George wrote:In most important top human versus human tournaments, they usually select Stockfish as the prefered program for analyzing all the current positions and all the best variations or possible next best 12 or 16 moves, but in my opinion the best feature that can and could be added to Stockfish is to be able to keep track of how many times during the tournament each top players select stockfish prefered best move and provide a statistical percentage. This would be a very interesting feature to keep track and see how the human brain of the best top players as they get higher in rating select similar moves to the best top chess program like Stockfish.
You can achieve this feature right now.
Using a tool like pgn2fen, create a list of positions together with the chosen moves.
Analyze the positions with stockfish and see if they agree.
I found (after deep analysis) that Stockfish agrees with Louis Eichborn about 55% of the time.
With a ponder hit that high, we need to check his brain to see if some stockfish code is embedded in there. Anybody got a shovel?
Can you create a list of positions from actual high level tournament where GM Calsen, GM, Caruana, GM Nakamura, GM Anand, GM Aronian, GM Kramnik participated in the last years with the chosen moves made by those players, and not only see if Stockfish agree, but to create a statistical bar chart showing which player selected or agreed to played the most moves chosen by Stockfish?
PS: But the ideal thing to do is to be able to do this for every high level tournament and provide a constant statistical report as what player has chosen the most moves recommended by Stockfish during the entire tournament, and to see if that player coincide to be the winner of that particular tournament
It is a lot more interesting to me to do the analysis for long dead players. That removes a variable:
"Did the player in question use Stockfish to analyze the opening before-hand?"
That's why I did the analysis for Louis Eichborn (that and also the fact that I did not know much about him until I recently found a web page about his exploits).
It took about a machine-week to do Eichborn. He does not have very many games. Do do someone like Kasparov would probably take a machine-year.
That is the reason why I posted that it would be much easier to create an extra feature in Stockfish that as the top 10 players participate in tournaments and select moves to have Stockfish automatically keeping track of of the players and who has selected most of the moves recommended by Stoskfish
Stockfish has no knowledge even of whose moves it is analyzing.
The tracking of something like this would require a second external program or extensive modifications to stockfish.
I do not think that this feature would be easy to create.
The program would need to be able to read PGN (or Epd OR Fen), figure out the name of the player it is analyzing, understand the time or depth wanted, and go through all of the moves for that player, storing the result in a database or in a file.
We would probably also want a GUI to control this stuff as well.
Dan this would be also beneficial for human versus human using Stockfish, or human versus Stockfish, and after we the human make a move we can later see what was stockfish best move
George wrote:In most important top human versus human tournaments, they usually select Stockfish as the prefered program for analyzing all the current positions and all the best variations or possible next best 12 or 16 moves, but in my opinion the best feature that can and could be added to Stockfish is to be able to keep track of how many times during the tournament each top players select stockfish prefered best move and provide a statistical percentage. This would be a very interesting feature to keep track and see how the human brain of the best top players as they get higher in rating select similar moves to the best top chess program like Stockfish.
Last year I created a tool to see which engine is similar to human, I also use the similarity tool but replaced it with positions from the subject player.
Here is an example.
Carlsen as white using 10389 positions white to play from games in yr, 2010-2014.
Engine: 64mb hash, single thread, 100ms per position.
sim version 3
------ id name Carlsen-white (time: 100 ms scale: 1.0) ------
54.04 id name Stockfish 5 64 SSE4.2 (time: 100 ms scale: 1.0)
53.47 id name Critter 1.6a 64-bit (time: 100 ms scale: 1.0)
52.94 id name Senpai 1.0 (time: 100 ms scale: 1.0)
52.77 id name Houdini 4 x64 (time: 100 ms scale: 1.0)
52.31 id name iCE 2.0 v2240 x64/popcnt (time: 100 ms scale: 1.0)
51.98 id name spark-1.0 (time: 100 ms scale: 1.0)
51.93 id name Fire 3.0 x64 (time: 100 ms scale: 1.0)
51.55 id name Hannibal 1.4x64 (time: 100 ms scale: 1.0)
51.42 id name Protector 1.7.0 (time: 100 ms scale: 1.0)
50.81 id name Deuterium v14.3.34.130 (time: 100 ms scale: 1.0)
50.36 id name Texel 1.04 64-bit (time: 100 ms scale: 1.0)
50.13 id name Vajolet2 1.45 (time: 100 ms scale: 1.0)
49.88 id name GNU Chess 5.60-64 (time: 100 ms scale: 1.0)
49.79 id name Gaviota v1.0 (time: 100 ms scale: 1.0)
49.76 id name Rhetoric 1.4.1 x64 (time: 100 ms scale: 1.0)
49.45 id name Andscacs 0.70 (time: 100 ms scale: 1.0)
49.37 id name Rodent 1.6 (build 6) (time: 100 ms scale: 1.0)
49.15 id name Nebula 2.0 (time: 100 ms scale: 1.0)
49.03 id name Yace 0.99.87 (time: 100 ms scale: 1.0)
47.98 id name Alaric 707 (time: 100 ms scale: 1.0)
45.81 id name Ares 1.005-64advW (time: 100 ms scale: 1.0)
George wrote:In most important top human versus human tournaments, they usually select Stockfish as the prefered program for analyzing all the current positions and all the best variations or possible next best 12 or 16 moves, but in my opinion the best feature that can and could be added to Stockfish is to be able to keep track of how many times during the tournament each top players select stockfish prefered best move and provide a statistical percentage. This would be a very interesting feature to keep track and see how the human brain of the best top players as they get higher in rating select similar moves to the best top chess program like Stockfish.
Last year I created a tool to see which engine is similar to human, I also use the similarity tool but replaced it with positions from the subject player.
