Stylistic bias in computer chess programs

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: Stylistic bias in computer chess programs

Post by Don »

JuLieN wrote:Well, some very typically styled players are:

- Petrossian (defense!)
- Shirov (attack!)
- Nimzovitch (logical but unnatural)
- Capablanca (natural talent: only smooth moves)
- Alekhine (ancient attacking style, very sharp but still positional sound)
- Kasparov (like Alekhine, but even more positional).

And I don't think players like Anand or Carlsen have so personal style... But Fischer and Karpov do.

Also, the measurement should only take place in the middle game, not the opening nor the endings (too technical, no place for style).
The effect of the openings are somewhat minimized by the fact that each position must be unique for that player so very little weight would be given for opening moves. However, I still do not include moves that are almost surely in book, my arbitrary cutoff is the first 8 moves or 16 ply are not considered. I am not quite as sure as you about the endings, but I also cut off games beyond move 60.

I can only include players if I can come up with a pretty large sample of games, but your list seems pretty good, I had planned on including games of Kasparov already and it would not seem right not to include Capablanca and some of the others on your list.

Don
User avatar
JuLieN
Posts: 2949
Joined: Mon May 05, 2008 12:16 pm
Location: Bordeaux (France)
Full name: Julien Marcel

Re: Stylistic bias in computer chess programs

Post by JuLieN »

Other players with very personal style that would be interesting to test:

- Pillsbury (self-taught, his style is the result of a theorization of the game more than from natural talent, so he sometimes committed awful blunders)
- Reti (deep natural talent + theorization, interesting mix!)
- Morphy (natural, logical player. Problem is, Morphy's level today would be around 2400, so we're back to the first posts of this thread.)
"The only good bug is a dead bug." (Don Dailey)
[Blog: http://tinyurl.com/predateur ] [Facebook: http://tinyurl.com/fbpredateur ] [MacEngines: http://tinyurl.com/macengines ]
Uri Blass
Posts: 10269
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Stylistic bias in computer chess programs

Post by Uri Blass »

JuLieN wrote:Other players with very personal style that would be interesting to test:

- Pillsbury (self-taught, his style is the result of a theorization of the game more than from natural talent, so he sometimes committed awful blunders)
- Reti (deep natural talent + theorization, interesting mix!)
- Morphy (natural, logical player. Problem is, Morphy's level today would be around 2400, so we're back to the first posts of this thread.)
I disagree with you about Pillsbury

http://en.wikipedia.org/wiki/Harry_Nelson_Pillsbury


I think it means that pillsbury probably had more talent for chess relative to other players.

I do not believe that it is possible or was possible to become a top player
without having a big natural talent.

I did not look at the games of pillsbury but I read that
pillsbury suffered from serious health problems and I guess this was the reason for the blunders that you find.

You also say:
"Problem is, Morphy's level today would be around 2400"

I also disagree here because I do not believe that Morphy was weaker than other players of the past that came after him like Reti(Maybe all of them were not better than 2400 but in this case I see no reason to mention it only about Morphy)
User avatar
JuLieN
Posts: 2949
Joined: Mon May 05, 2008 12:16 pm
Location: Bordeaux (France)
Full name: Julien Marcel

Re: Stylistic bias in computer chess programs

Post by JuLieN »

EDIT: MY MISTAKE, the player that made huge progresses by studying chess at home was Rubinstein, not Pillsbury, sorry :)
(So all I said about Pillsbury was actually about Rubinstein)
Last edited by JuLieN on Sat Dec 10, 2011 8:15 pm, edited 1 time in total.
"The only good bug is a dead bug." (Don Dailey)
[Blog: http://tinyurl.com/predateur ] [Facebook: http://tinyurl.com/fbpredateur ] [MacEngines: http://tinyurl.com/macengines ]
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: Stylistic bias in computer chess programs

Post by Don »

We may learn something from this study about the playing style of the players. It's my opinion there is some truth in the popular wisdom and anecdotes of various players but the top players are also romanticized a lot, and I suspect that much of what we hear is distorted. Sometimes we put a lot of weight on one or two well publicized incidents, a particular game or two where some player fought his way out of what seemed a certain loss, or other such happening that defines the reputation of a player, even to the point where the player himself starts believe it.

