chessgames.com
Members · Prefs · Laboratory · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing

🏆 Tata Steel Masters (2016)

Chessgames.com Chess Event Description
... [more]

Player: David Navara

 page 1 of 1; 13 games  PGN Download 
Game  ResultMoves YearEvent/LocaleOpening
1. Navara vs Carlsen ½-½312016Tata Steel MastersD37 Queen's Gambit Declined
2. Adams vs Navara  ½-½312016Tata Steel MastersB12 Caro-Kann Defense
3. Navara vs Giri ½-½422016Tata Steel MastersD97 Grunfeld, Russian
4. Yifan Hou vs Navara 1-0402016Tata Steel MastersB12 Caro-Kann Defense
5. Navara vs So ½-½462016Tata Steel MastersD38 Queen's Gambit Declined, Ragozin Variation
6. Ding Liren vs Navara ½-½412016Tata Steel MastersE60 King's Indian Defense
7. Navara vs Karjakin  ½-½352016Tata Steel MastersD45 Queen's Gambit Declined Semi-Slav
8. Navara vs Caruana 1-0552016Tata Steel MastersE32 Nimzo-Indian, Classical
9. Wei Yi vs Navara 1-0282016Tata Steel MastersC65 Ruy Lopez, Berlin Defense
10. Navara vs Mamedyarov ½-½312016Tata Steel MastersC76 Ruy Lopez, Modern Steinitz Defense, Fianchetto Variation
11. Van Wely vs Navara ½-½372016Tata Steel MastersA04 Reti Opening
12. Navara vs Tomashevsky  ½-½322016Tata Steel MastersE15 Queen's Indian
13. Eljanov vs Navara 1-0362016Tata Steel MastersE12 Queen's Indian
  REFINE SEARCH:   White wins (1-0) | Black wins (0-1) | Draws (1/2-1/2) | Navara wins | Navara loses  

Kibitzer's Corner
< Earlier Kibitzing  · PAGE 65 OF 67 ·  Later Kibitzing>
Feb-15-16  frogbert: <Olavi: With a rating period of say 6 months it is fully possible to lose or gain 30-40 points because of colours.>

Fully possible, or theoretically possible? Consider these numbers, Olavi:

These days the rating period is 1 month in FIDE. Over time, players tend to play more or less an equal number of black and white games. Of course, these days it would make sense to add colours to the rating calculations (since they're now done by computer programs and not humans), but in the long run I doubt it would make much of a difference.

At the elite level I think I've seen numbers saying that white scores 54% versus 46% for black. In 10 games this almost translates to white scoring 5,5 points versus 4,5 points for black; the difference is slightly less.

However, for huge samples (350-450 games, ref the stats I gave in a previous post), players play around 50% of their games with either colour - and less than 51% for those I sampled.

With a K of 10 for all FIDE players rated 2400+, 10 games with white would statistically give (only) 8 ratingpoints extra [ (0,54 - 0,46) * 10 * 10] for having the white pieces. In reality players have less than 51% white. Hence, the rating gain in 10 games <if> you have 51% white games instead of exactly 50% (this is only a theoretical example, of course, since it's impossible to play white in exactly 51% of 10 games), is as follows:

0,02 (2% extra whites: 51% = 102% of 50%) * 8 rating points = 0,16 rating points.

Hence, in 20 rated games it amounts to about 0,3 rating points, which on average is rounded up (1 point) once every 3 monthly list.

For Carlsen's stats, which indicate only 1% extra whites (50,5% = 101% of 50%), this translates to 0,01 * 8 points = 0,08 points per 10 games. Hence in 20 games in a rating period, there's a gain of 0,16 points, which on average is rounded up (1 point) once every 7 (seven) monthly list.

Is this a problem in the grander scheme of things? Obviously not. It's also worth noting that a rating increase impacts the score expectancy formula, thereby causing a correction in the other direction.

In short: Yeah, one would slightly increase the accuracy of the rating formula by taking colours into consideration, but no, it would have essentially no practical consequences except making it harder to do rating calculations for anyone not using a fancy computer program.

Feb-15-16  frogbert: < The significant thing is what rating did the player you won, draw or lost against have when you played with each colour.>

Again, this averages out over a realistic number of games. If you don't "buy" that, even after reading my latest post, I don't think I can do much better. There's nothing to buy here, it's simply about understanding math and statistics.

Feb-15-16  Olavi: <frogbert> You did not really need to post those numbers. They are self-evident.

