< Earlier Kibitzing · PAGE 30 OF 31 ·
|Dec-31-11|| ||alexmagnus: Heh, the so liked by the inflationists average rating of the top-100 has actually fallen in the last rating period by 1 point. Interestingly, btw, the 2700 club has no true new member in this list - that is, each 2700+ on the January 2012 list has at least once been at 2700 or higher in the past.|
|Jan-01-12|| ||alexmagnus: Carlsen's domination level:
No new peak in terms of my "first domination definition" (distance to #10): 74 points, the peak 78 is in July 2010.
New peak in terms of my "second domination definition" (expected score in a DRR against five highest ranked players who are not himself): 5.64, previous peak was 5.56 in July 2010.
|Jan-08-12|| ||alexmagnus: From my recent blitz game:
click for larger view
Black to move :D
|Jan-18-12|| ||alexmagnus: By the way, to those interested in the issue of rating inflation (or its absence): have you seen this? http://www.cse.buffalo.edu/~regan/p...|
|Feb-18-12|| ||alexmagnus: Average Elo rating of world chess championship matches, just for the historical reference:|
Karpov-Korchnoi I 2695
Karpov-Korchnoi II 2697.5
Karpov-Kasparov I 2705
Karpov-Kasparov II 2710
Karpov-Kasparov III 2722.5
Karpov-Kasparov IV 2720
Karpov-Kasparov V 2765
FIDE during the split:
|Feb-19-12|| ||alexmagnus: As I said, there are two kinds of domination: general domination (which I measure in two ways - gap to #10 and expected score against 5 best players who are not the measured player) and domination over the toughest rival (rating-wise, not in terms of head-to-head). I listed largest gaps to #10, here is how the six #1 players made it in terms of gaps to #2:|
Fischer 125 (1972)
Kasparov 82 (January 2000)
Karpov 65 (January 1982)
Topalov 34 (July 2006, October 2006)
Carlsen 30 (January 2012)
Anand 23 (July 2007)
|Feb-25-12|| ||frogbert: funny position in your blitz game, alexmagnus!
from the diagram 1... Nxe5!? looks fun, with 2. Bxe8 Ba6+ to follow. then there are two moves to choose from
a) 3. Bb5 Bxb5+ 4. axb5 Rxa2+ 5. Ke1/f1 Rb8 (i think), which looks rather devastating for black. or
b) 3. Ke1 Nd3+ and then
b1) 4. Qxd3 Bxd3 and white can't cover both his knight on b1 and bishop on e8, so black wins
b2) 4. Ke2 Nxf2+ 5. Kf3 (best?) Nxd1 6. Rxd1 Rxe8 and black's winning.
i might have overlooked something, as i only looked at the diagram for the above analysis, but it indeed looks like a nice pseudo queen-sacrifice.
how did the game conclude?
|Feb-25-12|| ||frogbert: in line a), maybe the greedy 5... Nxf2 (with a fork on d1 and h1) is as good or better than 5... Rb8, but both look winning.|
|Feb-25-12|| ||alexmagnus: <how did the game conclude?>|
The game ended 19... Nxe5 20. Bxe8 Ba6 21. Ke1 Nd3 22. Qxd3 Bxd3 23. Bc6 Rac8 24. Bd7 Rc7 25. Nd2 Rxd7 26. Nf3 Rb7 27. Ra1 Rb2 28. Ne5 Ba6 29. a5 Re2 30. Kd1 Nxf2 0-1
Patzerish from both players, but I was quite proud to have found Nxe5 in a blitz game :)
|Feb-25-12|| ||frogbert: nah, you converted easily.|
|Feb-27-12|| ||alexmagnus: Some moves were unnecessary though. Like Rac8 or Ba6. I know it's blitz and I'm far from an engine (or even from you :D) when it comes to playing level, but that doesn' hold me away from a "perfectionist" look at own games (resulting in being not completely satisified with almost each game, but that's a way to improve. Though now I have doubts around it - I'm stuck at 1500-1600 level for years).|
|Feb-27-12|| ||frogbert: well, i had a period where my rating was stuck at 1600-something (norwegian rating, roughly 1800 fide - but at the time there didn't exist fide ratings on that level). from january 1996 to january 2001 i played 20-25 games per year on average, but only increased my rating by a net 100 points, from 1603 to 1709, after going up and down a lot, touching 1700 in -98, dropping back to 1630-something in -99, etc.|
i finished my master in 2000 and started working, and in the first norwegian championship i played after that, i jumped 110 points up and have been 1800+ (in norway) ever since, with a max of about 1950. (my current fide rating of 2046 is also the highest it's been, and i turn 40 this year.)
maybe you'll get a boost now that you're done with your master? i don't think i did anything special to make my jump into the 3rd highest (of 7) rating classes in norway - it appears that suddenly i simply had understood a little more. however, i'm pretty sure i could make another jump of ca. 100 points now, only by working systematically on issues i'm aware of, that has little or nothing to do with my chess understanding, only with certain bad habits in otb play that i *know* cost me points on a regular basis.
|Feb-29-12|| ||alexmagnus: When I finally find a job I'll look for a club.
And now, regular update re. new rating list:
Distance #1-#2: 15 points (Carlsen's personal best: 30)
Distance #1-#10: 71 points (Carlsen's personal best: 78)
"Domination list": Carlsen and Aronian qualify, neither improves his personal best.