I don't know if this will happen, but we may be able to get a computerized view of the similarity of individual players by looking at how the programs rank them according to their own style.


Uri Blass wrote:
JuLieN wrote:Other players with very personal style that would be interesting to test:

- Pillsbury (self-taught, his style is the result of a theorization of the game more than from natural talent, so he sometimes committed awful blunders)
- Reti (deep natural talent + theorization, interesting mix!)
- Morphy (natural, logical player. Problem is, Morphy's level today would be around 2400, so we're back to the first posts of this thread.)
I disagree with you about Pillsbury

http://en.wikipedia.org/wiki/Harry_Nelson_Pillsbury


I think it means that pillsbury probably had more talent for chess relative to other players.

I do not believe that it is possible or was possible to become a top player
without having a big natural talent.

I did not look at the games of pillsbury but I read that
pillsbury suffered from serious health problems and I guess this was the reason for the blunders that you find.

You also say:
"Problem is, Morphy's level today would be around 2400"

I also disagree here because I do not believe that Morphy was weaker than other players of the past that came after him like Reti(Maybe all of them were not better than 2400 but in this case I see no reason to mention it only about Morphy)
Dan Andersson
Posts: 442
Joined: Wed Mar 08, 2006 8:54 pm

Re: Stylistic bias in computer chess programs

Post by Dan Andersson »

Akiba Rubinstein, in many ways the first truly modern player. Pawn structures he understood. He was arguably the first player to take the types of endgames a certain opening produced in to account. His opening preparation was legendary and there are many, many Rubinstein variations in most modern openings. He played endgames extremely well and is still considered one of the five best endgame players by many. He faded away fro numerous reasons. Lack of financial backing to mount a WC challenge, mental and bodily sickness, traumatic war experiences and the rise of J R Capablanca.
User avatar
JuLieN
Posts: 2949
Joined: Mon May 05, 2008 12:16 pm
Location: Bordeaux (France)
Full name: Julien Marcel

Re: Stylistic bias in computer chess programs

Post by JuLieN »

Dan Andersson wrote:Akiba Rubinstein, in many ways the first truly modern player. Pawn structures he understood. He was arguably the first player to take the types of endgames a certain opening produced in to account. His opening preparation was legendary and there are many, many Rubinstein variations in most modern openings. He played endgames extremely well and is still considered one of the five best endgame players by many. He faded away fro numerous reasons. Lack of financial backing to mount a WC challenge, mental and bodily sickness, traumatic war experiences and the rise of J R Capablanca.
Just to be clear, and prevent people from jumping on me because of Rubinstein, I said what I said because one of my favorite book, when I was a teen club player, was french IM Nicolas Giffard's Chess Bible ("Le Guide des Echecs"), especially because of the detailed biographies of famous players. So, there you go, I'll translate the 3 pages (pp 406-409 of my 1992 edition) for you all. That's a big work, but hopefully you'll find it interesting ;) It's an insertion in the chapter about Lasker.

Purity made man

After his frightening fight against Schlechter, Lasker wasn't in a hurry to put his title in the stakes again. The world chess had to wait ten years before a new world championship could take place. Nevertheless, a candidate was knocking at the door and had Akiba Rubinstein as a name. This little polish jew, just like Steinitz, was to become a rabbi. But he discovered chess when he was 16 and never got interested in anything else during the rest of his life.

Réti tells about his first youth achievement, that seems nearly unbelievable. Having left his birth town, Stawiski, Rubinstein was passing most of his time at the Lodz's chess club, despite he had come to this town to study the Talmud. The club's champion was the master George Salwe (1862-1920) and Rubinstein, as the other amateurs, used to play against him with being a rook up. Rubinstein stayed like that during several months without making any noticeable progress, apparently not any more talented than Salwe's other victims. Came a moment when nobody saw him anymore at the club. Two months later, Rubinstein came back, went directly to Salwe and defied him in a match, this time without any material handicap. And the unthinkable happened. Salwe lost one, two, three games. Thanks to a struggling personal training, the young Rubinstein had trained himself into a master.