But it is not so unusual to play 8 blacks in a row. It happens in team tournaments, has happened to me. Or check Nunn in Debrecen 1992, or Sokolov in Dubai 1986 (the reverse), if I remember correctly.

Feb-15-16  frogbert: <Olavi> Maybe there are readers out there who consider such numbers less self-evident than you do. :)

Yes, these things happens, particularly in team events. I played quite a few games for my club Akademisk while we were part of the Norwegian National League ("Eliteserien"), and for some reason about 75% of those games were with black. But at my level (even though most of my opponents were master level), colour seldom decides the outcome of the game. ;)

Still, 8 games in a row with black - at the elite level - only gives a theoretical drawback of 0,08 * 10 * 8 points = 6,4 points.

In order to lose 30-40 points, you would need to play 5-6 times as many games with black in a rating period, assuming a K of 10, right? (And assuming opposition of equal strength.)

Feb-15-16  Clemens Scheitz: <frogbert>, Of course things have a tendency to "average out" over a large number of games, I'm only stating that the formula is far from perfect (like statistics itself). You said it yourself "one would increase the accuracy of the formula by talking colours into consideration". As far as how slight or what kind of practical consequences, that's subjective and a bit irrelevant now with the power of computers. I say let's push for an improvement.
Feb-15-16  frogbert: <As far as how slight or what kind of practical consequences, that's subjective>

Only to a certain extent. I gave a couple of real, objective examples, which should convince most people that we're only talking about a very minimal "improvement" in accuracy.

If you think about it, the fact that it's only possible to score 1, 1/2 or 0 points in a single game, the closest thing to the expected score you might achieve in a single tournament is somewhere between -0,25 and +0,25 of your expected score.

Hence, even if you perform as close to your expected score as humanly possible, you will gain or lose between 0 and 2.5 points in such a tournament. This deviation is <much larger> than what your colour distribution might lead to in a normal event - and sometimes even larger than what 3 extra games with white/black will result in, theoretically. It's all about keeping a perspective on matters.

Feb-15-16  frogbert: <Hence, even if you perform as close to your expected score as humanly possible, you will gain or lose between 0 and 2.5 points in such a tournament.>

And that's assuming a K of 10. With Ks of 20 or 40, you can double or quadruple that number.

Feb-15-16  Rolfo: I happen to know that frogbert is a computer expert undrstanding math and computing, at the same time he is a decent chessplayer. So , very good argument frogbert
Feb-15-16  Sally Simpson: HI Frogbert,

Re: A computer determining the world champion. Of course I was joking.

These discussions should always be conducted on a light hearted matter. All to often after any slight disagreement on here they descend into name calling on the first post.

A bank of computers all running on the same mainframe with 30 seconds a move should do it.

Instead of thinking of the best move you would be thinking of what would the computer play. Any undetected cheat using Rybka 13 would become world champion.

I picked Rybka because I pass their Edinburgh office a few times every week. It has stopped trying to solve chess and is now working on climate control.

Thats's not joke.

http://www.rybka.co.uk/documents/Cl...

I've never been worried about or phased about how good computers are v humans and it may kill chess.

People will always play chess, but this climate control thing has me worried.

Once they feed in all the dats it will come to the conclusion that the best way to save the planet's climate is to get rid of all the humans.

And as computers now control everything I think working together and led by Rybka they could do it.

Feb-15-16  frogbert: <Once they feed in all the dats it will come to the conclusion that the best way to save the planet's climate is to get rid of all the humans.>

Hehehe. :) Unless our technology will advance to a point where we're able to destroy a deadly meteor before its impact with earth, of course. In that case humans would be the only species on earth capable of saving our environment from disaster. ;)

Feb-15-16  Sally Simpson: Hi Frogbert,

You are laughing now but Rolfo has told Rybka (it reads these threads) that you are a computer expert. You will targeted as a threat.

Mr too because I know where the Rybka Edinburgh Office is.

Feb-15-16
Premium Chessgames Member
  AylerKupp: <frogbert> The reason why I think that your definition of systemic inflation requires a perfectly closed system with no players entering or leaving the pool is because I thought that's what you said so in Hans Arild Runde (kibitz #4317):

"what do i mean with systemic inflation? well, if we assume the simple scenario with say N players in the pool, and no players entering or leaving the pool, then we would witness a systemic inflation if the average (or here, equivalently, sum) of all the ratings weren't constant, but were gradually and systematically increasing as a result of games being played and rated."

