a period, his new rating will be based solely on those games." This is very awkward, but it is clear that calculation mistakes for players with 35+ played games are not the programming bug. Simply, FIDE calculate Elo in their own way, probably stupid enough that they should not display it publicly, but this does not prevent them to apply it in practice. Anyway, I have never seen the FIDE rating regulations for Rapid and Blitz, if they exist at all.

Consider now the standard FIDE rating and new FIDE magic from July 1st 2014. The development coefficient K is kept at K=10 for players who reached 2400. However, mere mortals like me get K=20, instead of the earlier K=15, which means that the amplitude increased by one third. New players, until their first 30 games, have K=40, instead of the older K=30, or antic K=25.

The novelty here is K=40 for all players until their 18th birthday, as long as their rating remains under 2300. FIDE do not respect this, beacuse they keep K=40 after 18th birthday till the new year, so for FIDE, birthday means the calendar year. You might notice huge Elo profits for good play of juniors in recent months. However, with such K=40, we do not need a large number of games that whould occur so-called rating lag. Thus, I believe that FIDE have to find some solutions like my former suggestion How to prevent rating lag?

Why? Because it is ridiculous that someone can get much higher rating than the performance that has. It is nice for a junior to catch up his real Elo strenght, but FIDE regulations allowed nice beautiful hypothetical exhibitions. Or not so hypothetical? Look at the concrete example of youth Parviz Gasimov (born in 2000) from Azerbaijan. On the first day of October he had 1949, but today, three months later, he has 2517! Riplays Believe It or Not, makes 568 Elo points in three months! Let us analyse his results from the last three rating lists.

If we look carefully into these results we can see that they are not for grandmaster 2500+ level, but they are contributed to this at the very end. I shall use the real rating performace (detailed description one can see in texts The Truth about Chess Rating and Chess Rating for Laymen) instead of poor FIDE rating performance. The real rating performance (AP) is the rating that would be preserved during the process of rating of the games in question (game by game) and it can easily be calculated iteratively.

Thus, 56 point out of 77 games rated by FIDE in the last three lists worth AP=2277, which is 240 Elo points less than his current rating. If we ignore warming-up from the Novemeber list, we have 40.5 out of 50, which worth AP=2363, which is higher, but also drastically below the current 2517. Finally, if we calculate just the last January 2015 list, we get that his 23.5 out of 32 worth AP=2326, which is even weaker than the previous calculation. Please notice, that one of the rated tournament (from January 2015) is Abu Dhabi Chess Festival 2014 Open played on August 2014! There he had 1964 Elo and get 6 out of 8 against under 2000 oppositions, and now, 4 months later, he earns 41 Elo point for that poor result, clean way to accomplish masterfully condition of 2500 Elo.

Shame on you FIDE!