TrueChess Computer Analysis: Fischer is strongest ever. Botvinnik > Kasparov

[EGarrett]

For those of you who haven't seen it...

http://www.truechess.com/web/champs.html

Basically, this guy did a very detailed computer analysis using the 2007 version of Rybka and a modified version of Crafty to test the middle-game moves of all the World Champions (to avoid progress in opening theory), and he ranked the various players based on rate of blunders and percentage improvement over the average Grandmaster's move...with tables showing the rankings over 1, 2, 3, 5, 10 and 15-year periods.

It may or may not be a surprise that Fischer was found to be the most accurate human chess player that's ever lived on almost every list. The only one who topped a list other than Bobby was Capablanca, over the 15-year period.

Perhaps it may be another surprise that, over the 1, 2 and 3-year peaks, Mikhail Botvinnik was found to be more accurate than Garry Kasparov.

As an armchair fan, this was certainly interesting to discover, as I've been led to believe that Capablanca, Fischer and Kasparov are the three best players ever, however, it may seem that Botvinnik was actually the most talented Russian chess player ever (though Kasparov may remain the hardest-working).

This also, perhaps controversially, jibes with Nakamura's claim that Kasparov's strength came from his superior opening preparation, since over all short-term peaks, his middle-game accuracy is very good but not outstanding amongst the other World Champions.

I found this all to be interesting stuff, and I would love for a very similar analysis to be done now with Houdini 1.5...though it may be that at a certain level above human-strength, the computers extra calculating ability helps only when competing against each other, and they will all find the same mistakes in the human game database. Though it would be worth it just to see where Magnus and Aronian would rank.

[fezjones]

USA wins

[Do it Right]

These studies are invariably performed by less than stellar individuals.

This one took about 3 seconds to bust. I'm not even going to comment on using "a modified version of Crafty" to analyze world champion's games (not to mention effectively using -on average- a single core from 2007 hardware per game), but he also chose to arbitrarily exclude drawn games for his primary report. That's simply comical.

I'd put more effort into busting his report but he seems to have done a great job of invalidating himself from the get go.

[RoundTower]

Originally Posted by EGarrett

using the 2007 version of Rybka and a modified version of Crafty

so he just used one engine?

[KyleJRM82]

Stupid question: Would an increasing standard of play over time cause older players to have a higher accuracy score, because they would face less challenging positions?

[PyramidScheme]

I'm not sure if play outside of prepared ideas HAS improved much since 1950 or so. Now, obviously everyone in the 1800's was terrible.

[Do it Right]

Originally Posted by PyramidScheme

I'm not sure if play outside of prepared ideas HAS improved much since 1950 or so. Now, obviously everyone in the 1800's was terrible.

You can't be serious?

[Do it Right]

Originally Posted by KyleJRM82

Stupid question: Would an increasing standard of play over time cause older players to have a higher accuracy score, because they would face less challenging positions?

That's not at all a stupid question. It's one all these 'studies' somehow completely miss which is why I'm so harsh on the people involved in the study. They're ignoring the fact that it's inherently impossible to compare players in an imbiased fashion against peers from different eras.

Take this imaginary game:

1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. Bc4 Bc5 5. d3 d6 6. h3 h6 7. O-O O-O 8. Nd5 Nd4 9. Nxf6+ Qxf6 10. Nxd4 Bxd4 11. c3 Bb6 12. Be3 Bxe3 13. fxe3 Qe7 14. Qe2 Be6 15. Bxe6 fxe6 16. Rxf8+ Rxf8 17. Rf1 Rxf1+ 18. Qxf1 Qf7 19. Qxf7+ Kxf7 20. Kf2 1/2

Wow that game must have been played by two incredibly strong players. If you look at the computer eval I doubt it sways by much more than 0.1 or so at any given time!! Of course the reality is that the only people who'd play a game like that would be two very weak players. But their mutual mediocrity ensures that neither was able to make much of a mistake in computer terms. It very well may be that imbalancing the game at some point could give one player or another a theoretically decisive advantage, but computers are nowhere near the point of being able to determine that. All the computer can do is look at their moves and go: "Yip.. that's not terrible. I think 'x' is a little bit better but this is okay too." If one of our players, or the other, made some ridiculous blunder then the other player would likely be able to win the game with similar ease and accuracy through mutual mediocrity.

