No Sparing the Rod at the NH

Now (move 61) Nakamura has 12 minutes left on the clock [and three connected passers to deal with]. Will he "lose on time" again?

"In the ICC live commentary you (or Joel Benjamin) stated that Nakamura won numerous strong tournaments. "

Thomas, in my book he's only won two if I'm generous - and I exclude his US championships, cause they aren't elite events at all IMHO.

Both his wins are in spain:

2007: Magistral Ciudad de Barcelona (cat. 15)

(Rc = 2602), round robin with amongst others Krasenkow (2668), Gashimov (2663), Vaganian (2600), Beliavsky (2646) and Dominguez (2683) plus 4 2500-players. Due to Dominguez and Gashimov's later success and being improving players at the time, I consider this win a good one for Naka - and we can count the event as cat. 16 to be generous.

2009: San Sebastion (cat. 18)

(Rc = 2679) I think that's his only really strong tournament win in classical chess so far.

Of course, people might consider wins in the swiss open Gibraltar Masters noteworthy, but when Nakamura won Gibraltar in 2008, he faced an average of 2507 - equivalent to cat. 11 (eleven). In this context, I disregard that completely as a relevant event.

Of course, his US championship wins are comparable to the 2007 Barcelona win, but still - for elite players we don't normally count cat. 15 tournaments as "strong tournaments".

In summary, I think saying that Nakmaura has won _several_ strong tournaments is seriously stretching it. That doesn't mean he won't be able to establish himself as a top 15 player. Personally I think he will.

More answers to come in London and Wijk aan Zee - Nakamura is unofficially number 24 today, one place after Vachier-Lagrave.

"Most of the opinions I see here either rank Nakamura as a no-hoper or a World Champ, depending on his performance on the day."

Yeah, that's silly. Nakamura has been sick, and regardless, it wasn't obvious that he would be able to keep up his recent good form.

Consider how Ivanchuk varies and has streaks of good and bad form. It's not that long ago Nakamura had bad results in the Austrian league and in Foxwoods Open - after that he had some extremely good results, like the French League and San Sebastian in particular. And now this - probably a mixture of sickness, much travelling and some bad form too. (And like Mig mentioned, he was rather "lucky" to score some of his early wins in Ordix Open in rapids.)

No reason to jump to conclusions. Nakamura should forget about this event as soon as it's over and move on. London will be exciting.

"he or she."

You need fortune cookie.

Thanks frogbert - Nakamura seems to like Spain (including Gibraltar), but it probably still takes some time before he gets to play Linares and/or Bilbao ,:).

Mig's and/or Joel Benjamin's "stretchy" statement was part of an answer to a listener's question (qualified as "typically European") regarding Nakamura's true strength at present. Another part was saying that he is undoubtedly the world's strongest rapid player - it has been discussed [and doubted] before if this can be deduced from his ICC records ... . [maybe I misquote slightly, don't remember what they _exactly_ said].

In response to chesshire cat's comment: I consider Naka neither a no-hoper nor a World Champ (yet), but somewhere in between with potential for improvement. But I like facts ... .
BTW, as the live rating list is unofficial (and the NH Tournament probably won't be rated for the upcoming FIDE list), Nakamura should be officially top 20 in a few days (1st September).

Another short BTW: seems like Ljubojevic finally did today what Mig predicted to happen several times in the tournament - spoiling a position in time trouble.

"Nakamura should be officially top 20 in a few days"

I think #15 - but like I explained in a previous post: IMHO you're not a top 15 player due to once getting a top 15 rank. If we would count that way, we would have 25 top 15 players and at least 15 top 10 players at the moment, and that appears a bit nonsensical, doesn't it?

"The weak U.S. opposition during his formulative decade 1996-2005, during which even the unpolished Kamsky had retired. "

You obviously have an axe to grind. In terms of top 10 players the US is the 6th strongest country in the world and is only 2 and 1 points behind the 4th China and 5th Azerbaijan ranked countries. The 2008 & 2006 US Olympiad teams won the bronze medal. The 2004 team took 4th and included Alexander Onischuk, Alexander Shabalov, Alexander Goldin, Gregory Kaidanov, Igor Novikov, and former USSR Champion Boris Gulko. These are not weak players.

Regardless if you quote 100% correctly or not, anyone who thinks Nakamura is "undoubtedly the world's strongest rapid player" should... well, stop thinking that :) It's _clearly_ not true that he is _undoubtedly_ the strongest, and I would say it's not even likely at all, and seriously I have no idea why anyone would think that.

BTW, as per frogbert's post, I don't remember the time control in Barcelona, but it is highly disputable if San Sebastian counts as "classical" -- although it was rated.

I would say at any moment in time there are exactly 15 top 15 players - well, maybe 15 official ones (latest FIDE list) and 15 unofficial ones (live rating list), of course many names would be the same ,:).
What you mean I would actually call "world elite" or "super-GM". My opinion: In the still-official FIDE list (July 2009) #30 Ivanchuk would fall in that category - he has just re-reconfirmed his status. But I am not that sure about a few names ahead of him (Malakhov, Motylev, Eljanov, Bacrot?).

"I don't remember the time control in Barcelona..."

I looked it up for you: 40 moves in 90 minutes with a 30-second increment each move, then 30 minutes with a 30-second increment for the rest of the game.

"I would say at any moment in time there are exactly 15 top 15 players"

I explained it with regards to being a top 10 player; in order for _me_ to consider someone a top 10 player, that someone needs to be able to _defend_ a top 10 position over time.

Also, technically, since the "expectancy" goes up with a higher rating, it's "easier" in a sense to go from 2700 to 2730 due to a good streak, than to maintain the 2730 rating. Of course, usually one performs above 2730 in a number of games to jump from 2700 to 2730, but theoretically one can go from 2700 to 2730 by means of a 2715 performance over a sufficient number of games. This becomes harder and harder, though, due to the shortening of rating periods resulting in fewer games before the rating is adjusted.

