< Earlier Kibitzing · PAGE 54 OF 54 ·
|Sep-23-17|| ||visayanbraindoctor: <Why do you believe that Ding Liren is underrated?>|
I decided to answer in your forum in order to avoid rating-robots (kibitzers that are obsessed with ratings) butting in with repetitive strawmen and red herring arguments. In brief, Ding has played a lot of games in recent years in Chinese tournaments, which have more lower rated players. I expect he doesn't gain as many points (and occasionally might even lose points) as he would have had he always been playing in Europe; while playing the same quality chess. I believe that if he had been playing purely in Europe for the past few years, his rating would be higher than So's.
My opinion on Elo ratings and their use in the World Championship cycle are summarized in my profile, #8 entry.
<8. On ratings:
Elo ratings reflect relative and not absolute chess strength.
Chessplayers are naturally arranged in populations partitioned by geopolitical regions & time periods that have infrequent contacts with one another. Within such a population, players get to play each other more frequently, thus forming a quasi-equilibrium group wherein individual ratings would tend to equilibrate quickly; but not with outside groups. With caveats & in the proper context, FIDE/Elo ratings are simply fallible descriptors & predictors of an active player's near-past & near-future performances against other rated players, & only within the same quasi-equilibrium group.
As corollaries: the best way to evaluate a player's strength is to analyze his games & not his ratings; one cannot use ratings to accurately compare the quality of play of players from the past and present, or even the same player say a decade ago and today; & care should be taken in the use of ratings as a criterion in choosing which players to seed into the upper levels of the WC cycle. All the above often entail comparisons between players from different quasi-equilibrium groups separated by space and/or time.>
|Sep-23-17|| ||visayanbraindoctor: Nice profile!
<after looking at the data I think that ratings inflation, which I define to be the unwarranted increase in ratings not necessarily accompanied by a corresponding increase in playing strength, is real, but it is a slow process. I refer to this as my "Bottom Feeder" hypothesis>
Thanks for this fascinating hypothesis.
For myself, I believe that the top players of the past, such as <Fischer, Reshevsky, Spassky, Petrosian> whom you mentioned, are as every bit as good as the top players today. Fischer IMO is even better than Carlsen. I base my opinion on a gut-level feeling after studying their games. Fischer for example plays better chess than anyone else today.
Yet Fischer's highest rating was only 2789, only good for # 18. This by itself IMO constitutes an unambiguous evidence that rating inflation exists.
I grew up in the Karpov era. I've followed his games over the years. I am convinced he was a stronger player in the 1970s and early 1980s. It's impossible for him to have a higher rating in the 1990s, if rating inflation did not exist.
I actually believe that Capablanca in his prime would beat even Fischer and Karpov. After studying many of Capa's games, I concluded that I would never believe that a human being could play that many nearly errorless complicated games in real time, if they were not fully documented to have been played by a human being. (For example see my post in Jonathan Sarfati chessforum, and my analysis in many of Capablanca's game pages.) He had no Elo rating.
Yet many of today's rating-robots would ignore these top players of the past just because they never hit 2800.
|Sep-24-17|| ||AylerKupp: <visayanbraindoctor> Thanks for taking the time to answer and for the nice word about my profile. I'm not that familiar with Ding Liren's games and I wasn't even sure that he had had his own page, but I'm glad to see that he does. Although, with only 19 pages of kibitzing, it's clearly not that well known or popular.|
I pretty much agree (and knew) about most of your observations about ratings. One thing I had not thought of in the past is the localization of players and how it influences their rating. It made me think of a recent book I was reading about linear algebra where they were characterizing sparse matrices as to whether they had block regions, regions of the matrix which had a lot of non-zero elements in a few localized and adjacent rows and columns but typically only non-zero elements in the rest of the rows and columns. Apparently there are special algorithms that can more efficiently solve systems of linear equations that have this type of matrix as their coefficients.