This occurrence is unique in the whole history of chess. All the greatest champions got their progresses through the mean of direct confrontations with superior opponents, not through a fold over themselves.

Encouraged by this success, Rubinstein put himself back to work, giving himself an ascetic life style, discarding anything that could disturb his progress. The international beginnings of Rubinstein took place in Kiev, in 1903. He ended up at the 5th place. Disappointed, Rubinstein went back to his "laboratory". His come back, sensational, baffled the world. The polish man won in a row the tournaments of Ostende (ex aequo with Bernstein) and, most of all, Carsbad's in 1907 where he finished before Maroczy, Schlechter, Tchigorine and Janowski, amongst others.

The traveling professional players of those days were stupefied by this quiet man, leaving the tournament hall only to go directly to his hotel room, where he would eat his meals. Rubinstein had no other thoughts than those devoted to chess.

The main characteristic of his game was the search for the absolute perfection, for purity. Rubinstein is the only player in the world that left so many exemplary games, reproduced in many chess manuals as examples of clarity in how some positions have to be played. He would only play a move if he was certain that this move was not only satisfying but also that it would be fitting the logics of the position, and, facing even the smallest defect, he would be searching for another move, more fitting to the position. From a sportsman point of view, this way of playing had the inconvenience of being extremely tiring and had for main consequence the enormous amount of blunders that plagued Rubinstein's games during his whole life.

Akiba Rubinstein owns a record. He won, throughout 1912, 4 tournaments in a row. But that's most of all how he did it that was impressive. His opponents were under the devastating impression of facing the God of chess himself, with all his perfect play. His rare defeats weren't caused by a superiority of his opponents, but by a momentary weak time, during which he would make a beginner's blunder.

Rubinstein's pure, spare style was especially efficient during the endgames. When the queens were gone, Rubinstein was even stronger, and the best players lived in fear of facing him during an endgame. When most of the masters would agree for a draw, Rubinstein enjoyed taking the time to deepen an apparently equal position, continued to play it to the end and, most of the time, would won the game.

Rubinstein remains famous for his rooks endgames, for which he remains one of the best specialists ever. Here's a typical example played during San-Sebastian's tournament in 1912:

Schlechter-Rubinstein, San-Sebastian, 1912 (black to play):
[d]8/p3kp2/1pr3p1/7p/8/2P1P3/PP5P/1K4R1 b - - 0 1
Rubinstein has a small advantage, because white has a defect in their game: the e3 pawn. His technics will work magics:
1. ..., Re6; First thing first, putting white in a defensive situation.
2. Re1, Rf6! ; Menacing Rf2
3. Re2, Ke6
4. Kc2, Ke5 ; centralizing the king
5. c4, Ke4
6. b4 ; Schlechter is playing his only card: a potential passed pawn on queen side
6. ..., g5!
7. Kc3, g4
8. c5, h4
9. Rg2, Rg6
10. Kc4, g3
11. hxg3, hxg3
12. Kb5, Kf3
Each side has a passed pawn. With one twist: whereas the black king is about to get his passed pawn's blocker to move, the white king will soon have to go back.
13. Rg1, bxc5
14. bxc5, a6+!
15. resign
Schlechter is about to lose his rook.

Despite being quiet, Rubinstein was aware of the impression he made to his contemporaries and was working on his fearsome image, as the following anecdote tells. Before the last round of a tournament, he was a full point ahead and had to face a player from the middle of the rankings. A draw was enough for him to end up alone at the first place. At the tenth move, his opponent offered him the draw. But to everyone's amazement, Rubinstein refused. The game went on and soon Rubinstein acquired an enormous superiority. What a ravishing finish, were thinking the spectators. but before giving his opponent the coup de grâce, Rubinstein offered him a draw that he, of course, was quick to accept. Of course, everyone wanted to understand why the Polish player had done that. He explained, all the more seriously, that he, Akiba Rubinstein, was the only one who chose when, and with who, he would share a point.

That's about when the first signs of mental trouble appeared. Rubinstein had a strange habit: he would went up and would seek refuge in a corner of the tournament hall. Nothing too strange up to now, but then one would see him approaching his hands toward his face and one could ear a small whispering and see him smile to his own stories.