So, no, I don't think that it was given by <Bureaucrat>, unless you were paraphrasing him without attribution. But your post above was dated Oct-2009 and the only post by <Bureaucrat> I could found with the word "systemic" was this one in Apr-2012: Hans Arild Runde (kibitz #5825).

No bid deal, memory often fails me also. And your definition is a reasonable one. If I misunderstood, I apologize.

Feb-15-16
Premium Chessgames Member
  AylerKupp: <Sally Simpson> I also have never heard of the Mannerheim/Schuller effect, but that could easily have to do with my ignorance. But, more to the point, a quick web check did not list any references to the Mannerheim/Schuller effect. Please educate me on it.

As far as Alexandru Crisan is concerned, I had never heard of him either. It apparently took 14 years but FIDE supposedly updated its rating information in Aug-2015, lowering Crisan's rating from 2588 in 2001 to 2132 in 2015. I don't know what, if anything, they did to adjust the ratings of all the people he played with the impact that his bogus rating had on their ratings. But I think that you would agree that the effort would have been considerable, given the ripple effect (after the ratings of the people he played had been adjusted, then the ratings of the people those people had played would need to be adjusted, etc.). Given the self-correcting nature of the Elo system, that correction probably would have taken place in the intervening years. So I think that the decision not to correct the ratings of everyone he played and those who played the ones he played was a reasonably pragmatic thing to do. Which doesn't make it right, of course, just understandable from my perspective.

As far as the Elo system not taking into consideration the player's color, that is correct. But, as <frogbert> said, on the average players have an equal number of White and Black games. And the Elo system is only an approximation anyway, and the probability distributions it uses (Gaussian or Logistic) are only appropriate given that a large number of games are considered.

When the number of games is < 30 other probability distributions such as the t-distributions are more appropriate. FWIW I'm currently working on a predictor for the result of individual games in tournaments, and I can tell with confidence that using either Gaussian, Logistic, or t-distributions does not have a significant effect on the results.

Feb-15-16
Premium Chessgames Member
  AylerKupp: <Clemens Scheitz> Dr. Elo in his "The Ratings of Chessplayers, Past and Present" indicated (section 8.93) that he considered it and even acknowledged that "In Swiss parings color assignments may contribute significantly to the outcome and can become critical for the final round." But he then dismisses its implementation because "Any incorporation of colors into the rating system, however, would again inordinately expand the bookkeeping requirements with small prospect of any utility for it, in the final analysis."

Perhaps that was too cavalier of him. But remember that at the time that he developed his system in 1959 – 1960 there was no ready access to computers and all calculations had to be done either by hand or by mechanical calculating machines (also not easily – read costly – available). So the concern about bookkeeping requirements was a valid one at the time. Besides, Dr. Elo was concerned about the validity of ratings over a much longer period of time when the individual color played would indeed average out to approximately 50%. Only when considering a small number of games (as in a specific tournament) does the piece color have any significant influence on the results.

Feb-15-16  Sally Simpson: Hi Aylerkupp.

The Mannerheim/Schuller effect. I made it up on the spot. (sounds like something from a Marx Brothers movie. I sometimes refer to Dr. Hackenbush when stuck for a name. If the Mannerheim/Schuller effect is not a mathematical term then it should be.)

I think I mentioned the impossible task in logistics in correcting all the grades after FIDE finally got around to sorting out the Crisan affair.

Have a wee peek at:

https://en.wikipedia.org/wiki/Cheat...

I recalled reading years ago that Dr. Elo did consider adding in colours and did mention as I did the last rounds of a Swiss event when colour choice is critical.

Yes it is 50-50 anything rare likes Nunn's 8 in a row is a by the by.

But should it be added?

I am sure even I could write a routine to take colours into consideration. In fact it would be very easy.

However I only mentioned it to show a flaw in the system.

Adding in my original suggestion would really falsify things and turn the system into a complete joke. (at the moment it is only half a joke.)

For instance using my added 5 pts. a number I plucked from the air. ( The Mannerheim/Schuller catch) Carlsen by my count won 12 games in classical chess with Black in 2015.

So we add 60 pts to his grade. He is going to keep climbing and climbing and so will everyone else. There will be no way to pull them back.

If Carlsen plays till he is 40 and wins 10 Blacks a year (a very conservative estimate that one) he will have gained something like 800 pts.