The usual problem of not being able to compare the relative strength of players between different eras, at least with the Elo rating system because they belong to different populations is well known. It's too bad that Jeff Sonas abandoned his calculations of Chessmetrics rankings in 2005, although I'm still not sure of its validity to compare the relative strengths (i.e. ratings) of players from different eras and populations.
I'm not sure about the feasibility of determining players' strengths by looking at their games, at least for a lot of players. In my case the obvious problem is my personal inability (because of my low playing strength) to determine the true quality of their play. Many attempts have been made to base this evaluation by comparing these top players' moves with the moves suggested by various computer engines but, in my opinion, all these attempts have been seriously flawed and it's not worth attaching much value to their conclusions.
It's also important to realize that today's top players have more tools at their disposal, mostly computer engines, databases, and tablebases, that earlier players did. And they also have more opportunities to play in tournaments against other top players, and that can't help but improve their game. So they may indeed be better, in the sense of making fewer mistakes, than players from older eras, and their higher ratings just reflect that. I just don't know. But I have no doubt that if the top players from other eras; the Capablancas, Alekhines, Fischers, Spasskys, etc. were somehow transported into the current time and given adequate time and exposure to current chess analysis tools that they would be able to hold their own against today's best players.
You might be interested in downloading the summary spreadsheet from the link in my forum header. It has a lot of charts comparing the ratings of players at different rating levels since 1966. I update it once/year so I will be doing that in Jan-2018 and it will be interesting to see if ratings inflation has indeed plateaud and is heading downwards for all rating levels. I had predicted that, based on trends, that Carlsen's rating would fall below 2800 this year but, although it seemed headed in that direction, it looks like that was premature.
|Sep-30-17|| ||visayanbraindoctor: <it will be interesting to see if ratings inflation has indeed plateaud and is heading downwards for all rating levels>|
This will be interesting indeed. However, I'm not a mathematician, and so I will take your word for it when <ratings inflation has indeed plateaud>
I do know that if the top players confine themselves to playing mostly each other, then they will form a quasi-equilibrium group that will maintain their current high ratings, regardless of the quality of their games or chess strength. But you surely will know how to factor this in your calculations.
<So they may indeed be better, in the sense of making fewer mistakes, than players from older eras, and their higher ratings just reflect that.>
This is the crux of the issue. I used to think this way too. Then Bridgeburner and I made a detailed study of the Lasker - Schlechter World Championship Match (1910)
Since we had to go through their games move for move, as though they were playing in real time, every brilliancy and error of theirs hit us with as much impact as seeing modern GM games being played live in the internet. I soon came to subjectively realize that Lasker and Schlechter were playing their middlegames and endgames more or less as well as modern Champions, and as objectively confirmed by a computer.
I believe that it's their openings that are objectively worse (in the sense of being less accurate and more dubious or <more mistakes> as you say) compared to today's. However, once they got out of the book and into the middlegame, they were every bit as good as today's best players.
<the top players from other eras; the Capablancas, Alekhines, Fischers, Spasskys, etc. were somehow transported into the current time and given adequate time and exposure to current chess analysis tools that they would be able to hold their own against today's best players.>
A quick game genius like Capablanca I believe would win the blitz and rapid championship of the world even without modern opening preparations more than 50% of the time, simply by deploying quiet openings such as QGD, Spanish, or Italian, and then out blitzing his opponent in a more or less equal middlegame.
Carlsen (and Karpov in his heyday) does exactly this stuff even in classical time controls. It's ironic but it seems to be Carlsen fans that keep on claiming that Carlsen plays better because he was born in the computer age, when among top players, he is the one most likely to play 1920s 'classical' openings and eschew sharp computer opening lines. (He rarely plays Indian openings and asymmetrical openings such as the Sicilian. Faced with the Sicilian himself, he opts to steer it into 'closed' variations, as he did in his matches with Anand. Carlsen is extremely 'classical' in his approach to openings, preferring to directly occupy the center with his pawns, rather than control it indirectly by fianchettoes or by counterpunching asymmetrical openings. The way he plays his openings is similar to a 1920s master.)