He would furthermore complain about a fly constantly turning around his head and preventing him from focusing, sleeping, etc.

The first world war seriously impacted Rubinstein who met awful times, was deprived of food and came back to the chess board in a very bad physical and mental health.

He played until 1930 but could never achieve again the level he had before the war.He had kept his perfectionist style, but sadly the blunders were now too many and would often came to waste the endgame of an up to then well played game.

Just like Pillsbury, Rubinstein was certainly deserving a world championship match against Lasker. Some even says that Lasker would ignore any question related to the Polish man... Whatever, Rubinstein put an end to his chess player career in 1932 and went to live in Brussel where he remained until the end of his life. We know about nothing of his late years except that he seemed to have recover some sanity as soon as he had put an end to his colossal training in chess.

He was forgotten by the world of chess and died the year the young Bobby Fischer became champion of the USA, in 1961.

As a tribute to Rubinstein, here's a game he played in Hastings in 1922, characteristic of the brilliant moments he had after the war:

Rubinstein-Tarrasch, Hastings, 1922, Dutch Defence:

1. d4 ; Rubinstein was a sucker for the queen game
1. ..., e6; 2. c4, f5; 3. g3, c5;
Tarrasch himself considered this move to be bad.
4. Nf3, cxd4; 5. Nxd4, Nf6; 6. Bg2, Nc6; 7. O-O, Bc5; 8. e3!, O-O; 9. Nc3, a6; 10. a3, Qc7; 11. b4, Be7; 12. Bb2, Ne5; 13. c5, Nc4;
It seems Rubinstein will have to give up having the pair of bishops.
[d]r1b2rk1/1pqpb1pp/p3pn2/2P2p2/1PnN4/P1N1P1P1/1B3PBP/R2Q1RK1 w - - 1 14
14. Nxf5!!
A magic trick that will seriously weaken black. If Tarrasch capture the knight, white will play Qb3, and if Tarrasch takes the bishop then White plays Nxe7+.
14. ..., Bxc5
The best choice. Black wins a pawn.
15. bxc5, Dxc5!; 16. Nh6+!, gxh6; 17. Bc1, Qe5?
An error that will cost Tarrasch a precious tempo.
18. Qd3!, Qc5
Tarrasch now sees that he can't support the knight with d5 because of e4.
19. a4!, d6; 20. e4, Kg7; 21. Nd1, Bd7; 22. Ne3!
Very simple and strong. Tarrasch will have to concede his most active chessman.
22. ..., b5; 23. Nxc4, bxc4; 24. Qd2!
The winning retreat.
24. ..., Ne8; 25. e5!, Rb8;
The praeceptor germaniae couldn't cut the white square bishop's path with d6-d5 without opening the one bishop's one (Ba3).
26. Qxh6+, Kg8; 27. Qg5+!, Kf7;
If fleeing to h8, Rubinstein would win a chessman by playing Qe7.
28. Bf3, Rg8; 29. Qh5+, Kg7; 30. Be4, Qxe5; 31. Qxh7+, Kf8; 32. Qxd7! Qxe4; 33. Bh6+, resign.
Taking the rook would cost the queen (because Ng7 is countered with Bxg7 followed with Qxd6+ that wins a rook.)

(The bad english translation is 100% my fault, but hopefully I didn't betray Giffard's very pleasant style in french).

Here's the pgn of the game:


[Event "Hastings 1922"]
[Site "Hastings"]
[Date "1922"]
[Round ""]
[White "Akiba Rubinstein"]
[Black "Siegbert Tarrasch"]
[ECO "A84"]
[Result "1-0"]