If Giri does a Korchnoi he could very easily crack the 50,000 barrier.

And who is going to stop the arguments when you are playing a weak team in a league match. Everyone will be wanting the Black pieces.

So probably best to leave it out or make the Black Bonus something like 0.5 of a point. (Carlsen then gets 6pts for last years efforts.)

or...

Have two grades. One for Black and one for White.....

:)

Feb-15-16  frogbert: <AylerKupp> If you search my kibitzing, you'll find a recent definition of <systemic inflation>, which explains in detail how I calculate it.

The calculations on my player page are done slightly differently, and I came to realize that I had to change the algorithm slightly. This is reflected in the recent description.

I think the main reason I "required" the rating pool to be closed in that old post of mine, was to make it very clear (for the reader) what the concept was. I've never used that as a requirement in my calculations.

I'll see if I can find that recent definition, but you'll probably find it quicker by making a search.

Feb-15-16  schweigzwang: <schweigzwang,s solution to wash out just a one game glitch is for every player to play every other player a 10 game match every month but even then adds "...and the glitch will have mostly been washed out."

All that time and effort to 'mostly remove' a one game glitch. Since 2001 how many games has Gelfand alone played? and those infected with the Gelfand Virus, how many have they played?>

Ha ha, this is <Sally> deliberately playing obtuse. I wouldn't call it trolling, but neither can I explain why he sticks to it so closely.

There's no Gelfand virus.

The ELO system by its nature tends to wash out the glitches, whether or not they are of nefarious origin.

Feb-15-16  nok: <I sometimes refer to Dr. Hackenbush when stuck for a name.>

Close. https://en.wikipedia.org/wiki/Wolfg...

Feb-15-16  Sally Simpson: No Nok,

It's this Doctor Hackenbush.

https://en.wikipedia.org/wiki/A_Day...(film)

There is a scene in it where Chico is selling Groucho books on horse racing form. He has to buy about 16 books because each book is useless without the other 15.

Reminds me of Batsford in the 70's. They published The Modern and The Pirc as two separate books and each book kept referring to the other book for plausible moves. You had to buy both books.

Here:

https://www.youtube.com/watch?v=Dqy...

Feb-15-16  nok: You got me there.
Feb-16-16  tuttifrutty: There is no rating inflation. Elo ratings are accumulated through past perdormances, ie volume of activity. Elo rating is not a commodity that changes value/prices due to calamity or rumors.

Frogbert mumbo jumbo numbers work of art are nothing but the result of smoking banana leaves. Elo rating inflation belongs to the trash bin.

Feb-16-16
Premium Chessgames Member
  AylerKupp: <tuttifrutty> How do <you> define rating inflation? There is not much point on discussing a subject unless there's agreement on what the subject is.
Feb-16-16
Premium Chessgames Member
  plang: An obvious example of rating inflation is when more points are added than are subtracted. In the US there are some examples of rating floors. I believe a Life Masters rating can't fall below 2200 so if you defeat him your rating goes up but his doesn't go down.
Feb-16-16  iking: The <42nd World Chess Olympiad > originally scheduled September 17-30, 2016 has been moved to a new date <September 1-14, 2016>

FIDE@Fide_chess: <Change for the Olympiad and Congress dates http://www.fide.com/component/conte... 2:32 AM - 16 Feb 2016 -https://mobile.twitter.com/Fide_che...

"sinochess@sinachess: <Olympiad in Baku changes dates from 17-30 to 1-14 Sept 2016. Overlap with Chinese League stage 4 & Sinquefield Cup>. China won last Olympiad! 6:01 AM - 16 Feb 2016" - https://mobile.twitter.com/sinaches...

"Mark Crowther@MarkTWIC: <Olympiad in Baku changes dates from 17th-30th September to 1st-14th. September 2016>". 2:53 AM - 16 Feb 2016 - https://mobile.twitter.com/MarkTWIC...

Tarjei J. Svensen@TarjeiJS: <The dates for the Chess Olympiad in Baku changed, now starting Sep 1st instead of Sept 17th.> 2:56 AM - 16 Feb 2016 - https://mobile.twitter.com/TarjeiJS...

This means that the <42nd World Chess Olympiad now scheduled September 1-14, 2016> effectively <overlaps> with the <2016 Sinquefield Cup taking place on August 19-September 2, 2016>.