Regarding Capablanca, if he were transported to the modern era, he would probably play his openings exactly like Carlsens'. Get right into a 'safe' semiclosed ot closed middlegame, and then hope to outplay his opponent. Capa would probably be just as successful too, and perhaps more so as I have reason to believe that he is a better tactician than Carlsen.
Alekhine on the other hand would prepare the sharpest of openings, and he would be overjoyed to have computers assist him. Alekhine from all accounts had an eidetic chess memory. It would be no problem for him to update himself in the sharpest of opening variations in short order from a laptop. The chess world would soon see him blasting his opponents off the board with non stop sacs and brillancies, exactly like Kasparov, live in the internet.
There is another thing that I've noticed. The stronger individual kibitzers are, the more they tend to think that rating inflation exists (but not all of course). It's mostly the (pardon the expression) patzers in CG.com that tend to think that ratings reflect absolute chess strength and totally deny any form of rating inflation. They can go through Carlsen vs Bu Xiangzhi, 2017 and Capablanca vs Marshall, 1918, and fail to realize that Capablanca was defending the position and handling its tactics (in a similar situation) better than Carlsen, on the assumption that since Capablanca had no Elo rating, he would not be able to do such a thing better than Carlsen.
It's like for them, chess has been reduced to ratings. When they see two chess players play a game, they see only how the players' respective ratings can change and fail to see the game itself.
|Oct-17-17|| ||Octavia: < The more you reply to him, the more trash he'll post.|
I'm not sure about that. I stopped responding to him for a while (i.e. taking the bait) and it didn't seem to reduce his posting volume.> Of course, he'll keep on posting hoping for some others to answer him. If nobody answered he'd stop eventually.
You don't need to worry about others believing him. What does it matter?
|Nov-01-17|| ||takchess: AK, your notes on computing reminded me of Complexity Theory; |
In the 1950's and 1960's, American meteorologist Edward Lorenz found that small rounding errors in his computer data (which has a limited number of significant figures) leads to large non-linear instabilities that expand exponentially in time and make long-term prediction impossible. This is the famous "Butterfly wings in Beijing" effect discovered in weather predictions.
found at the link above
|Nov-02-17|| ||Boomie: <I'm not sure about the feasibility of determining players' strengths by looking at their games.>|
Computers can measure the tactical strengths of players only. They are oblivious to psychology, aesthetics, and other human factors that raise the game to the level of an art.
One measure which hasn't been mentioned here is the opinion of world champions and other strong players. For example, Capa, who was not effusive in his praise of other players, said he was flattered to be considered as talented as Morphy. Fischer worshipped Morphy and Botvinnik praised him. I suggest that their opinions carry more weight than pages of computer screed. They all knew that Morphy would be a formidable opponent in their times. Plus I'd wager that they would all love the opportunity to play him.
|Nov-02-17|| ||takchess: Aagaard in his Attacking Manual 1 and 2 has some interesting views on chess computer analysis. Worth checking out.|
|Dec-09-17|| ||kwid: <AylerKupp:>
As a member of team black in the Traxler 5.Bc4 challenge game I am disappointed from your expressed wishes to withdraw.|
Please reconsider since your already made contributions to the opening theory will be of historical value.
Since we are only in an early stage of the opening it would really be interesting to see the conclusion, especially if team white has worked out the refutation of the Traxler opening.