But as a curiosity: when Kamsky got his famous 2650 rating and top 10 position as a teenager, he did that without _ever_ having a performance above 2620 - I once tried to calculate it accurately but I lacked some results (only a few), and the TPR over the games I did calculate was about 2610 - for the games that brought him to 2650 back in July 1990. One could consider that a "bug" in the system; the system simply didn't account for an improving player playing so many games in a "short" period. History did show, however, that in 1990 Kamsky was heavily overrated at 2650 - he wasn't at all able to defend that rating.

In summary: to me it's two different things,

1) once being ranked top 10, and

2) being an established top 10 player

A lot of players can make a visit to top 15 after a good streak - but _establishing_ oneself as a top 15 (or top 10) player is different - and harder.

She probably heard that she needs to be at the edge of her chair to do well here.

Well, here is another patzer who thinks that Lube job should have held the draw. If it was me, I would have run my King over and snatched the h6 pawn before it could hurt me. Something like 90...Kf7 followed by Kg6 and snatch. Draw. As it turned out, the h6 pawn killed Lube job.

By the way, how are Alexander Goldin and Igor Novikov doing? I haven't seen them mentioned a whole lot. I really liked Igor Novikov's games in the Semi Slav a lot.

@Luke

> That sounds crazy.

It is weird, but it usually doesn't matter much, because you have to play a huge number of games in one rating period, scoring consistently above your current rating - and that implies you're on an upward trajectory anyway. And in the next rating list the same effect works against you. If you're rated 2730 and score 2715 over a sufficient number of games, you're going down to 2700 again.

We patzers (I also thought Ljubo could hold the position) have company: In his live coverage on ICC, around move 40 GM Joel Benjamin was also at least 90% sure that black had a fortress - or that he could even try to play for a win with his passed d-pawn!? When Benjamin and Mig "signed off" around move 60-70, the final verdict was "the game may go on forever, but it's a draw".

Things may have started to go wrong for Ljubo when he lost his g-pawn (or gave it up? what was wrong with 72.-Kd6?). It still seems to be a fortress, but "it becomes easier to blunder" and at some stage, both players presumably had to rely on 30 second increments.

I wonder (and asked at Chessvibes) who the onsite experts were. When I was visiting earlier this week, I was curious enough to catch some glimpses of the action behind the (open) door marked "Press only". Among other people, GM John Nunn was discussing one of the games, with journalists typing notes into their laptops - Nunn also features on the video of Naka's KID win against Beliavsky. I guess and hope that today's experts weren't (just) Rybka and friends, engines are still pretty bad in assessing fortress positions.

On rating issues: I will try to explain, of course frogbert is welcome to correct me if I am wrong ,:) . In each tournament, your rating gain depends on the difference between expected score and actual score. Actually, this is even done per game (see below). So, taking frogbert's numbers, each time a 2700 player has a 2715 performance he gains rating points. The "trick" or system bug is that his expected score is still calculated from his official 2700 rating, until the rating list is updated. So let's assume he gains 2 rating points per tournament and plays 15 tournaments in one rating period, his new rating will be 2730.
But in the next rating period, if he still plays as much and at the same level, his rating will go down again back to 2700 ... because now he always scores less than expected for his 2730 ELO. As Bartleby already said, in practice it's rather unlikely to happen.

As I said, these calculations are actually done per game, that's how the live list is updated daily (FIDE does the same, but publishes the results only once every few months). Let's take Kramnik as an example and click on his name in the live rating list, then the last 10 entries (200907...) correspond to Dortmund 2009. He gained 0.2 points by drawing the higher-rated Carlsen, lost 0.5 points twice by drawing Bacrot, "+0.0" corresponds to his draws against Jakovenko and Leko [actually it should be something like +0.005 and -0.013 ,:)], which leaves +4.1 twice for beating Naiditsch, and +5.2 for winning against Carlsen.

"think you've chosen an entirely wrong example."

Not at all. I've done the math.

In july 1990, it didn't at all make sense for Kamsky to have a 2650 rating. He got from 2510 to 2650 while his performance over the games he played wasn't above 2610-20. I once made some research on it and posted a series of posts on chessgames.com.

1,5 years later - in January 1992 - he _was_ indeed able to defend his rating of 2650-ish. But 1,5 year is a long time for an improving teenager.

I'll try to find my old posts.

"By the way, how are Alexander Goldin and Igor Novikov doing? I haven't seen them mentioned a whole lot. I really liked Igor Novikov's games in the Semi Slav a lot."

Goldin is currently #15 in the US at 2557 and Novikov has a FIDE rating of 2560, but retired in 2006. Here's a Novikov link you may enjoy: http://www.chessgames.com/perl/chess.pl?pid=15649&playercomp=black&opening=D43-D49&title=Igor%20Novikov%20playing%20the%20Semi-Slav%20as%20Black

"Why not?"

Because FIDE ratings aren't adjusted based on rating performances per se - based on differences between "expected" and "actual" score. It's simply how the system works.

If rating changes always were calculated based on the latest rating numbers (after the players' previous game), this would've been a non-issue, but it's not really a big problem either - as long as organizers submit rating reports roughly on time; that is, at the end of the event, not several months later.

You can read about the FIDE rating system in the handbook, section B, on FIDE's site.

Thank you (and Thomas) for the explanation.

Warning: Long post, only for those who care a lot about Kamsky's 1990 rating of 2650 (and why I consider it a bit "odd").

Here are some excerpts from previous posts of mine in a different forum:

---

regarding kamsky, here's my projections based on his results:

july 1989: 2345 - heavily underrated
january 1990: 2510 - still underrated
july 1990: 2650 - somewhat overrated
january 1991: 2640 - still overrated
july 1991: 2595 - somewhat underrated (again)
january 1992: 2655 - "correctly" rated (finally)

the 6 months rating intervals introduced a bigger uncertainty (room for error) than
today's ratings, for obvious reasons. it can be noted that kamsky achieved his first
2650 rating without having faced more than 3+ 2600+ players - the scores i've got,
are:

kamsky - vaganian 2605 0-1
dreev 2605 - kamsky 1/2
vaganian 2605 - kamsky 1/2

he may have met a couple more in the us open championship 1990 and in bern open
1990. anyone with old tournament reports from the us open 1990, giving
the schedule or anything? :o)

more than that, though, it's clear that kamsky was terribly uneven in 1990 and 1991,
ranging from a 2470 performance in the manila interzonal 1990 (13 games) and a 2409
performance in linares 1991 (13 games) [while being rated 2650 and 2640
respectively], via 2580 in reggio emilia 1991, to 2714 (!) and 2674 in tilburg 1990
and 1991.

