- 12 May '08 22:42

He's really kicking butts this time! And not any butts they are all Super GM butts!*Originally posted by Korch***http://www.chessbase.com/newsdetail.asp?newsid=4626**

Ivanchuk is real monster - 5 wins in first 5 rounds, over Radjabov, Topalov, Bu Xiangzhi, Cheparinov and Aronian!!!

He really loves chess and has great fun while playing it and that's why he's one of the guys I like to see winning. - 12 May '08 23:43

We'll see if he can avoid self-destructing. He's been known to play brilliantly and then lose a couple of horrible games, usually in time pressure. Topalov is still in striking distance, but it won't be easy.*Originally posted by Korch***http://www.chessbase.com/newsdetail.asp?newsid=4626**

Ivanchuk is real monster - 5 wins in first 5 rounds, over Radjabov, Topalov, Bu Xiangzhi, Cheparinov and Aronian!!! - 13 May '08 23:50 / 2 edits

That's the number that Chessbase is reporting (2740 rating plus a performance of +797). I don't know how the performance is calculated. I guess it's possible for Chessbase to be wrong, but I can't prove or disprove the number.*Originally posted by gaychessplayer***The only way to have a performance rating of 3537 is to be undefeated against players rated 3137. Since no one in the world is rated over 2825 or so, such a PR would be impossible.**

Edit - OK, I just checked the elo performance rating formula on Wikipedia. If I understand the formula correctly, you get +397. So my guess is that Chessbase did indeed make a big typo, only off by 400 points.

Edit 2 - Of course, I'm assuming that FIDE doesn't use some bizarre variation of the standard elo performance rating formula. And considering that it's FIDE, I might be making a bad assumption. - 14 May '08 07:32

I think you gain (and lose) rating points at 1/2 the rate once above a certain rating (I think it is around 2400 but it might be a little lower).*Originally posted by Mad Rook***That's the number that Chessbase is reporting (2740 rating plus a performance of +797). I don't know how the performance is calculated. I guess it's possible for Chessbase to be wrong, but I can't prove or disprove the number.**

Edit - OK, I just checked the elo performance rating formula on Wikipedia. If I understand the formula correctly, you get +397. S ...[text shortened]... mance rating formula. And considering that it's FIDE, I might be making a bad assumption.

but this is still an amazing performance rating in excess of 3000. Even GMs rarely win against players +2700 and 5/5 is trully awesome. - 14 May '08 17:52The non draw rule applies, so they can't offer a draw unless it's a theory draw or umpire suggests so.

And keeping that in mind it still is a great achievment. If you know you can't just equalize and head for easy draw maybe some less played and "unsound" lines will appear if they feel the opponents preparation is bad.

Not sure if that would work at their level but still it is fodder for fun games I think!

Gogo chucky ;p - 15 May '08 01:01 / 2 editsOK, I need to revisit the rating performance calculation issue for Ivanchuk. I think we were all off base with how it's determined. After Round 5, Chessbase reported that Ivanchuk's performance was +797 points. And after Round 6, Chessbase reported that with Ivanchuk's Round 6 draw, his performance was now "only" +415 points. So clearly, either Chessbase is wrong, or I'm wrong in my understanding of the calculations. After looking at the FIDE handbook, I think Chessbase was right, and I was wrong.

Apparently the way it should be calculated is as follows:

1) Calculate the average elo of Ivanchuk's opponents for the rounds played. (If this is being done after Round 6, then you have to add Radjabov's rating twice, since Ivanchuk played him twice. After summing the 6 rating numbers, you then divide by 6.)

2) Using the table in the FIDE handbook, section B.02.10.1.a, calculate the difference points (dp) according to Ivanchuk's percentage score (p=Points obtained/Number of games).

Apparently, this table is just an approximation of the exact equation:

dp = -400 x log(1/p - 1)

(in the above equation, the x means multiply)

3) Then Ivanchuk's performance is obtained by adding the results of 1) and 2) above. That is, Performance = Average ELO of Rivals + dp

I used this method to calculate the Round 6 performance for Ivanchuk and Topalov (I used the exact equation instead of the chart), and I got within 2 points of the value Chessbase got for Ivanchuk and within 1 point of the value Chessbase got for Topalov.

For example, for Topalov after Round 6, the average elo of his opponents was 2736.8333. Topalov's percentage score was 4.5/6, or 0.75. Plugging these numbers into the dp equation, we get dp = 190.85. Therefore, Topalov's performance was 2736.8333 + 190.85 = 2927.68, which is about 160.7 points above his rating of 2767. Chessbase gave a value of +160, which is about the same as I got, considering roundoff error.

The curious thing you can see from the table is that at a percentage score of about 0.92, the value for the difference points is 400 points. And as the percentage score goes above 0.92, the difference points continue going up. (At a p of 0.99, the dp is 677.) And for Ivanchuk after Round 5, his p was 1.0, which is undefined for the equation, and it basically means that the difference points have an infinite value. (I suspect this is the reason that the Chessbase report for Round 5 carried the title, "MTel R5: Ivanchuk aiming for infinity with 5.0/5".)

The only thing I don't understand is how Chessbase came up with the +797 performance number for Ivanchuk after Round 5. (Technically, his performance was infinite.) It appears that they arbitrarily assigned a value of 800 for the dp, correlating to a p of 1/1.01, or 0.990099. If you use these arbitrary numbers, you get a +797 performance as Chessbase reported.

Dang, I hope I don't get any questions on this - my head hurts. - 15 May '08 06:19

I got the bit about the x meaning multiply.*Originally posted by Mad Rook***OK, I need to revisit the rating performance calculation issue for Ivanchuk. I think we were all off base with how it's determined. After Round 5, Chessbase reported that Ivanchuk's performance was +797 points. And after Round 6, Chessbase reported that with Ivanchuk's Round 6 draw, his performance was now "only" +415 points. So clearly, either Chessbase is ...[text shortened]... reported.**

Dang, I hope I don't get any questions on this - my head hurts. - 15 May '08 09:35

If they do it like the guy on chessetrics they invent some draws against lower rated opponents and see what it gets. The thing is that with a few draws you always get a finite answer for the performance rating and if done correctly they don't affect too much the estimation of the*Originally posted by Mad Rook***The only thing I don't understand is how Chessbase came up with the +797 performance number for Ivanchuk after Round 5. (Technically, his performance was infinite.) It appears that they arbitrarily assigned a value of 800 for the dp, correlating to a p of 1/1.01, or 0.990099. If you use these arbitrary numbers, you get a +797 performance as Chessbase reported.**

Dang, I hope I don't get any questions on this - my head hurts.*real*result.