This game could be of great theoretical value if you would help us to play the best possible variations for black to counter RV's contributions for the white side.
|Dec-11-17|| ||AylerKupp: <kwid> Sorry, but I can't reconsider. It was not a snap decision, I had been thinking about it for a while because of all the other things that I need to do. I had misgivings about joining the game from the beginning because of the known demands on my time but I decided to give it a try anyway. Unfortunately it got to the point where I didn't think I could do a good enough job and it was very frustrating, I wasn't enjoying it. And, if you're not going to enjoy the game, what's the point? So, if I was going to withdraw, I thought that it would be better for the team that I do it now rather than later.|
Anyway, thanks for the kind words from you and other team members. And best of luck during the rest of the game.
|Feb-12-18|| ||AylerKupp: <<tpstar> Re: Wesley So (kibitz #214349)>|
How typical. You can't dispute the assertions I made to your last post (Tata Steel (2018) (kibitz #1174)) so you withdraw into the Chessgames Home Page page cocoon hoping that I and others won't follow. Well, as you can see, that won't work.
There definitely IS a controversy around the So forum and it's growing larger every day. As far as "witnessing a bunch of sore loser crybabies who are dying to post here, but they can't, because they are anti-Wesley." that's certainly not true. These '"crybabies" can't post in the Chessgames Home Page pages because they have been banned by its webmaster based on the "suggestions" of users like you, not because they are anti-So. Then again, you think that because most users are not 110% fanatical So worshipers they are automatically anti-So. You will apparently never learn the difference.
As far as I'm concerned I've told you that I don't care one way or another whether I post on the Chessgames Home Page page. The number of times I've done that is small, and you can easily verify that. I doubt that too many users will lose any sleep about being banned from posting on the Chessgames Home Page page.
If Wesley So did not want quarrels spoiling his player page then perhaps he should not be visiting the page. At any rate, "spoilage" is often in the eye of the beholder.
As far as asking chess fans to stay out of his personal business, that's a reasonable request. But it's only a request. He is a public figure by voluntarily participating in the public arena and, as US courts have noted (https://www.rcfp.org/federal-foia-a... ), a public figure has significantly diminished privacy interest than others." Therefore, if he does not wish people to dig into his personal business, he should not have decided to become a public figure.
I for one would also like for <BatangaLista> to periodically update the list of so-called anti-Wesley posters in order to facilitate also banning them from posting on the Chessgames Home Page page. Then I can create a spreadsheet to show how the number of banned posters increases over time. I'm sure that <chessgames.com> would also be interested in this information.
|Feb-12-18|| ||tpstar: <AylerKupp> It doesn't matter what I think. It matters what Wesley thinks, and he has always been touchy about his player page. Anyone who has read that page from the top understands this - Dot Dot Dot - and anyone who remembers when he left this site before understands this - Dot Dot Dot. |
Since moving to the U.S., Wesley had a rocky split with Webster, a rocky split with his parents, and a rocky split with the Barangay Wesley. Without dredging it all back up, he said some mean things about Norlito, and vice versa, then we all picked sides. Except you can't spot Norlito, and you can't spot Joselito, and you can't spot Francis, and you can't spot Glenn. They just now tried to sneak back in using the former handle of a dead person, which rightfully got banished during the latest scuffle. For the fourth time, this "fanatical contingent" you keep referencing is anti-Wesley. Blaming pro-Wesley supporters for anti-Wesley antics is a false narrative.
Read Tata Steel (2018) from the top and then try to tell me, "There is NO anti-So contingent" here. This event ended two weeks ago and has fully degenerated into a pity party bitch session by sore loser crybabies who cannot post on the Wesley So page because they are repeat offenders. Moreover, even a child could notice the blatant antagonism toward Wesley and his pro-Wesley supporters. When you burst onto his page in August 2017 and declared your opinion two years after the Great Banishment of May 2015, well, you couldn't have been more wrong. You also couldn't have been more wrong here:
<This "ever present malice" is not very real and it is mostly a figment of yours and very few others' imagination who think that anyone that is not (fanatically) pro-So is automatically (fanatically) anti-So.>
Perhaps you saw your intervention as enlightened verbosity, but apparently Team Wesley perceived it as quarreling.