---

i can say alredy now that almost 70 (!) of those 140 points were due to one
single tournament result, played when kamsky unofficially would've been rated 2510
(and still be underrated) but officially was rated 2345 (! - maybe 200+ points
underrated, with a k of 15) - making him gain an insane amount of rating points in
just 9 games. :o) of course, he had an extremely good tournament result, as well, to
make that happen. (7/9 against 2523 average).

so, when kamsky started the year 1990 having and being counted as 2510, he already
was almost 2580 (unofficially), but enjoyed another 5 months of getting his rating
gain calculated based on a 2510 rating.

---

"I believe to have read here somewhere that the high numbers of games
Kamsky played prior to his 2650 rating had something to do with this too (?) Was
his average ratingperfermance in that periode LOWER than 2650 ?"

probably it was, but i lack about 10 individual results from that rating
period to know for sure. anyway, the most important thing is the massive gain of 69
rating points in december 1989, where kamsky earned an extra 23 points due to the
k-factor alone. if we assume a hypothetical october 1989 rating of 2401 (attributing
less than 1/3 of his rating gain from june to november to the first 3 months),
kamsky would've gained 40 points instead of 69 for exactly the same performance in
december. that alone would've made him nr 13 instead of nr 8, at 2620.

as things were, though, he gained 69 points which weren't counted yet - but kamsky's
strength obviously was getting close to 2600 +/- at the time. to see the effect of
having these 69 points "hidden in the luggage", we can consider the following: based
on the 42 rated games i currently have (including the 9 where he gained 69 points),
i've calculated _a gain of 105 points_. [so kamsky must have gained between 32,5 and
37,4 points in the "missing" 9 games to make the sum 140.].

however, if these 69 points had been counted in his january 1990 rating, giving him
2579 then instead, the total gain for these 42 games _wouldn't be 105, but 76_ -
alas he'd earned 29 points less for the same results in those 33 other games. of
course, technically or "legally" there was nothing wrong - fide followed its
procedures and rules. the "problem" was that the system didn't "treat very well" the
case of an underrated youngster improving fast while playing a tournament each
month. :o) that is, if one considers it worse that a young player might become
overrated at 16, than that he is more conservatively rated (and then possibly
underrated).

comparing to how it is today, when the top 10 start at 2740 - miles away from the
2400-border, we see another unusual thing about kamsky's case: he could jump into
top 10 helped by a rating gain boosted by a k of 15. that obviously isn't possible
anymore, not even theoretically - unless some rating report is delayed for so long
that it really should be dismissed rather than rated.

---

what seems to be the fact regarding kamsky, is that he

1) didn't perform consistently at a 2650-level prior to getting that rating

2) he certainly didn't perform consistently at a 2650-level (or higher) after he got
it either (despite the tilburg performance).

considering my database, we have the following:

july 1990 rating: 2650
next 6 months: 27 games
tpr: 2594 against 2585 average: +8 -7 =12
(tilburg, manila interzonal)

january 1991 rating: 2640
next 6 months: 33 games
tpr: 2535 against 2615 average: +8 -14 =11
(reggio emilia, linares, new york open)

was kamsky a top 10 player between his 16th and 17th birthday? technically, yes - in
reality, i doubt it very much.

i still prefer trying to obtain more results from tournaments i know kamsky
participated in, which isn't part of this data material yet, specifically the us
open championship 1990 and the national open 1990. there might be more, too - and i
need to know if games were rapid or classical. :o)

---

i can't know exactly, but my estimate is that kamsky in the roughly 50 games rated
for the july 1990 list had a performance of around 2620 - or the same level that
would've gained a 13th place in the july list (disregarding kamsky's 2650).

in those 42 games i have, where he gained roughly 105 (of 140) points, it was 2610
against a 2484 average, with only 3 games against 2600-players.

So what (moving the topic from ratings to chess) did Kamsky have to learn to do in order to hang with the élite?

Opening repertoire (some things never change) was part of his problem, but I also remember Seirawan (?) in Inside Chess saying that the Linares crowd had identified a technical weakness in Kamsky's game & were ganging up on it.

This remark may have been completely off-base, or it might suggest the work that Nakamura still has to do to crack the top 10. (The encouraging thing is that Kamsky worked through this phase very quickly: last place in Linares is a cheap price to pay for such an education.)

I believe your maths for the jump from 2510 to 2650. But I take issue with your notion that he wasn't at all able to defend it. Quite the contrary: in the next rating period he won one of the strongest tournaments of the year (performance must have been over 2700) and settled with 2640.

The interesting question would be: Where would Kamsky have been in January 1991 if ratings and expectancies would be adjusted game by game? It's clear he wouldn't have gone to 2650 in July 90, but I guess he wouldn't have been far off 2640 in January 91.

"But I take issue with your notion that he wasn't at all able to defend it. Quite the contrary: in the next rating period he won one of the strongest tournaments of the year (performance must have been over 2700) and settled with 2640."

Bartleby, in order to defend your rating, it's not enough to do _one_ good (or extremely good) tournament in the next period. See what I wrote one more time:

"
july 1990 rating: 2650
next 6 months: 27 games
tpr: 2594 against 2585 average: +8 -7 =12
(tilburg, manila interzonal)
"

INCLUDING the marvellous tilburg, his performance in the next rating period was 66 points lower than his 2650 rating - at 2594.

If you can say "quite to the contrary" and focus on Tilburg, I can say "he totally flunked it in the next rating period", focusing on the Manila Interzonal - his TPR there was 2470 over equally many games as his great Tilburg event.