Average of top-100: 2700
"Lifetime" 2700 club: one new member (Grachev)
Carlsen's "domination by expected score": 5.56 (best: 5.64).
|Apr-30-12|| ||alexmagnus: The new rating list didn't change much:
Distance #1-#2: 10
Distance #1-#10: 71
"Domination list": Aronian reaches, but doesn't beat his personal best.
|May-01-12|| ||alexmagnus: Average rating of the upcoming WC match: 2759.|
|May-10-12|| ||solskytz: Hey Alexmagnus!
I love that article which you reference. Great that somebody already did the research and conclude what I felt was true - that rating inflation doesn't exist (and that actually some deflation might exist)
And to Frogbert -
well funny... I just turned 40. I also play the strongest I ever had, also feel that I can improve further whenever I get to work on my chess...
also in my case I attribute my improvement to suddenly better understanding, having gone through some life issues and challenges - but not to any particular work on chess!!
|Jun-05-12|| ||alexmagnus: As one user pointed out, my "domination list" was based on faulty pre-2000 data. I corrected them. Notable changes: |
1)Kramnik comes ahead of Ivanchuk (on the faulty list he was behind)
2)Anand enters the "pantheon" (score over 100, ob the faulty list it was 95)
3)Carlsen moves from 11th to 10th (as pre-1971 lists were unofficial and don't count, Spassky moves down)
4)total rehash at the bottom of the list.
Here the corrected "domination list":
Kasparov 175 Jan 90
Fischer 160 Jul 72
Karpov 130 Jan 89
Kramnik 110 Jan 98
Tal 105 Jan 80
Ivanchuk 105 Jul 91
Anand 105 Jul 98
Korchnoi 95 Jan 80
Topalov 84 Jul 06
Carlsen 78 Jul 10
Spassky 70 Jan 71
Shirov 65 Jul 94
Aronian 61 Mrz 11 Mai 12
Gelfand 60 Jan 91
Kamsky 60 Jan 96 Jul 96
Morozevich 57 Jul 99
Portisch 55 Jan 80
Jussupow 55 Jul 86
Timman 55 Jan 90
Adams 52 Okt 00
Bareev 50 Jul 91
|Jun-22-12|| ||Benzol: <alex> Some info for you at the Biographer Bistro|
|Jun-22-12|| ||alexmagnus: Thx <Benzol>.|
|Jul-01-12|| ||alexmagnus: July ratings: Carlsen the only player qualifying for the domination list (10 points below his personal best).|
As for my second domination definition, Carlsen's score is 5.62, slightly below his PB of 5.64.
|Jul-17-12|| ||whiteshark: HMK alert 4u, she's playin bd 5: http://liveschach-schau.de/turniere...|
|Jul-17-12|| ||alexmagnus: I knew of that Germany-Poland match, but thx :)|
|Aug-08-12|| ||achieve: Hi <alexmagnus> - I just now read these posts of yours (see below), and am interested in the calculations behind the numbers. I am not knowledgable wrt this subject, although I recently did some research into Jeff Sonas' "simple formula", realizing that the ELO calculation is different. Sonas simple method, if I understood correctly, is that if you score 10% over 50% against an average field/strength of X, then your Performance Rating is determined to be X + 85 ==> If one scores 6/10 against 2400 opposition, one's playing strength is determined to be 2485.|
This system awards therefore 8.5 points for each percentage point scored above "playing even", i.e., a 50% score. 76% would in that case indicate a difference in strength of 221 points, compared to the exact 200 points you give in your table (below).
Could you please explain what, if the case, I am presenting that is wrong? Enlighten me on the current "conventions/Formula(s)" used? (There are small differences in TPR calculations on different sites.)
I also read this article - http://www.chessbase.com/newsdetail... - and although I can not grasp every detail, it is helpful in addressing, and "learning" to understand, the different Formulas that can be used, are used, to calculate ratings and expected scores.
- - - - -
<alexmagnus: Or, in other words, to be expected to win oftener than to draw, the rating difference between an unbeatable player and his opponent must be at least 193.>
<alexmagnus: 100: 64%
|Aug-08-12|| ||alexmagnus: <achieve>
1) Sonas' method is actually somewhat more complicated than this - but exactly for the reason of that complication it appears a bit nonsensical to me. Wishing to make the performance dependent on the number of games, he adds to the actually played games four fake draws against players rated 2300, then applies to this sample of "true plus four" games the "85-point-formula" you mentioned, and then adding 43 points. Without any explanation why 43, some empirical values.
2)Sonas long advocated the "85-point-formula", but recently changed his mind. When he checked the actual distribution of results, he said the (real) Elo formula is correct (pointing out that it is wrong with the currently existing 400-point-rule - which is obviously true).
3)The percentage scores I mentioned were taken from the table on FIDE's own website.
4)Different results from different TPR calculators:
a)Some calculators use the table I mentioned, rounding the percentage score to the nearest integer.
b)Some don't round the percentage score and use FIDE's table, just extrapolating the score to non-integer values.
c)Some use the approximation:
Expected score=(number of games)/(1+10^(rating difference/400)).
This formula works the better the closer the rating difference is to 0. At huge rating differences, say, 500 points, it produces quite a big deviation.
5)Some use another approximation: normal distribution with the SD of 200 points. It is a better approximation than the formula above,
What the FIDE itself uses as the base for it's table - no idea.
|Aug-08-12|| ||alexmagnus: Also, FIDE's own TPR calculator used to have a bug (not sure if it is still there) - instead of calculating TPR it calculated the initial rating based on the score. Which is equal to the TPR only in case of the score being below or equal 50%.|
< Earlier Kibitzing · PAGE 30 OF 31 ·