I warned you to drop it, and I was right. The next person will take the hint. Meanwhile, good luck getting unbanished.
|Feb-12-18|| ||AylerKupp: <tpstar> You still don't get it. You keep trying to lump together the anti-So antagonisms and the anti-So<bot> antagonisms and they are not the same. There is no substantial anti-So antagonism; although there is beginning to be a substantial anti-So-contingent antagonism and it is increasing as a result of your silly and extreme antics.|
As far as http://www.chessgames.com/perl/ches... there is NO anti-So contingent here. There IS an anti-<So>bot antagonism but, since you are not willing to accept that they are not the same you think that any antagonism is a reflection on So himself and that is just not the case.
And as far as blaming pro-Wesley supporters for anti-Wesley antics I have done no such thing. I have said that <some> of these pro-Wesley "supporters" reflect poorly on him, although they delude themselves into thinking that their extreme annoying behavior actually serves him. They could not be any more wrong and have no one but themselves to blame.
I did not see any supposed "intervention" on my part as enlightened in any way, just verbose as usual. If Team Wesley perceived it as quarreling they couldn't be more wrong nor could I care less.
As far as my getting unbanished from the Wesley So page I also couldn't care less. Even if I were unbanished I wouldn't bother posting anything there. Why bother doing so when you all have a closed mind and are unwilling to tolerate any other views but your own, even if they are expressed in an objective and non-derogatory way? You extreme Wesley So fanatics just want to live in your fictional world and tolerate no discussion, just fawning and worshiping. Who in their right mind wants to be a part of that?
|Feb-13-18|| ||tuttifrutty: You forgot to say " You tell me" at the end of your question. :-)|
|Feb-14-18|| ||AylerKupp: <<tuttifrutty> You forgot to say " You tell me" at the end of your question. :-)>|
You're absolutely right, sorry about that. In fact, in my ever futile attempts to be less verbose, I think that I'll start abbreviating that as "YTM". Unless, of course, you register the phrase as a trademark. I think that some day this acronym will be as famous and ubiquitous as LOL, IMO, and BTW. And don't worry, if anyone ever asks me what YTM stands for, don't worry, I'll give you full credit. I already have many things to be infamous for, and I'm more than willing to share.
|Mar-03-18|| ||AylerKupp: <2018 Candidates Tournament Simulation> (part 1 of 4)|
<Lambda> Below is my concept of how a tournament simulation might be implemented in Excel and how it could be used to determine each player's tournament winning probabilities. I would greatly appreciate if you could take a look at it, tell me if it seems like a reasonable approach, and let me know if it's in any way similar to what you've done.
a. Assumed Draw percentage = 4/7 or p(Draw) ~ 0.571429
b. Assumed White advantage = 35 Elo points
c. White p(Win or Draw) with no Elo rating point advantage = 0.500000
d. White p(Win or Draw) with 35 Elo rating point advantage = 0.549241
3. White's p(Win or Draw ) advantage = 0.549241- 0.500000=0.049241
Note: I display 6 digits (rounded) in the p(Win or Draw) calculations because that's the number of digits needed to ensure that each p(Win or Draw) is unique for each 1 point Elo rating difference. After rounding to 6 decimal places this allows me to do a table lookup in Excel to get the p(Win or Draw) based on the rating differential, which is faster than a search. But all the calculations are made using Excel's internal 15-digit precision.
And, while I hopefully have your attention, perhaps you could answer a few more questions:
1. What language / package did you use to implement your simulation?
2. How long does it take to run 1 million simulations?
3. Did you ever determine how many simulations needed to be run in order for the results to be statistically valid? I typically use the criteria that a result is statistically valid if there is less than a 0.05 probability that the result was due to chance.
|Mar-03-18|| ||AylerKupp: <2018 Candidates Tournament Simulation> (part 2 of 4)|
<1. Calculate p(Win), p(Loss) >
For each player / player and White / Black combination, calculate the p(Win) and p(Loss) for the White player based on their rating differences per the Mar-2018 FIDE rating list. For an 8-player double round robin there will be 2 * 8 * 7 = 112 combinations.
