25 Mar 13
2 edits
Originally posted by RJHindsPlease be reasonable.
On the following website is given the analysis of the games for the candidates for the World Championship:
http://london2013.fide.com/analysis/index.html
I was interested in looking at the round 8 match between the two leaders Aronian and Carlsen and noticed that the Houdini 2.0c x64 computer chess engine is being used for the evaluation analysis. Eac ...[text shortened]... ppears to me that both players have a high matchup rate with this Houdini computer chess engine.
The first seven moves belong to the opening.
Nobody ever tried to nail a computer user because the opening moves fit. It is in the nature of openings, that they repeat and well analysed.
Thus when analysing computer assistance the analyser is giving only values for moves after the longest recorded game with that move order....
EDIT I went to the analysis and the 365openings database knows 377 games with the move order leading to that game. Both players stayed on the most popular line until move 13, when Radjabov selected a move as black which is only 6 times in the databse (83 % white wins) instead of the short rochade which was chose 65 times in the database. Only in the 14th move Radjabov selected a move, not contained in that one.
Originally posted by PonderableYou must have a problem reading English because I am referring to the round 8 game between the leaders of the tournament at the time Carlsen as white and Aronian as black. Also I did not suggest that the computer matchup be compared until all pieces were develped for both sides and it appears that the middlegame had begun. That would be after move 13 and starting with move 14 when both playesrs seem to me to have a high matchup rate. I did not bother trying to figure out the % because I don't do that sort of thing. However, others do and have accused me of using a computer engine and claimed that grandmasters don't have high matchup rates with computer engines. This was the first game that I looked at that I noticed the evaluation analysis scores and the suggested moves by the Houdini computer. I have not bothered looking at any other games because I really don't care and do not think they are somehow cheating by using computers. I only wanted to point out that it apparently is not true that grandmasters do not sometimes have high matchup rates with computer engines.
Please be reasonable.
The first seven moves belong to the opening.
Nobody ever tried to nail a computer user because the opening moves fit. It is in the nature of openings, that they repeat and well analysed.
Thus when analysing computer assistance the analyser is giving only values for moves after the longest recorded game with that move ...[text shortened]... imes in the database. Only in the 14th move Radjabov selected a move, not contained in that one.
25 Mar 13
According to Houdini's analysis on this site:
http://london2013.fide.com/analysis/index.html
Game 8 Aronian - Carlsen
Match up with Houdini from move 14 on:
White (Aronian)
top 1 20/28 71.4%
top 2 21/28 75.9%
top 3 24/28 85.7%
Black (Carlsen)
top 1 18/27 66.7%
top 2 21/27 77.8%
top 3 23/27 85.2%
Originally posted by Dewi JonesThe OTB games I play are at much faster time controls than they are playing in the candidates matches and I am sure my matchup rates would probably be very low compared to theirs. However, my matchup rates are just for RHP games where I have much longer time for moves than they do and I also use the RHP analyze board, which they do not do OTB.
OK, we'll give you that one, as long as your OTB rating is something close to Carlsens
25 Mar 13
Originally posted by gambit05This is backward because Carlsen had the white pieces.
According to Houdini's analysis on this site:
http://london2013.fide.com/analysis/index.html
Game 8 Aronian - Carlsen
Match up with Houdini from move 14 on:
White (Aronian)
top 1 20/28 71.4%
top 2 21/28 75.9%
top 3 24/28 85.7%
Black (Carlsen)
top 1 18/27 66.7%
top 2 21/27 77.8%
top 3 23/27 85.2%
25 Mar 13
4 edits
Originally posted by RJHindsI only wanted to point out that it apparently is not true that grandmasters do not sometimes have high matchup rates with computer engines.
You must have a problem reading English because I am referring to the round 8 game between the leaders of the tournament at the time Carlsen as white and Aronian as black. Also I did not suggest that the computer matchup be compared until all pieces were develped for both sides and it appears that the middlegame had begun. That would be after move 13 and s is not true that grandmasters do not sometimes have high matchup rates with computer engines.
No one has ever claimed that sometimes grandmasters don't have high computer matchups. Who ever said that?
The computer analysis starts when the game goes 'out of book', here's the nearest game to the Carlsen game -
[Event "Spanish Team Championship (Honor Division)"]
[Site "Leon ESP"]
[Date "2012.11.11"]
[EventDate "2012.11.05"]
[Round "7.2"]
[Result "1/2-1/2"]
[White "Ivan Cheparinov"]
[Black "Pentala Harikrishna"]
[ECO "E06"]
[WhiteElo "2710"]
[BlackElo "2692"]
[PlyCount "65"]
The games deviate on move 15 for black.