1.d4 e6 2.c4 f5 3.g3 c5 4.Nf3 cxd4 5.Nxd4 Nf6 6.Bg2 Nc6 7.O-O
Bc5 8.e3 O-O 9.Nc3 a6 10.a3 Qc7 11.b4 Be7 12.Bb2 Ne5 13.c5 Nc4
14.Nxf5 Bxc5 15.bxc5 Qxc5 16.Nh6+ gxh6 17.Bc1 Qe5 18.Qd3 Qc5
19.a4 d6 20.e4 Kg7 21.Nd1 Bd7 22.Ne3 b5 23.Nxc4 bxc4 24.Qd2 Ne8
25.e5 Rb8 26.Qxh6+ Kg8 27.Qg5+ Kf7 28.Bf3 Rg8 29.Qh5+ Kg7 30.Be4
Qxe5 31.Qxh7+ Kf8 32.Qxd7 Qxe4 33.Bh6+ 1-0
Last edited by JuLieN on Sun Dec 11, 2011 3:35 am, edited 4 times in total.
"The only good bug is a dead bug." (Don Dailey)
[Blog: http://tinyurl.com/predateur ] [Facebook: http://tinyurl.com/fbpredateur ] [MacEngines: http://tinyurl.com/macengines ]
User avatar
JuLieN
Posts: 2949
Joined: Mon May 05, 2008 12:16 pm
Location: Bordeaux (France)
Full name: Julien Marcel

Re: Stylistic bias in computer chess programs

Post by JuLieN »

I just checked and my own engine, Prédateur, will play Nxf5 nearly instantly! ^^

[d]r1b2rk1/1pqpb1pp/p3pn2/2P2p2/1PnN4/P1N1P1P1/1B3PBP/R2Q1RK1 w - - 1 14[/b]
"The only good bug is a dead bug." (Don Dailey)
[Blog: http://tinyurl.com/predateur ] [Facebook: http://tinyurl.com/fbpredateur ] [MacEngines: http://tinyurl.com/macengines ]
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: Stylistic bias in computer chess programs

Post by Don »

Uri Blass wrote:
JuLieN wrote:Other players with very personal style that would be interesting to test:

- Pillsbury (self-taught, his style is the result of a theorization of the game more than from natural talent, so he sometimes committed awful blunders)
- Reti (deep natural talent + theorization, interesting mix!)
- Morphy (natural, logical player. Problem is, Morphy's level today would be around 2400, so we're back to the first posts of this thread.)
I disagree with you about Pillsbury

http://en.wikipedia.org/wiki/Harry_Nelson_Pillsbury


I think it means that pillsbury probably had more talent for chess relative to other players.

I do not believe that it is possible or was possible to become a top player
without having a big natural talent.
I'm not sure I agree with that but it's certainly arguable. There is a very high correlation between how much time (not just time but effort too) people spend on chess on how good they are. Fischer is an example of someone obsessed with chess and who played and studied more chess than we can imagine.

Rubenstein from the article that was just translated apparently was relatively weak and could not accept rook odds until he became obsessed with study and perfection. I think in athletics natural talent is much more important than in intellectual endeavors. I also believe that motivation is often mistaken for talent.

I know in Virginia there was a player who played tournament chess for many years at a very mediocre level, something like 1400 but he suddenly got very interested in improving and really applied himself to serious study, becoming something like a 2300+ player in a very short period of time. A friend of mine who is a solid USCF master told me about him and is also convinced that chess skill is much more about study than it is talent. It's about time as well as intensity or effort put into it.

I have no doubt that natural ability has some role, but I suspect it's not as important as most people think. Of course I'm perfectly willing to be wrong about this, I am certainly not an expert in whatever branch of science or psychology that studies this.

I did not look at the games of pillsbury but I read that
pillsbury suffered from serious health problems and I guess this was the reason for the blunders that you find.

You also say:
"Problem is, Morphy's level today would be around 2400"

I also disagree here because I do not believe that Morphy was weaker than other players of the past that came after him like Reti(Maybe all of them were not better than 2400 but in this case I see no reason to mention it only about Morphy)
BubbaTough
Posts: 1154
Joined: Fri Jun 23, 2006 5:18 am

Re: Stylistic bias in computer chess programs

Post by BubbaTough »

I have heard a lot of people downplay talent and claim anyone can be a master with enough study. Oddly enough, very few of them have spent much time giving chess lessons. For every story about some 1400 that worked hard and became a master there are a thousand stories about 1400's that worked hard and never broke 1600.

talent AND hard work are needed to excel in my opinion, and I always find it confusing that anyone could think differently as that principle is considered an obvious truth in almost everything else. The idea that the main difference between Anand and my Aunt Helga is study time is just silly.

-Sam