<Example> Mamedyarov (highest rated player in the tournament) vs. Karjakin (lowest rated player). First let's consider Mamedyarov playing White:
Mamedyarov's Pre-tournament rating = 2809
Karjakin's pre-tournament rating = 2763
Rating difference (RDiff) = 46 Elo rating points
Ignoring any White advantage:
a. Mamedyarov's [ p(Win or Draw | RDiff = 46 ] = 0.564597
b. Mamedyarov's [ p(Loss or Draw | RDiff = 46 ] = 1 - 0.564597 = 0.435403
Since White's advantage is assumed to be 35 rating points and this corresponds to a 0.049241 difference in White's p(Win or Draw), I add 1/2 this amount to White's p(Win or Draw) and subtract 1/2 this amount from White's p(Loss or Draw) to get:
a. Mamedyarov's p(Win or Draw | RDiff = 46 | Mamedyarov playing White) = 0.589217
b. Mamedyarov's p(Loss or Draw | RDiff = 46 | Mamedyarov playing White) = 0.410783
Since p(Draw) = 0.571429, subtract 1/2 of this amount from Mamedyarov's p(Win or Draw) and p(Loss or Draw) so:
a. Mamedyarov's [ p(Win) | RDiff = 46 | Mamedyarov playing White ] = 0.589217 - 0.571429 / 2 = 0.303503
b. Mamedyarov's [ p(Loss) | RDiff = 46 | Mamedyarov playing White ] = 0.410783 - 0.571429 / 2 = 0.125069
As a check, p(Win) +54:54p(Draw) +54:54P)Loss) must = 1. Therefore, given that Mamedyarov is playing White and his rating advantage over Karjakin is 46 rating points:
Mamedyarov's p(Win) + p(Draw) + p(Loss) = 0.303503 + 0.571429 + 0.125069 = 1.000000
Now, let's assume that Karjakin is playing White. Then RDiff = -46 Elo rating points, Then, ignoring any White advantage:
a. Karjakin's [ p(Win or Draw | RDiff = -46 ] = 0.435403
b. Karjakin's [ p(Loss or Draw | RDiff = -46 ] = 1 - 0.435403 = 0.564596913
a. Karjakin's p(Win or Draw | RDiff = --46 | Karjakin playing White) = 0.460024
b. Karjakin's p(Loss or Draw | RDiff = -46 | Karjakin playing White) = 0.539976
c. Karjakin's [ p(Win) | RDiff = -46 | Karjakin playing White ] = 0.460024 - 0.571429 / 2 = 0.1743095
d. Karjakin's [ p(Loss) | RDiff = -46 | Karjakin playing White ] = 0.539976 - 0.571429 / 2 = 0.2542615
Again, as a check, given that Karjakin is playing White and his rating advantage over Karjakin is 46 rating points:
Karjakin's p(Win) + p(Draw) + p(Loss) = 0.174310 + 0.571429 + 0.254262 = 1.000000
|Mar-03-18|| ||AylerKupp: <2018 Candidates Tournament Simulation> (part 3 of 4)|
<2. Calculate Win, Draw, and Loss ranges>
For each player / player, White / Black combination also calculate 3 sets of values:
a. Win Range = [ 0 , p(Win) ]
b. Draw Range = [ p(Win) , p(Win) + p(Draw) ]
c. Loss Range = [ p(Win) + p(Draw) , 1 ]
<Example> For Mamedyarov vs. Karjakin, Mamedyarov playing White:
a. Win Range = [ 0.000000 , 0.303503 ]
b. Draw Range = [ 0.303503 , 0.874932 ]
c. Loss Range = [ 0.874932 , 1.000000 ]
And for Mamedyarov vs. Karjakin, Karjakin playing White:
a. Win Range = [ 0.000000 , 0.174310 ]
b. Draw Range = [ 0.174310 , 0.745739 ]
c. Loss Range = [ 0.745739 , 1.000000 ]
|Mar-03-18|| ||AylerKupp: <2018 Candidates Tournament Simulation> (part 4 of 4)|
<3. Determine the winner of one tournament by simulation>
For each player / player, White / Black combination, calculate a random number (RN) between 0 and 1. Score each game as follows:
a. White player wins if RN <= Win Range(2)
b. White player draws if RN > Draw Range(1), RN <= Draw Range(2)
c. White player loses if RN > Loss Range(1) (i.e. otherwise)
<Example> For Mamedyarov vs. Karjakin, Mamedyarov playing White:
a. A win for Mamedyarov if the value is <= to 0.303503