One simply must consider BOTH events. In particular, considering how Kamsky did in the _next_ rating period, where his performance over 33 games (an even bigger sample than the previous period) was a lowly 2535 - more than 100 points under his official rating.

When considering whether or not Kamsky was able to defend his 2640-50 rating, you focus on 13-14 games at Tilburg, out of a total of 60 games played that year. Despite a 2714 (!) performance at Tilburg 1990 over 13-14 games, one must acknowledge that his performance over 3 times as many games (roughly 45) the same year was around 2500 - 150 points under his rating.

Or put differently: one very strong event, one decent (TPR 2587), and three (3) lackluster ones, considering his 2640-50 ratings.

But of course, if you define "defend a rating" as occasionally (here: 1 time out of 5) being able to perform at or (much) above it, then... Personally I consider this summary quite balanced and fact-based:

july 1989: 2345 - heavily underrated
january 1990: 2510 - still underrated
july 1990: 2650 - somewhat overrated
january 1991: 2640 - still overrated
july 1991: 2595 - somewhat underrated (again)
january 1992: 2655 - "correctly" rated (finally)

If you still disagree that Kamsky was "somewhat overrated" in july 1990, then I'm out of arguments. In my "unofficial" book, Kamsky wasn't a top 10 player before somewhere between july 1991 and january 1992.

Interestingly, Chessmetrics (missing the same 9 game results that I lack), "supports" the idea that FIDE's system had an odd effect on Kamsky's rating gain:

"based on slightly incomplete data, chessmetrics still somewhat enlightens the
"flaw" in kamsky's 2650 fide rating from 1990 - and i say "flaw" even if kamsky
delivered a totally amazing tilburg 1990 performance _after_ he got that 2650
rating. the highest ranking kamsky got in 1990 according to chessmetrics was #87 in
the world (june 1990), and kamsky actually didn't achieve a top 10 position until
december 1993, when he was 19. with complete data, i think chessmetrics would've had
kamsky somewhat higher in 1990, but hardly within top 40."

They started as scheduled, round 3 was played last night (CEST) - Onischuk lost against Bluvshtein in round 3 and is at 1,5/3, but had another "guest visit" above 2700 after his round 1 win. I'll update live ratings soon.

See http://www.echecsmontreal.ca/parties2009.htm for details.

"Opening repertoire (some things never change) was part of his problem, but I also remember Seirawan (?) in Inside Chess saying that the Linares crowd had identified a technical weakness in Kamsky's game & were ganging up on it."

Considering Kamsky's Manila Interzonal result, I guess Seirawan's (?) point - while probably having some truth to it - must be considered mostly anecdotal. I simply think Kamsky was still "immature" as a chess player at 16, and that this to a notable degree was the explanation for his huge variances in performances during 1990 and 1991. He was simply very uneven yet.

Or he might have been sick during some events - but it simply wasn't customary to make such excuses back then, so it didn't stick in the history books. ;o)

Seriously, during 1989 Kamsky seems to have followed a similar "system" to that of Carlsen almost up until now - using tournaments as a training vehicle. In addition Kamsky also studied A LOT - much more than I think Carlsen did at 16. And that's another reason why performances probably varied back then - I wouldn't be surprised if Gata simply was tired at times, but that his father didn't necessarily understand that _rest_ was an important part, too.

"History did show, however, that in 1990 Kamsky was heavily overrated at 2650 - he wasn't at all able to defend that rating."

Looking back I agree that "heavily overrated" was overstating it more than a bit, so I understand your initial reaction. But I maintain that he wasn't at all able to defend the 2650 rating. :o)

I was more careful with my wording 1,5 years ago, when I first considered the story behind Kamsky's early top 10 position in 1990.

"somewhat overrated" is probably something you can agree to, Bartleby.

What happened to Onischuk in his game against Bluvshtein? I'm asking you because you seem to follow his every move. Isn't he winning at the end with 37...Rxb1+ 38.Bxb1 cb4? Isn't this simple, obvious, and almost forced? Did he lose on time before making his 37th move? Really strong players should not lose on time. When did Kasparov, Fischer, Anand, Kramnik, Capablanca, etc. ever lose on time?

Botvinnik I can understand because he frequently burned up his clock, as did several of his Soviet contemporaries back in the 1950s (Bronstein and Korchnoi come to mind). If I remember correctly, Short flagged in a winning position against Kasparov in 1993. Stupid.

What about Morphy? Tal? Alekhine? Lasker? Spassky? Steinitz? Petrosian? Topalov?

Losing on time is inexcusable, especially when you have the advantage on the board. If that's what Onischuk did against Bluvshtein, he should be ashamed.

No, I have never lost on time, but have come close.

If you go that far back in time (Morphy, Tal, Alekhine, Lasker, Capablanca), don't forget that time controls used to be slower back then - I am not even sure that chess clocks already existed when Morphy was playing.
From my own experience, I still sort of remember one rapid game ~15 years ago when I won a piece with a series of subtle and "time-consuming" moves and then flagged. Several people, including my honest opponent, considered me the "moral winner" of the game - of course it didn't affect the official result ... . And I don't see why such things cannot, should not or must not happen at a higher level. I concede that rapid games (without increments) are not directly comparable to classical time controls.

@frogbert

No, no. Don't be so cautious.
If your number of 2594 for the second half of 1990 is roughly correct, and I have no reason to doubt it, I come over to your side.
You are right, I've been wrong.

I liked the fairy tale of the young immigrant who was ridiculed for his Swiss-inflated rating by the established GM, and then went on to win the very first elite event which he was invited to.
The story is still true of course, but as you showed, the top 10 rating is just an artifact of the rating process. It certainly got him publicity back then.

When I looked up the numbers I first thought they would corroborate my version. I don't have the individual tournament evaluations, just the old informators with the tournament results and rating lists. Somehow my numbers still feel strange: He gained 140 points by performing roughly 100 points above his rating over 51 games. By that standard - if he had 2650 rating, and an under-2600 performance, shouldn't he lose more than 10 points over 29 games? (The rating list says 29 games.) But that's probably another technicality. Maybe Manila was calculated with his old 2510 rating, or my feeling for the bell curve is wrong.