Originally posted by Proper KnobI guess they have pretty good memories then. However, 15... c6 was not one of the 3 Houdini computer moves. So I guess we can conclude that Aronian was not using Houdini.
[b]I only wanted to point out that it apparently is not true that grandmasters do not sometimes have high matchup rates with computer engines.
No one has ever claimed that sometimes grandmasters don't have high computer matchups. Who ever said that?
The computer analysis starts when the game goes 'out of book', here's the nearest game to the Ca d6 33. axb5 1/2-1/2[/pgn]
The games deviate on move 15 for black.[/b]
Since Carlsen is the highest rated and he matched Houdini better than Aronian, I decided to look and see how he matched up with Houdini on his next game in round 9 against Kramnik. Like Aronian he also made some moves in this game against Kramnik that were not one of the 3 choices of Houdini as did Kramnik. Neither player seemed to me to match up as well with Houdini as in the first game that I looked at in round 8. However, what was interesting to me about the evaluation of the moves by Houdini was that two of Carlsen's moves seemed to have been better than the 3 choices given by Houdini because the two moves by Calsen's reduced Kramnik's advantage more than is shown by any of the choices given by the Houdini computer. I don't know if I really understand this and I probably don't. Another strange thing I noticed wa that move 28. a3 by Kramnik was not on Houdini's list and reduced Kramnik's advantage to an evaluation of 0.00 making the game dead even. That would make that move almost a blunder even though it looks like a perfectly reasonable move to me. Perhaps, this Houdini computer is flawed. That is my conclusion.
27 Mar 13
Originally posted by RJHindsHi Ronald,
Since Carlsen is the highest rated and he matched Houdini better than Aronian, I decided to look and see how he matched up with Houdini on his next game in round 9 against Kramnik. Like Aronian he also made some moves in this game against Kramnik that were not one of the 3 choices of Houdini as did Kramnik. Neither player seemed to me to match up as well ...[text shortened]... reasonable move to me. Perhaps, this Houdini computer is flawed. That is my conclusion.
yes the computer is flawed, since it is in fact a clculation amchine. So how does a chess computer work?
* the principal rules are coded in.
* an evaluation algorithm is coded in. In this algorithm a (human) programmer applies his understanding into the software. So piece count is an evident measure to evaluate a position. Combination planing is not (as long as we can't code plans).
* The computer evaluates a position, and tries various legal moves. Since it can calculate at an amazing speed it takes not too much time. After creating the move it is evaluated, then the possible countermoves are made and the ensuiing position evaluated. As can easily be seen the amount of possibilities is inxcreasing fast. So another method is used called "pruning". all positions below a threshold are scrapped and only with the "good" positions the above mentuioned procedure is repeated.
There is something called the "hotizon effect". This is describes the problem, that even the fastest computer can't do a complete evaluation of all possible moves and countermoves. At the point this would be possible Chess would be "solved". So it has to prune (not follow unpromising lines).
So a well placed scarifice which effect shows up only several moves later is an efficient weapon against a computer. (If you can find it) As is a good positional move.
Computers don't plan and can't understand plans. Computer evaluate positions and look for an increase inn score. So there are typical computer moves due to evaluation algorithms, which can be picked out by the experienced expert.
This is also the reason why a computer will give some line at a given position, but will alter this after the next move. It was not a plan and the horizon shifted...
It is important t
Originally posted by RJHindsHave you put houdini to test your games here at RHP?
Since Carlsen is the highest rated and he matched Houdini better than Aronian, I decided to look and see how he matched up with Houdini on his next game in round 9 against Kramnik. Like Aronian he also made some moves in this game against Kramnik that were not one of the 3 choices of Houdini as did Kramnik. Neither player seemed to me to match up as well ...[text shortened]... reasonable move to me. Perhaps, this Houdini computer is flawed. That is my conclusion.
27 Mar 13
2 edits
Originally posted by RJHindsI have no idea, are there websites that do just that? Why can't you run your own games through Houdini? What is Houdini anyway? Never mind, I just found it. It's not free, and the latest is H 3 and there is a regular price and a pro price which can run on much fancier machines. How does it stand up in games against fritz or Rybka? So it has a feature where you can pgn in a game and it analyzes it for comparison of play, how close to computer match someone does?
No. What website can you go to to test your own games?