b. A draw for Mamedyarov if the value is > 0.303503, < 0.874932)
c. A loss for Mamedyarov if the value is > 0.874932
And for Mamedyarov vs. Karjakin, Karjakin playing White:
a. A win for Karjakin if the value is <= 0.174310
b. A draw for Karjakin if the value is > 0.174310, <=0.428571
c. A loss for Karjakin if the value is > 0.428571
The winner of the tournament, of course, is the player with the highest score. If two or more players have the same score, consider that each player won the tournament.
<4. Determine the tournament win probabilities>
Run a simulation of as many tournaments as desired, or as required to establish statistical significance. Determine for each players their p(Tournament Win) as the ratio between the number of tournament wins by that player and the total number of tournament wins by all players. Note that the total number of tournament wins by all players will likely be greater than the number of simulations run because more than one player might tie for first place in any simulated tournament.
Hopefully all this makes sense to you.
|Mar-03-18|| ||AylerKupp: <Lambda> I don't know how you arrived at your assumptions of a draw percentage = 4/7 ~ 57.14% or your assumed White advantage = 35 Elo points. But in case you're interested here's some data that validates your assumptions using the ChessTempo database.|
The ChessTempo database (https://chesstempo.com/game-databas...) currently contains over 1.7 million chess games of all types (Classic time control, Blitz, rapid, etc.). One thing that makes it useful is that you can easily filter it to consider only games where both players were rated 2200+, 2300+, ..., 2700+. The latter is particularly useful for determining % Win, % Draw, and % Loss for the White player in a super strong tournament like the 2018 Candidates.
The last time I looked at the database in detail was in May-2017. At that time it had 14,502 games of all types where both players were rated 2700+. I filtered the database to exclude games played in events that had Blitz, Rapid, Exhib(ition), Blind(fold), and Simul(taneous) in their title, and the remaining 8,879 games I assumed to have been played at Classic time controls. Not perfect, but probably close.
Of these 8,879 games all were played since 2000 so they are probably all relevant. White won 26.23% of the games, lost 15.73% of the games, and 58.04% of the games were drawn. Clearly the 58.04% is very close to your 57.14% assumption for a draw percentage.
The percentage of White wins + the percentage of Black wins was 41.96%. Calculating the White advantage as White Win % - 1/2 * (White + Black win % ) = 26.33% - 20.92% = 5.25% or a White p(Win) of 0.552. This corresponds to a rating differential of +37 Elo rating points. Again, very close to your assumed +35 Elo rating points.
Just thought that you might be interested to know.
|Mar-04-18|| ||Lambda: My simulation tool is written in Python, and it takes slightly over a minute to run a million simulations. (Less time once the tournament starts and some of the game results are already determined.)|
I haven't attempted to define "statistical validity", but the results from a million trials don't tend to change from run to run by more than 0.1%.
I have no insights about what a good way to use Excel to do this because I've never used Excel in my life, and indeed I've never willingly used any tool from any "office suite" in my life. Markup languages for text formatting, and programming languages for data processing is my attitude.