Sorry for interrupting. I appreciate your concept of an established top 10 player. I would even like to see it formalized, and base the designation "Real GM" on it.

"I simply think Kamsky was still "immature" as a chess player at 16, and that this to a notable degree was the explanation for his huge variances in performances during 1990 and 1991."

To some extent, huge(?!) variances in performances also seem typical for the "current" Kamsky - post-comeback, older and wiser (at least old and wise enough that his father no longer plays a dominant role in his life). So it may be not only a question of age, but also of personality and playing style. Kamsky had an excellent result winning the World Cup, at least an acceptable result in his match against Topalov [many people, not only Topalov's fans, had predicted a very one-sided affair] and some mediocre or poor tournament results. Only Bazna 2009 (3/10, losing 18 ELO points) would be very poor in my book, though.

"Implications" might be mostly an American affair:
Is Nakamura now far stronger than Kamsky, and will keep his lead for the rest of their lives (or chess careers)? At face value, current numbers (July 2009 FIDE list) are:
Kamsky 2717, expected change -25
Nakamura 2710, expected change +25 (not including his NH result)
I wouldn't consider this an irreversible trend yet ... . This discussion would be on a more secure ground if the USA had organized one GP event, handing a wildcard to Nakamura thus allowing him to regularly play "Kamsky's opponents" as well as Kamsky himself more frequently (so far, strangely their draw at the US Championship seems to be their one-and-only game at classical time control).
More ugly were comments mostly on Susan Polgar's site saying that Kamsky should quit professional chess because he is simply not good enough ... . Given the tone of some comments, it hadn't even surprised me if some "chess fans" suggested that Kamsky should commit suicide (maybe I missed such comments, or they were moderated away).

"And I don't see why such things cannot, should not or must not happen at a higher level." (Thomas)

It should not happen because you are supposed to be like a general on the battlefield in the heat of battle with bullets flying and your soldiers are depending on you to issue the orders necessary for their survival...and the general goes mute with brainlock? Absolutely inexcusable.

"More ugly were comments mostly on Susan Polgar's site..." (Thomas)

I am shocked.

From the ICC live coverage: According to Stellwagen's second l'Ami the game was home preparation until and including 27.-e4. Somehow they then switched off their computer as black looks completely winning - if they had continued just a few minutes the mate would have been part of their home preparation !!?

Quote from Thomas:

Given the tone of some comments, it hadn't even surprised me if some "chess fans" suggested that Kamsky should commit suicide (maybe I missed such comments, or they were moderated away)

Man, this is so annoying...

Now I have only taken a brief look at the game, but chances are that if you prepare a complicated line to move 27, you will have had a look at a myriad of sidelines as well. You simply don't have infinite time to look at everything all the way through to mate.

Take Lékó's win against Gelfand in the Nalchik Grand Prix as an example. Everything up to 26.Qg3 was prepared, but he had to find the very nice winning continuation after 26..Ne5 on his own. He had just noted that the computer said White's winning there, but not how.

"I'll update live ratings soon."

01 Topalov off 2813,0 0 0 0 1975 id-card
02 Anand off 2788,0 0 0 0 1969 id-card
03 Aronian off 2773,4 +5,4 13 1 1982 id-card
04 Carlsen off 2772,4 +0,4 10 1 1990 id-card
05 Kramnik off 2771,6 +12,6 10 1 1975 id-card
06 Leko off 2761,9 +5,9 23 2 1979 id-card
07 Radjabov off 2756,6 +0,6 10 1 1987 id-card
08 Gelfand off 2755,9 +0,9 33 3 1968 id-card
09 Ivanchuk off 2755,8 +52,8 42 4 1969 id-card
10 Morozevich off 2750,0 -1 18 2 1977 id-card
11 Jakovenko off 2741,7 -18,3 24 3 1983 id-card
12 Ponomariov off 2740,9 +13,9 9 1 1983 id-card
13 Gashimov off 2740,0 0 0 0 1986 id-card
14 Svidler off 2739,7 +0,7 17 2 1976 id-card
15 Grischuk off 2733,0 0 0 0 1983 id-card
16 Shirov off 2729,9 -2,1 18 2 1972 id-card
17 Wang Yue off 2727,8 -8,2 10 2 1987 id-card
18 Alekseev off 2725,1 +11,1 23 2 1985 id-card
19 Karjakin off 2722,4 +5,4 13 1 1990 id-card
20 Malakhov off 2721,5 +14,5 15 2 1980 id-card
21 Mamedyarov off 2721,2 +4,2 2 1 1985 id-card
22 Dominguez off 2719,0 +3 13 2 1983 id-card
23 Vachier-Lagrave off 2718,3 +15,3 31 4 1990 id-card
24 Nakamura off 2717,8 +7,8 25 4 1987 id-card
25 Eljanov off 2716,5 +0,5 13 1 1983 id-card
26 Motylev off 2716,3 +6,3 6 1 1979 id-card
27 Bacrot off 2711,3 -9,7 25 3 1983 id-card
27 Movsesian off 2711,3 -4,7 9 1 1978 id-card
29 Short off 2705,6 +21,6 10 1 1965 id-card
30 Rublevsky off 2703,0 0 0 0 1974 id-card
31 Wang Hao off 2702,4 +12,4 6 1 1989 id-card
32 Kasimdzhanov off 2701,9 +29,9 30 3 1979 id-card
33 Onischuk off 2700,4

So what? An updated rating list. Big deal.

"Somehow my numbers still feel strange: He gained 140 points by performing roughly 100 points above his rating over 51 games. By that standard - if he had 2650 rating, and an under-2600 performance, shouldn't he lose more than 10 points over 29 games?"