Obviously that's not to say that any world champion and his peers are a case from mutual mediocrity. But the problem still exists and inherently invalidates any of these 'studies'. Analyzing one player against entirely different peers than another is going to inherently bias the results and there simply is no reasonable way to balance for this. If one of these studies included Claude Bloodgood's game, whose story you can read on Wiki, he would almost certainly rank well above plenty of GMs even though he was a strong amateur at best. But his opponents were very weak amateurs for the most part.

[Do it Right]

And these studies inherently hide their absurdity by only including players who people can be reasonably swayed to believe may be stronger, or more accurate or whatever else, than another. It may be a little backwards but it doesn't inherently send off any major bull**** bells to say Fischer was a stronger player than Kasparov. But if the studied included large numbers of players and all the sudden you're getting unusual outliers like claiming 2000 rated "Mike 'DrawMaster' Demeister" (who had the strange quirk of refusing to ever play anybody anywhere near his strength or who was willing to imbalance positions) was a stronger or more accurate player than Kasparov then obviously you know the people involved in the study simply screwed something up somewhere.

[Cadaz]

US chess propaganda wooo

Erm I mean..

[EGarrett]

Originally Posted by Do it Right

These studies are invariably performed by less than stellar individuals.
...
he also chose to arbitrarily exclude drawn games for his primary report. That's simply comical.

Originally Posted by Do it Right

Take this imaginary game:

1. e4 e5 2. Nf3 Nc6 3. Nc3 Nf6 4. Bc4 Bc5 5. d3 d6 6. h3 h6 7. O-O O-O 8. Nd5 Nd4 9. Nxf6+ Qxf6 10. Nxd4 Bxd4 11. c3 Bb6 12. Be3 Bxe3 13. fxe3 Qe7 14. Qe2 Be6 15. Bxe6 fxe6 16. Rxf8+ Rxf8 17. Rf1 Rxf1+ 18. Qxf1 Qf7 19. Qxf7+ Kxf7 20. Kf2 1/2

Wow that game must have been played by two incredibly strong players. If you look at the computer eval I doubt it sways by much more than 0.1 or so at any given time!! Of course the reality is that the only people who'd play a game like that would be two very weak players.

Originally Posted by Do it Right

...if the studied included large numbers of players and all the sudden you're getting unusual outliers like claiming 2000 rated "Mike 'DrawMaster' Demeister" (who had the strange quirk of refusing to ever play anybody anywhere near his strength or who was willing to imbalance positions) was a stronger or more accurate player than Kasparov then obviously you know the people involved in the study simply screwed something up somewhere.

You seem to dislike the study because it excluded drawn games, then your examples of potential problems are a drawn game and a player nicknamed "Drawmaster."

It would seem that you've either outright contradicted yourself in your criticisms, or you've answered your own question about why the study didn't include draws.

[Noir_Desir]

i don't even get what makes Fischer the greatest according to his data. There does not seem to be a measure that determines the standings. It looks like taking a football table and saying, well this team won more often but the other scored more goals and a third had less players sent off and then making an arbitrary ranking.

[Do it Right]

Originally Posted by EGarrett

You seem to dislike the study because it excluded drawn games, then your examples of potential problems are a drawn game and a player nicknamed "Drawmaster."

It would seem that you've either outright contradicted yourself in your criticisms, or you've answered your own question about why the study didn't include draws.

Oh stop being so damn obtuse. Do you honestly not realize what excluding draws does? I refuse to believe you're that dumb and prefer to believe you're simply trolling me. And if you don't see the point the above game illustrates as it applies to decisive games then add a move where one of the 'players' hangs a piece and they continue to play weak, but computer approved, moves until one resigns. Wow, such accurate and amazing play from the winner!

[wlrs]

It's so much harder to play "accurately" when your opponents are constantly putting pressure on you (which I guess is one of the points DIR is trying to make).

[EGarrett]

Originally Posted by PyramidScheme

I'm not sure if play outside of prepared ideas HAS improved much since 1950 or so. Now, obviously everyone in the 1800's was terrible.

This is why Botvinnik's placement is interesting to me. Everyone has such appreciation for Kasparov's strength, but by these indications, Botvinnik was equally as (or more) talented. At least as far as calculating ability went.