Interesting. Gives some insight into the meaning of computer analysis (what with some moves having good responses, but being part of an inferior plan). Thanks for sharing.
Somewhat off topic: I'd like to ask if anyone knows of annotations for Capablanca - Alekhine World Championship Game 1. I was looking over this game the other week and there were a few moves that I needed clarification on. Thanks.
I think it is very interesting to see how the greats of yesteryear stack up against the silicon. I also think to a certain extent it shows that Alekhine and Capablanca were not really any weaker than top players today whose match ups seem similar. The key thing for me in producing a control sample of what is humanly possible is to see what match ups are achieved by players playing in similar conditions to server based chess. Those pre-engine CC champions who played to a high level OTB like Ragozin and O'Kelly de Galway would be interesting as you might expect a higher match up in their CC games, although there might actually be little difference. Other cross over players like Ulf Andersson and Johnathon Penrose could be interesting too, although they are probably too close to the engine era to be 100% sure. Penrose might be even more interesting because his opening preparation - for example the Queens Gambit Accepted - also crossed over.
I wonder what would have been the match-up rate if you gave Fritz 11 60 seconds instead of 30. Would the results be too diferent? Maybe not.
Thanks for sharing your analysis with the community.
Originally posted by Squelchbelch 12th World Chess Championships
Capablanca vs Alekhine 1927
Buenos Aires
Why did I choose this match?
There are several reasons why I wanted to analyse this famous World Championships.
Capablanca is often quoted as one of the most accurate players of all time. He is statistically one of the least blunder-prone, so therefore I would expect qui ...[text shortened]... 2,5% )
Top 3 Match: 34/40 (85,0% )
My friend, you have far, far, far, too much time on your hands 😛. Just kidding- very nice analysis, I was also struck by the fact that Capablanca matched more often, but lost the match- perhaps it was that the moves that he didn't match were the most critical ones
Originally posted by chesskid001 My friend, you have far, far, far, too much time on your hands 😛. Just kidding- very nice analysis, I was also struck by the fact that Capablanca matched more often, but lost the match- perhaps it was that the moves that he didn't match were the most critical ones
Or perhaps the engine isn't always correct. That seems pretty plausible since Rybka 3 is stronger than Fritz so Rybka won't match up with Fritz 100%
Also, I'm not sure on this, but don't the top players beat Fritz?
Yes, but Fritz's strength lies in it's analysis capability.
It quickly realises if it made an error in judgement and re-adjusts the scoring if a player makes a stronger move than any of Fritz's top 3 30 second choices.
The strength of play in the games I'm analysing means that most of the time, Fritz has a good idea of the values of the lines relative to each other because they feature in it's hash tables.
With weaker players who regularly pick moves outside the top 3, Fritz has to do far more reassessing throughout games.
Of course it is also interesting to compare the OTB Super GM masters & the World Correspondence champs from the pre-computer age with other online player's matchup rates.
I'm analysing the 1972-'75 CC World Championships now, and maybe after that will finish on Kasparov-Karpov 1985.
I think then I will have plenty of high standard benchmark statistics.
Smaia; I tried 60 seconds vs 30, and the scores for 3 games were almost exactly the same, but analysis took too long for practical purposes. It simply isn't worth the hassle.
Originally posted by tomtom232 Also, I'm not sure on this, but don't the top players beat Fritz?
no, they are helpless against fritz at this stage in equal conditions. But I guess super GMs would beat fritz moving every 30 seconds in multi variation mode from time to time.
Game 27 is interesting. Almost a 30 point difference in the top 3 match ups and Alekhine pulled out a draw. With all the high match ups did Fritz evaluate Capablanca as having the better position throughout the game? Or did Fritz score the game about even give or take a half pawn?
Edit: Fritz 10 scored Capablanca 4.5 points ahead of Alekhine until 38.Kf2. Ke2 would have prevented the draw and maintained Capablanca's lead. All those good moves undone by one mistake.
Originally posted by KneeCaps Game 27 is interesting. Almost a 30 point difference in the top 3 match ups and Alekhine pulled out a draw. With all the high match ups did Fritz evaluate Capablanca as having the better position throughout the game? Or did Fritz score the game about even give or take a half pawn?
Edit: Fritz 10 scored Capablanca 4.5 points ahead of Alekhine until 3 ...[text shortened]... evented the draw and maintained Capablanca's lead. All those good moves undone by one mistake.
Yes; Capablanca chose more top 3 moves but at the same time played about 4 or 5 real howlers in the 1927 match.
Alekhine made very few -1.0+ 'blunders' in comparison. But often chose moves just outside the top 3 when the difference in his choice & the Fritz's number 3 was rather insignificant.
The top 3 matchup figures only reveal part of the story!
Originally posted by Squelchbelch If I do another Super GM match it'll probably be Kasparov-Karpov WC 1985.
Next up I think will be analysis of a pre-computer era correspondence chess World Championships either from the late '60s or early 1970s.
One thing I noticed throughout Capablanca-Alekhine '27 is that although Capa had slightly better total average matchup stats, he was actual ...[text shortened]... ds of
top 1 match = 60%
top 2 match = 75%
top 3 match = 85%
which was not so surprising.
How about Hans Berliner vs Yakov Estrin 1965 CC Championship?
Has anyone tried an analysis of, say, Rybka games with this Fritz match-up method? It would be interesting to see how another engine scores on the match-ups.
I think this sort of analysis would appeal to the larger chess community. If you've got the skills, you should register a domain and share all your data. Seriously.