The answer to that is partly in my _very_ long post above. The main points were:

* about 70 of the 140 points were gained based on a very strong performance (2734! or nearly 400 points his official rating) in Baleares open in december 1989, where
- Kamsky had a 2345 official rating, and hence
- the event was rated using a K of 15 for Kamsky
* he gained a lot ratingwise by having 70 "secret" points "in his bag"

This event was played after the deadline for the 1990 January list (they actually needed a full month back then to do the math), and hence it only was included when we got to the July 1990 list. Hence, Kamsky's "live rating" in January 1990 wasn't 2510 but 2579 - but he went on to play another 42 games being counted as 2510, which was very "advantageous" for his following rating gain.

Put differently, if either of the following would've been the case, he'd never been top 10 before 1992 (given the exact same game results):

* Baleares had taken place and been submitted for rating prior to December 1st 1990

* Baleares had been rated with a K of 10 (instead of 15)

* The rating periods had been 3 months (instead of 6 months)

But it does appear like I lack also two games in my data for Kamsky for the period June 1990 to November 1990 (inclusive), and 10 game reults for the period December 1990 to June 1991 (inclusive). This is in addition to the 9 game results I miss for the rating period December 1989 to June 1990.

It would be greatly appreciated if anyone would help me collecting the entire list of results. I'll put a summary of missing results in my next post.

Maybe Caruana should also get some credit for finding 32.-g5 33.hg5: h4 ? Even more surprising (and still a bit mysterious) to me is that Hou Yifan suddenly and quickly collapsed against Beliavsky, but this game also took two: one doing something wrong, the other one doing something right ... .

"Where is Naiditsch at the moment?"

Isn't he in Canada? ;o)

Yes, lots of credit to Caruana. I did not see the 32...g5 idea at the time, and even after he played it, I briefly saw 33...h4 as a possibility but discounted it thinking he needed to either bring his King to g6 first or play 33...hg4. Of course, Svidler's 35.f4 was totally senseless and he was definitely lost after that.

Hou Yifan just panicked for some reason. She played like a butterfly.

Informator 50 has a tournament New Delhi in December 1990. Category XI, 2508.

1-2. Anand, Kamsky 8 3-4. J.Polgar, Zsu.Polgar 6.5 5-6. Barua, Thipsay 6 8-9. A.Cernin, Zso.Polgar 5 10. Prasad 4.5 11. Ravisekhar 3 12.Sudakhar 2

That's probably where the Anand-Kamsky game is from. But it must have been more than 2 rated games, so it probably was too late for the January 91 list.

Bartleby, thanks.

Does Informator say anything about the length of the event? If it was classical, then that means 11 more games for the July 1991 list.

Reggio Emilia (12 games)
Linares (13 games)
New York open (8 games)

That means that I have an "overflow" - but now I see that round 1 in New York open was against an unrated opponent (at least the database has no rating for him). Then probably these are the 43 games for the July 1991 rating list:

Reggio Emilia (12 games)
Linares (13 games)
New York open (7 rated games)
New Delhi (11 rated games)

Kamsky scored 8/11 there, and games weren't rated game by game back then, so I have all info I need both for performance and rating calculation. Thanks again!

Then we're down to

July 1990 (9/52 games missing)
January 1991 (2/29 games missing)
January 1992 (10/43 games missing)

I'm easy to please, so closing these gaps would actually make me quite happy. :o)

With 35.f4, Svidler at least got rid of black's passed d-pawn. Else black would have played Kg6 to pick up the pawn on g5, and then the white pawns on e5 and even f3 may also be weak (ideas like Rc5 or Kf4 with Ng5 to follow). And to me it seems that black's pawn on h4 is "surprisingly healthy".
Conclusion: maybe there were better moves or defensive plans for white, but I wouldn't disqualify 35.f4 as "totally senseless".

Some information from the ICC live coverage: Much of the game was home preparation by Stellwagen and/or his second l'Ami. "First version" was until and including 27.-e4 (hence just one move before Stellwagen missed a forced mate), later it was said that the preparation ended after white's 25.Nd6: evaluating the position as fine for black. Then 26.-Nd5! ("a typical KID sacrifice", l'Ami's words) was found over the board.

And Stellwagen + l'Ami were primarily inspired by Radjabov, more precisely by the Olympiad game Van Wely - Radjabov (one of a series of games both players had in the Bayonet Attack of the KID over the years). Van Wely won that game, Stellwagen was following from nearby (they were on the same team!). Amazingly, only 24.-Qf3 was the novelty. Radjabov played 24.-h5, Van Wely first neutralized the threat Bh6+ with the simple 25.Ng7:, and after 25.-Kg7: played 26.Qb3 Qg2 27.Qe3 when black's compensation turned out to be insufficient. It remains to be seen if Stellwagen's novelty (again) changes the assessment of the entire line ... . L'Ami also stated that it makes sense to look for improvements in a game that your opponent had won, because he may not analyse the position that thoroughly.

The predecessor game can be found at
http://www.chessgames.com/perl/chessgame?gid=1519975

That's ok. Think what you want to think.

Well, I gave some ideas and plans (including the idea behind Svidler's 35.f4) - you didn't support your "totally senseless" statement [pun intended] by giving alternatives for white.
Think about it ... .

Nothing to think about. You think what you want to think. What's wrong with you? Don't start up with me.

> Does Informator say anything about the length of the event?

No, it doesn't mention anything special, which implies it was classical.

"1-2. Anand, Kamsky 8 3-4. J.Polgar, Zsu.Polgar 6.5 5-6. Barua, Thipsay 6 8-9. A.Cernin, Zso.Polgar 5 10. Prasad 4.5 11. Ravisekhar 3 12.Sudakhar 2"

Did it also say who finished in 7th place? :o)

I guess "A.Cernin" is a spelling mistake for Alexander Chernin, and I was unable to find "Sudakhar" too ...

As I also happened to find back Informator 50 (collecting dust somewhere on the bottom of the pile): 7th was Torre.

Excellent - thanks! And as "predicted", he was rated 2530 at the time. My first informator was number 60 - and they almost went out of "fashion" before I started buying them regularly ;o)

3/3 at Montreal for Naiditsch not 2.5/3

Then we have the following about Kamsky's first 12 months after achieving the 2650 rating (due to a chain of "fortunate" events):

Based on 70 of 72 rated games, he scored 34,5 points (of 70 of course) against an average of 2584 for a performance of 2577 - which is quite far away from both 2650 and 2640.

