tamar, not even carlsen can win drawn positions if the opponent defends perfectly. and against ivanchuk i have yet to see that there ever was a win for black.

also, carlsen usually is among those who work the hardest otb. it's the pre-game work he's been criticised for.

it can be argued that a truelly "professional" player would've managed to squeeze a full point from the games against kramnik and karjakin; personally i'm willing to "excuse" carlsen for not seeing Bc8 against kramnik (but as he wrote himself, he was probably a bit too relaxed after having reached the time control there - he should've tried to reset and really look at the position with fresh eyes after move 40). against karjakin though i think he probably should've been able to concentrate/focus enough before move 40 in order to see that Rxh4?! was flawed as a winning attempt.

37. hxg5! however is insanely complicated in several lines, and no human (not even kasparov in his prime) would be able to see through every possible line there. that's not to say that carlsen (or anybody else) couldn't have played 37. hxg5 on any given day - without being able to calculate it to its conclusion. of course it could've been played. but it's much more human to look for a clearer ending with a huge plus - the problem was that carlsen didn't take the time (or wasn't able to, if you like) to find the right one, the one after 39. Nxh4 (winning) instead of the one after 39. Rxh4 (drawing, against best defence).

still, i think that carlsen, without the first 3 draining games or with a rest day before the game against karjakin, would've had greater chances of finishing what he started yesterday. of course, it's not a great excuse - just an observation of the tough schedule in moscow (equal for everyone, of course) and the tough, complex games carlsen has played (not so equal, as there have been a few not so ambitious games from some players).

I do think Magnus could adopt some of Kasparov's suggestions with benefit to himself.

kasparov hovered around 2830, he never performed consistently around 2850 - that was (as usual) the result of an <untypical> string of good results, this time from kasparov.

as i argue for current ratings, i find it more interesting around which rating level a player appears consistent (regardless of what we may or may think about rating inflation). in that respect 2851 is only interesting in terms of "fide rating record" - it's not the most interesting rating number in terms of expressing kasparov's level or typical distance to his closest peers.

- Magnus Carlsen

in terms of systemic inflation, i think the difference is actually less than that - maybe at most 20 points in the last 10 years. (bronkenstein's wild guess is just that, a wild guess.)

however, even if we can assess the amount of systemic inflation, it can't really be used to "adjust" kasparov's 2850-ish rating from 1999/2000 to a 2011-equivalent. the ratings are relative, and when someone catches up with the rating leader, it will typically impact the ratings of the guys playing catch-up positively and the leader's rating negatively. moreover, a peak rating isn't - as i just argued - not the most interesting rating as a measure of level. (consider radjabov's, karjakin's and nakamura's recent rating peaks as reference.)

example: let's go back to 1998 for a brief moment and consider the rating lists from january that year until july 1999:

january 1998:
1. kasparov 2825 (+5)
2. kramnik 2790 (+20)
3. anand 2770 (+5)
4. topalov 2740 (-5)

july 1998:
1. kasparov 2815 (-10)
2. anand 2795 (+25)
3. kramnik 2780 (-10)

first observation: both kramnik and anand were hovering very close to 2800 and the gap up to kasparov was only 20-30 points. now, does anyone think that kasparov suddenly became a whole lot better sometime in 1999? at the age of 36? i for one do not. what happened rather, was that he had a string of unusually strong results (even for him), without any intervening bad events.

january 1999:
1. kasparov 2812 (-3)
2. anand 2784 (-12)
3. kramnik 2751 (-29)

ok, kramnik has had a rough year, down about 40 points since january 1998. has he become a much weaker player? no, he's had a bad patch.

july 1999:
1. kasparov 2851 (+39)
2. anand 2771 (-13)
3. kramnik 2760 (+10)

kasparov has an amazing 6 months, while anand's dropping back some more points. kramnik recovers slightly. now, do we think that kasparov suddenly is at a level 80-90 points above his closest peers? or did he rather have 6 wonderful months?

for the rest of 1999 kasparov was busy playing against "the world" and some rapid and simul events. no more classical games that year after he reached 2851.

then he had another great 6 months playing corus 2000, linares 2000 and sarajevo 2000, performing on average at ca. 2850, only dropping 2 rating points to 2849. hence, he had a peak period consisting of 3 events in 1999 and the same 3 events in 2000. incidentally his closest peers had a somewhat negative form swing at the same time that kasparov had his positive form swing (january 1999 to june 2000). but anand and kramnik did participate in several of "kasparov's" events in 1999 and 2000. what were the results?

1999:
kasparov - anand 1/2 (hoogoven/waz)
kramnik - kasparov 1/2 (hoogoven/waz)
anand - kasparov 0-1 (linares)
kasparov - anand 1/2 (linares)
kramnik - kasparov 1/2 (linares)
kasparov - kramnik 1/2 (linares)

in sarajevo, neither kramnik nor anand participated

2000:
kasparov - anand 1/2 (corus/waz)
kramnik - kasparov 1/2 (corus/waz)
anand - kasparov 0-1 (linares)
kasparov - anand 1/2 (linares)
kasparov - kramnik 1/2 (linares)
kramnik - kasparov 1/2 (linares)

in sarajevo, neither kramnik nor anand participated

the results look a bit like blueprints of each other, but that's how it was. kasparov managed in these best 12 months of his career (his own judgement) to score 2 wins, 10 draws:

a) 4/6 against anand (2 extra whites)
b) 3/6 against kramnik (2 extra blacks)

wins 1999:
reinderman (2542) - kasparov 0-1
kasparov - kasimdzhanov (2595) 1-0
yermolinsky (2597) - kasparov 0-1
kasparov - piket (2609) 1-0
kasparov - sokolov (2624) 1-0
kasparov - van wely (2636) 1-0
timman (2670) - kasparov 0-1