I bet the USCF together with Mr Sinquefield, prime backer of the 2016 Grand Chess Tour & organizer of the 2016 Sinquefield Cup will adjust the tourney date of Sinquefield Cup earlier by a week or so to possibly come up with a strong team for the chess olympiad with Nakamura, Caruana & So on board. Stay tune for further updates.

<Statement from the 2016 Sinquefield Cup organizer> - SaintLouis ChessClub@CCSCSL @LennartOotes @TarjeiJS <We are aware the Olympiad dates impact the Sinquefield Cup. We are working to have the schedule rectified shortly.> 12:52 PM - 16 Feb 2016 - https://mobile.twitter.com/CCSCSL/s...

- thanks jayZ ..

.

Feb-16-16
Premium Chessgames Member
  AylerKupp: <<Sally Simpson> The Mannerheim/Schuller effect. I made it up on the spot.>

LOL, that's a good one. It reminds me of a time years ago when I was the systems engineering lead on a project. I was conducting a staff meeting and I was telling everyone the projects difficulties could all be traced to the SLF. No one ever asked what SLF stood for. At the end of the meeting when everyone started to leave I said "STOP! How can you sit throughout this meeting and not ask what SLF stands for?" So, when asked, I responded "Software Lunatic Fringe".

And compensating for colors wouldn't take a routine. Looking at the ChessTempo database with 3M+ game results White's P(Win | Draw) is about 55% and Black's P(Win | Draw) is about 45%, give or take a percent. Looking at the FIDE tables a 0.55 P(Win | Draw) corresponds to a 33 to 39 point rating differential, or an average of 36 points. Somewhat more than your plucked from the air 5 points but, if they're Mannerheim/Schuller points (which are clearly worth more), then pretty close. :-)

So, to compensate for color advantage/disadvantage, when calculating the new ratings one could <subtract> 36 points from the White player's rating and <add> 36 points to the Black player's rating. This needs verification, of course, but at least it superficially sounds reasonable to me.

Why <subtract> points from White's rating and <add> points to Black's rating? Well, as an example, have two players, A and B, each rated 2400, play a 30-game match and have the result be, not unsurprisingly, 15 – 15 since the players are equally rated. But suppose that instead of alternating colors for each game as is usually done Player A plays White in each game. Now a 15 – 15 score would mean that Player A performed worse than expected because he had the advantage of the White pieces for each game and yet only managed to tie the match. Similarly Player B performed better than expected because he had the disadvantage of the Black pieces for each game and yet managed to tie the match.

So, in order for the score to be 15 – 15 under those circumstances Player A's rating would need to have been 2400 – 36 = 2364 and Player B's rating would need to have been 2400 + 36 = 2436 because Player A's rating would need to have been lower to compensate for having the White pieces and a 0.05 P(Win) advantage and yet only tie the match. Likewise Player B's rating would need to have been higher to compensate for having the Black pieces and a 0.05 P(Win) disadvantage and still be able to tie the match.

Therefore we should <subtract> the 36 points from the player with the White pieces and <add> 36 points to the players with the Black pieces in order to achieve the same result given the White player's P(Win) advantage of 0.55.

Jump to page #    (enter # from 1 to 67)
search thread:   
< Earlier Kibitzing  · PAGE 65 OF 67 ·  Later Kibitzing>

NOTE: Create an account today to post replies and access other powerful features which are available only to registered users. Becoming a member is free, anonymous, and takes less than 1 minute! If you already have a username, then simply login login under your username now to join the discussion.

Please observe our posting guidelines:

  1. No obscene, racist, sexist, or profane language.
  2. No spamming, advertising, duplicate, or gibberish posts.
  3. No vitriolic or systematic personal attacks against other members.
  4. Nothing in violation of United States law.
  5. No cyberstalking or malicious posting of negative or private information (doxing/doxxing) of members.
  6. No trolling.
  7. The use of "sock puppet" accounts to circumvent disciplinary action taken by moderators, create a false impression of consensus or support, or stage conversations, is prohibited.

Please try to maintain a semblance of civility at all times.

Blow the Whistle

See something that violates our rules? Blow the whistle and inform a moderator.


NOTE: Please keep all discussion on-topic. This forum is for this specific tournament only. To discuss chess or this site in general, visit the Kibitzer's Café.

Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors.
All moderator actions taken are ultimately at the sole discretion of the administration.

Spot an error? Please suggest your correction and help us eliminate database mistakes!

Copyright 2001-2021, Chessgames Services LLC