Hence, I consider it safe to say that his "actual" strength between July 1990 and July 1991 probably was a bit below 2600 rather than in the vicinity of 2650 and top 10 in the world.

Even counting the entire period december 1989 to july 1991 (about 120 games) his over-all performance was "only" around 2600 - in fact, probably a little below, but I can't tell for sure until I get the 9 (11) missing results.

Thomas, you don't happen to have anything in your Informators about Bern 1990 or US open 1990? Naiditsch is at 2692 sharp after yesterday's game, btw ;o)

Moiseenko (2682) - Naiditsch (2697) [D41]
TIM (3), 2009

1.d4 Cf6 2.c4 e6 3.Cc3 d5 4.Cf3 c5 5.cxd5 cxd4 6.Dxd4 exd5 7.e4 Cc6 8.Fc4 Cxe4 9.0-0 Cf6 10.Fg5 Fe7 11.Fxf6 Fxf6 12.Dxd5 0-0 1/2-1/2

The "(3)" means round 3. Taken from the official home page of the event.

"Bern 1990 or US open 1990?"
No, sorry - this should be Informator 49 which I don't have (maybe I once had it but I moved ~9 times since 1990 ,:) ). In any case, for open tournaments Informators would be of limited help: they list the final standings but not all opponents of a given player - at most the 'memorable' games featured in the main part.

Bartleby, I also recalculated Kamsky's performance during the 43 games for the July 1991 list - which became roughly 50 points higher when including the good 8/11 result in New Delhi (compared to my initial calculations based on only 32 games). We then have:

july 1990 rating: 2650
next 6 months: 29 games (2 missing)
TPR: 2594 against 2585 average: +8 -7 =12
(tilburg, manila interzonal)

January 1991 rating: 2640
next 6 months: 43 games
TPR: 2589 against 2603 average (20,5 / 43)
(reggio emilia, linares, new york open, new delhi)

Seems reasonable that this landed him on a 2595 rating for the July 1991 rating list.

Here are the rating adjustments for the 70 games (6 events) I now know:

Manila (January 1991 list):
E = 13 * 0,67 = 8,71
change = (5,5 - 8,71) * 10 = -31,2

Tilburg 1990 (January 1991 list):
E = 14 * 0,51 = 7,14
change = (8,5 - 7,14) * 10 = +13,6

Total: -17,6 points

Hence, in order to only drop 10 points, from 2650 to 2640, he must have gained at least 5,2 points on the two unknown games (which corresponds to 1,5/2 against someone rated roughly 2650); if his change had been -12,4, it would've given a 10 point drop.

---

New Delhi 1990 (for July 1991 list):
E = 11 * 0,71 = 7,81
change = (8 - 7,81) * 10 = +1,9

Reggio Emilia 1991 (for July 1991 list):
E = 12 * 0,51 = 6,12 points
change = (5 - 6,12) * 10 = -11,2

Linares 1991 (for July 1991 list):
E = 13 * 0,47 = 6,11
change = (2,5 - 6,11) * 10 = -36,1

New York open 1991 (for July 1991 list):
E = 7 * 0,71 = 4,97
change = (5 - 4,97) * 10 = +0,3

Total: 1,9 - 11,2 - 36,1 + 0,3 = -45,1

And that fits perfectly with Kamsky's drop from 2640 (in the January list) to 2595 (in the July 1991 list).

"Could just be that his "actual" strength (your quotes) was 2650ish, and he consumed a bad run?"

I you would reread (or read for the first time) the large number of posts, you would've seen that Kamsky did NOT perform at 2650 for any prolonged period of time neither prior to getting that rating, or for the next year after receiving it.

Specifically, in the 51 games leading up to the July 1990 list, his performance was at most ca. 2620 - hence, 2650 is a "feature" (flaw) of the rating system, due to some unlikely things that came together to produce that rating.

"this whole over/under rating business is largely nonsense."

I'm willing to discuss different terminlogy, but let me say that I only use over/under-rated about something in the past, where I can study rating development at a later time.

If you over _many games_ over an entire year perform 60-70 points under your official rating, you appear to have had a misleading rating number prior to that year. Since a rating number technically only describes PAST results, one could argue pramatically that nobody EVER is overrated or underrated, they simply have the rating number that their actual results dictate.

However, in the case of Kamsky in 1990, he never ever (again: for more than single events) had performed at the level of his 2650 rating prior to receiving it - and I meticulously explained how that happened. Hence viewing the period July 1990 to July 1991 as "variance" and a negative trend from a former, higher level - as you suggested - doesn't make sense here.

The fact is that Kamsky's performances were varying heavily from december 1989 to July 1991 - but he'd never played better before or performed on average as good in any prolonged period before those 18 months.

In reality, I think Kamsky was gradually improving all the time during those 18 months, but due to artifacts of the rating system, he got a too high rating in July 1990.

The performance of the 51 games in the first 6 months, estimated to ca. 2615-2620, isn't significantly higher than the calculated performance of ca. 2580 for the next 70 games over 12 months. I completely agree that this is within the "natural variance" one can expect - and if we'd put the sampling points a bit differently, those 120 games could've been divided into two periods of 2600 performances - and obviously there is no "right" points to take the samples.

The MAIN point is that 120 games usually is a more than big enough sample of games to get a very GOOD performance measure - and in this 18 months period his average performance was around 2600 and not anyplace near 2650.

"Your "safe" world lacks nominal variance."

I think you assume too much about "my world". :o)

"Could just be that his "actual" strength (your quotes) was 2650ish, and he consumed a bad run?"

Like I said - ratings describe prior results more than "strength", but of course there is correlation between the two. But your hypothesis that Kamsky's "strength" was 2650-ish around January 1990 and he went on to have a "bad run" with a 2600 performance over 18 months and 120 games does not compute.