kasparov - short (2697) 1-0
kasparov - topalov (2700) 1-0 (1)
topalov (2700) - kasparov 0-1 (2)
svidler (2713) - kasparov 0-1
kasparov - svidler (2713) 1-0
ivanchuk (2714) - kasparov 0-1
kasparov - ivanchuk (2714) 1-0
adams (2716) - kasparov 0-1
kasparov - adams (2716) 1-0
kasparov - adams (2716) 1-0
kasparov - morozevich (2723) 1-0

shirov (2726) - kasparov 0-1 (3)
anand (2781) - kasparov 0-1 (4)

losses 1999:
sokolov (2610) - kasparov 1-0

wins 2000:
kasparov - bacrot (2594) 1-0
kasparov - sokolov (2637) 1-0
kasparov - van wely (2646) 1-0
kasparov - timman (2655) 1-0
kasparov - polgar (2658) 1-0
kortschnoj (2659) - kasparov 0-1
nikolic (2659) - kasparov 0-1
movsesian (2668) - kasparov 0-1
kasparov - georgiev (2675) 1-0

kasparov - gurevich (2694) 1-0
kasparov - bareev (2709) 1-0
kasparov - morozevich (2748) 1-0

kasparov - shirov (2751) 1-0 (5)
anand (2769) - kasparov 0-1 (6)

selected draws 1999 - june 2000:

1) leko (2694) - kasparov
2) kasparov - leko (2694)
3) kasparov - topalov (2700)
4) topalov (2700) - kasparov
5) topalov (2702) - kasparov
6) kasparov - leko (2725)
7) kasparov - leko (2725)
8) leko - kasparov (2725)
9) shirov (2726) - kasparov
10) shirov (2751) - kasparov
11) shirov (2751) - kasparov
12) kramnik (2751) - kasparov
13) kasparov - kramnik (2751)
14) kramnik (2751) - kasparov
15) kasparov - kramnik (2758)
16) kramnik (2758) - kasparov
17) kramnik (2758) - kasparov
18) kasparov - anand (2769)
19) kasparov - anand (2769)
20) kasparov - anand (2784)
21) kasparov - anand (2784)

the 6 enumerated wins from the previous post can be contrasted with these 21 draws against the same players (except leko and kramnik, who didn't lose a single game out of 11 against kasparov).

so even including the kasparov-client shirov in the above (+2 -0 =3 here) and anand who didn't score very well against kasparov (+2 -0 =4 adv kasparov here) these 27 results against leko, topalov, shirov, kramnik and anand <do not represent rating gain for kasparov>.

it's quite clear that the reason kasparov outrated anand and kramnik by so much in this shorter period was <his score against the "non-contenders" (svidler, adams & co) plus lower-rated players>. those who have been playing for the crown after kasparov was dethroned (kramnik, leko, topalov, anand) did not directly "cause" his jump to and stay at 2850-level in 1999-2000.

interestingly kasparov only managed a single win against leko in 15 tries. he never lost, though ...

<the ratings are relative, and when someone catches up with the rating leader, it will typically impact the ratings of the guys playing catch-up positively and the leader's rating negatively.>

so, let's look at the july 2000 standings for a moment:

1. kasparov 2849
2. kramnik 2770
3. anand 2762
7. leko 2743
9. topalov 2707

now, let's imagine - hypothetically speaking - that

1) the players listed below kasparov above would reduce the gap between themselves and kasparov in terms of skills (and hence, results and ratings), and

2) all of these players also play a non-significant amount of games <against kasparov>.

given 1) above, all the players' ratings should get closer to each other than they used to be, and given 2) above, it follows logically that <kasparov's rating would go down even if his skills would be unchanged> - simply because his closest peers have improved, individually and as a group.

a consequence of 2) is that kasparov will <perform relatively worse> against these players, which again will have an impact on the <relative measure that ratings is>. in other words, a clear rating leader will lose rating points if he keeps playing against an <improving> group of lower-rated players, <despite keeping his absolute level>. this is a fact, but it seems that few are able to wrap their mind around it ...

[now, in reality i <think> that kasparov faded a little bit towards the end of his active career, mostly due to lack of motivation, but he didn't fall from some 2850-level either, because he was averaging some 2825-30 between 1998 and 2005, not anything 2850-ish.]

anyway, disregarding our hypothetical scenario from above, here are the actual numbers from april 2005:

1. kasparov 2812
2. anand 2785
3. topalov 2778
4. leko 2763
5. kramnik 2753 (*)

as in july 1998 the difference was again only in the 20-40 points range for his closest peers - and with kramnik fit and healthy there would probably have been another player in the 2780-90 area at that time, too.

*) kramnik's rating and results at that time were poorer than they should've been based on his skills - due to his health problems that weren't publicly known back in 2005. he'd been above 2800 between april 2001 and april 2003.

click for larger view

61 ... ♗c3x♘d4? 62 ♖c2-d2! <pin: undefended d5-rook>

click for larger view

click for larger view

Note how nasty chess can be. Here chess super-genius Carlsen overlooks that he must give the exchange after his blunder 61 ... ♗c3x♘d4?. What did he miss? If he allows White to play ♙x♗, then *White* wins!

http://www.chessvibes.com/reports/t...

<61. Nd4 Bxd4? 62. Rd2! Bc3

62... Rd8? 63. exd4 >

(VARIATION)
62 ... ♖d8-d5? 63 e3x♗d4

click for larger view

click for larger view

And it is *White* who wins because of the *enormous* strength of his passed d-pawn and the vulnerability of the Black kingside pawns to the White king.