But other than what you need to do to work around the limitations of your tool, that sounds about right to me. I haven't checked your details, but in approach, the only obvious differences I have are that:
My "draw area" is the first four sevenths of my random number, so I can generate the random number, immediately check whether it's a draw, and if it is, I don't have to do any further calculations for that game, for efficiency, and
At the end, I check for ties and try applying all the tie-breaks to my cross table to get the one winner, and if they're all tied too, call a tie a ninth result.
|Mar-04-18|| ||AylerKupp: <Lambda> Thanks for responding. I use Excel mainly for entering the parameters (players, ratings, draw %, etc.) so that they are easily visible and changeable. The bulk of the work I do using Visual Basic for Applications (VBE) mainly because I'm familiar with it, and I start the simulation by invoking a macro. I've been trying to learn Python off and on for years because I find it elegant, but I've never gotten it to work on my computer for unknown reasons. Then again I haven't tried very hard.|
I'm encouraged that no only do you think (at least at first glance) that my concept is reasonable (or, at least, not unreasonable) and that it only takes a little over a minute to run a million simulations. Since both Python and VBE are interpreted, VBE should be able to run in the ballpark, although I'm sure that Python is much more efficient. After all, VBE is a Microsoft product.
The reason I mentioned statistical significance testing was that, if it took 1 million simulations a long time to run, then the number of simulations probably could have been reduced to save time. But, if it only takes slightly over a minute (or two, or five), then it's not an issue.
Yes, for efficiency, I was going to check for a draw first since that's the statistically most likely result; it's just second nature to me after many years of writing real-time software. But since that was not relevant to the concept, I didn't bother to mention it. At any rate, given the small amount of time that it takes to run 1 million simulations, this type of efficiency is probably not significant.
I had not considered using the tie-breaks to get the one winner; thanks for the tip. It makes more sense. But, reviewing the tie-break rules for this tournament, after the first 3 tie-breakers the players need to first play 2 rapid games, then up to 4 blitz games, and then a sudden death game. To be more "accurate", I would think that additional simulations would have to be run with the player's rapid and blitz ratings used, and I doubt that there are any ratings for sudden death games. So that's more work. Then again, the likelihood that there would be a tie after using Sonnenborn-Berger is miniscule, so bothering to implement rapid, blitz, and sudden death tie-break simulations is probably not worth the effort.
FWIW, I like the tie-break sequence in the Candidates and I wish more tournaments would use it. It encourages trying for wins just in case and it's based primarily on the results obtained by the players in the actual tournament based on games at classic time controls. So all the arguments for and against using rapid and blitz time controls to determine the results of a tournament played at classic time controls can be avoided for the most part.
|Mar-31-18|| ||qqdos: <Dear AK> would you like to take a quick look at the invitation at Bobby's flawed Gem vs Geller [B89]. kind regards.|
|May-11-18|| ||yskid: I've just posted on "Naiditch game" site ;
12.a3 line played in the correspondence championship game
|May-18-18|| ||djvanscoy: <AylerKupp> "It made me think of a recent book I was reading about linear algebra where they were characterizing sparse matrices as to whether they had block regions, regions of the matrix which had a lot of non-zero elements in a few localized and adjacent rows and columns but typically only non-zero elements in the rest of the rows and columns."|
I'm guessing you meant to say, "...typically only zero entries in the rest of the rows and columns"? In other words, some block is dense but the rest of the matrix is sparse?
"But I have no doubt that if the top players from other eras; the Capablancas, Alekhines, Fischers, Spasskys, etc. were somehow transported into the current time and given adequate time and exposure to current chess analysis tools that they would be able to hold their own against today's best players."
I agree with you, and indeed I couldn't help but think that Carlsen's rook-and-pawn endgame blunder on move 54 of his game against Caruana in the first round of the 2018 GRENKE tournament (Caruana vs Carlsen, 2018) would not have been made by Capablanca. But maybe in this case I'm afflicted with a bit of hero-worship.
< Earlier Kibitzing · PAGE 54 OF 54 ·
from the Chessgames Store