Kamsky had a GREAT run over these 18 months, the best 18 months in his life so far. In November 1989 his "live rating" was 2510, but still he went on to perform at around 2600 for the next 18 months - with heavy variations, but still.

The "historical mistake" here, is that it seems that nobody realized that his July 1990 rating didn't describe his PAST results in an adequate way. Still, this probably helped Kamsky's career, even though it probably was frustrating seeing one's rating drop more than 50 points while Kamsky in reality CONTINUED to improve and play the best chess of his life so far.

"NEEEERRRRDDDDSS!!!"

Thanks, I take that as a compliment. One "nerd" I know of, Bill Gates, didn't do too bad. :o)

I may be the last man to be a fan a Naka, but give him a break... the guy's so sick he throws up and nosebleeds...

Oh, if you would made the kind of money Bill Gates has made, I'm sure you would've easily taken on the responsibility for Windows Media Player too. :o)

This was a case of Nakamura losing his tactical eye and not seeing that 24.Re1? was crushed by 24...Ng4! Otherwise, 24.Rd1 and he was ok.

He'll need to re-evaluate things after this event.

Looking at the final position, it seems Naka's main mistake was that he forgot to develop his queenside pieces? This was asking for trouble, even if he was objectively OK at a certain stage.

After this event, first and foremost he has to get healthy again, and maybe he needs to re-revaluate his travelling schedule. Will he take some rest now, or will he continue accumulating frequent flyer miles?

"You said that one should not heed the horrors one has unleashed as long as one has sufficient material reward."

Technically I didn't say that. I said that I think you would've taken the dollars if given a choice. But that was a mere speculation, of course.

So what _would_ you choose?

a) one billion dollars + the credit for WMP

b) no association with WMP whatsoever

Remember the 10 commandments when you answer!

:o)

According to his homepage, his next event is the European Club Cup in Ohrid (Macedonia), starting Oct 3 (day of arrival).

"Just because Kamsky blipped to 2650 due to systematic flaws, doesn't mean that he wasn't actually that strong, even if he only performed at 2600 for the next 70 games."

Technically right, but again - if we're taking your statistical approach - we should consider what is most likely - which possible interpretation has the highest probability, in other words.

And again, his performance over those 120 games was almost 2600, and "only" (at most) 2620 prior to the 2650 rating.

But I don't think there are big disagreements here. :o)

Who are you to call someone a troll? What about me? What about Thomas, Manu, chesshire cat?

What a disastrous tournament for Nakamura. Dont know if he should have withdrawn, admire him for his sportmanship to hang on and play through illness. I hope his fighting spirit will be noted by the Melody Amber Rapid so he gets an invitation

I agree. I guess the agrument is that the older players are likely to be less active, not as current on their theory, and not as spritely with their tactics. I quite like these match ups--they take the what-ifs off the table.

What kind of invitation would that be? To be a spectator? Certainly not to be a player.

This tournament had very little to do with team play and all about who gets the invitation to Melody Amber.

I don't think Nakamura ever thought of himself as part of a team. In his interview, he kept on saying stuff like "basically, the tournament is over for me." He was more concerned about not being able to catch Smeets. It was all me, me, me.

@frogbert
b) of course! Without question.
Do you remember the parable Luke 18,18-25? And there are many other texts, like Luke 6,24 etc...
Bless!

"I agree."

Thanks. The trolls and ratings freaks may drive me out of posting, but I'll still read Mig.

A glance at the latest list (Kramnik 2772; Carlsen 2772) shows the biggest problem with the rating system: rounding.

It's totally ridiculous that two players who might have significantly disparate ratings (e.g. 2771.51 and 2772.49) should be lumped together.

Without ratings calculations to at least two decimal places there's just no way to predict who would win a Kramnik-Carlsen game or match.

"ratings freaks may drive me out of posting"

Didn't you post an update from the live rating list on this very page not long ago? :o)
Onischuk freaks may drive me out of posting, but I'll still read Mig. ;o)

"Didn't you post an update from the live rating list on this very page not long ago?"

Yes. You said you were going to so I was trying to do you a favor. I'm not obessed with ratings calculations, the early 90s, Kamsky, and Nakamura. Some of us still live at home or school, others own homes. Nuff said.

"Without ratings calculations to at least two decimal places there's just no way to predict who would win a Kramnik-Carlsen game or match."

Haha. That's so very witty. And a real bulls eye as a comment to the recent rating discussion here, where you _could_ have observed that performance differences of 40 points were described as insignificant by for instance yours truely.

But I guess it's more fun to pretend that everyone who takes an interest in ratings hasn't got a clue about their meaning.

Isn't that so, Greg? ;o) It's good that straw men are good at taking bullets - they typically go straight through...

"Some of us still live at home or school, others own homes. Nuff said. "

Huh?

Actually the FIDE list still says 4 Carlsen 2772, 5 Kramnik 2772. The live rating list more accurately says 4 Carlsen 2772.0, 4 Kramnik 2772.0. Both lists put Aronian on third place with 2773 and 2773.0, respectively ... . (there is some irony in this, but it's not directed against frogbert!)

Of course these differences are meaningless, though such subtle differences could become relevant for the rating spots into the candidates tournament (if I remember correctly, two digits behind the comma will be considered if required). Good that Aronian is already qualified via the Grand Prix, else it might come down to a real lottery between the three players mentioned. [Yes, I know that the average of July 2009 and January 2010 will count, and the new September list is irrelevant]

Well, I guess it's no point telling Onischuk fan about my $900k house then or pointing out that the reason the Live Top List site isn't plagued by ads is that I don't want any there (it produces enough traffic to drive some income for a poor student).

Onischuk freak, you're right - I live in a small room in my parents' house, expences paid, and I still go to school at 37. That's why I can afford having a hobby. I don't understand why that makes me a freak, though. A rating freak, I mean. Living with one's parents at 37 is a bit awkward, of course, but we get along nicely...

I am more than certain that you have a deep understanding of chess ratings and a remarkably balanced and healthy attitude toward them.

My earlier post was, of course, directed at folks who obsess over every five or ten rating points.