Here is an example.
Carlsen as white using 10389 positions white to play from games in yr, 2010-2014.
Engine: 64mb hash, single thread, 100ms per position.
sim version 3
------ id name Carlsen-white (time: 100 ms scale: 1.0) ------
54.04 id name Stockfish 5 64 SSE4.2 (time: 100 ms scale: 1.0)
53.47 id name Critter 1.6a 64-bit (time: 100 ms scale: 1.0)
52.94 id name Senpai 1.0 (time: 100 ms scale: 1.0)
52.77 id name Houdini 4 x64 (time: 100 ms scale: 1.0)
52.31 id name iCE 2.0 v2240 x64/popcnt (time: 100 ms scale: 1.0)
51.98 id name spark-1.0 (time: 100 ms scale: 1.0)
51.93 id name Fire 3.0 x64 (time: 100 ms scale: 1.0)
51.55 id name Hannibal 1.4x64 (time: 100 ms scale: 1.0)
51.42 id name Protector 1.7.0 (time: 100 ms scale: 1.0)
50.81 id name Deuterium v14.3.34.130 (time: 100 ms scale: 1.0)
50.36 id name Texel 1.04 64-bit (time: 100 ms scale: 1.0)
50.13 id name Vajolet2 1.45 (time: 100 ms scale: 1.0)
49.88 id name GNU Chess 5.60-64 (time: 100 ms scale: 1.0)
49.79 id name Gaviota v1.0 (time: 100 ms scale: 1.0)
49.76 id name Rhetoric 1.4.1 x64 (time: 100 ms scale: 1.0)
49.45 id name Andscacs 0.70 (time: 100 ms scale: 1.0)
49.37 id name Rodent 1.6 (build 6) (time: 100 ms scale: 1.0)
49.15 id name Nebula 2.0 (time: 100 ms scale: 1.0)
49.03 id name Yace 0.99.87 (time: 100 ms scale: 1.0)
47.98 id name Alaric 707 (time: 100 ms scale: 1.0)
45.81 id name Ares 1.005-64advW (time: 100 ms scale: 1.0)
Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best and any other moves that is NOT recommended by stockfish or Komodo as the best should not be added. The information could be represented in a bar graph or simply by providing the total # of moves og the GM that had the highest percentage of moves recommended by stockfish or KOMODO, then we the observers can easily figure if the player that had the highest percentage also happen to win the tournament
PS: For example if GM Carlsen is playing versus GM Caruana and in move 51. Stockfish recommend 51. Ka1a2 and Carlsen played the same move GM Carlsen get an added moved, but if Carlsen makes another move 51. Ka1b2 even if it is of equal level of importance or as good as the move recommended by Stockfish or Komodo that move is NOT added and so on and on etc.....
In this example I believe that the only winning move is the one recommended by Stockfish or Komodo 51. Ka1a2
[d]6k1/7R/6P1/7P/8/6r1/8/K7 w - - 0 1
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best
How can we compare the move made by a human when we will not run the engine? There are softwares that can keep track of live pgn, if this is what you mean? But still the engine has to be run.
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best
How can we compare the move made by a human when we will not run the engine? There are softwares that can keep track of live pgn, if this is what you mean? But still the engine has to be run.
Ferdinand in most top human versus human tournaments Stockfish or Komodo, but mainly Stockfish is being used to analyze and recommend the next best moves so the observer or chess fans know in advanced which is the next best move, the only tool needed is to be able to track and add all the moves instantly by the top humans and to be able to provide some type of count so at the end of the tournament we can compare if indeed the top GM that selected the most moves recommended by Stockfish as the next best move really ends up being the winner
PS: For example in any given game where GM Carlsen played versus GM Caruana if GM Carlsen played 22 moves recommend by stockfish out of 68, and GM Caruana only selected 16 moves recommended by stockfish, but ended up winning that particular game, it only showed us that Good moves made by computers are NOT necessarily the only best moves which could win that particular game in the fastest way. Sometimes computer moves are only good against other computers, but due to the horizon effect it could end up being the worse move
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best
How can we compare the move made by a human when we will not run the engine? There are softwares that can keep track of live pgn, if this is what you mean? But still the engine has to be run.
Ferdinand in most top human versus human tournaments Stockfish or Komodo, but mainly Stockfish is being used to analyze and recommend the next best moves so the observer or chess fans know in advanced which is the next best move, the only tool needed is to be able to track and add all the moves instantly by the top humans and to be able to provide some type of count so at the end of the tournament we can compare if indeed the top GM that selected the most moves recommended by Stockfish as the next best move really ends up being the winner
PS: For example in any given game where GM Carlsen played versus GM Caruana if GM Carlsen played 22 moves recommend by stockfish out of 68, and GM Caruana only selected 16 moves recommended by stockfish, but ended up winning that particular game, it only showed us that Good moves made by computers are NOT necessarily the only best moves which could win that particular game in the fastest way. Sometimes computer moves are only good against other computers, but due to the horizon effect it could end up being the worse move
Sorry I am giving up, I don't know what you are saying .
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best
How can we compare the move made by a human when we will not run the engine? There are softwares that can keep track of live pgn, if this is what you mean? But still the engine has to be run.
Ferdinand in most top human versus human tournaments Stockfish or Komodo, but mainly Stockfish is being used to analyze and recommend the next best moves so the observer or chess fans know in advanced which is the next best move, the only tool needed is to be able to track and add all the moves instantly by the top humans and to be able to provide some type of count so at the end of the tournament we can compare if indeed the top GM that selected the most moves recommended by Stockfish as the next best move really ends up being the winner
PS: For example in any given game where GM Carlsen played versus GM Caruana if GM Carlsen played 22 moves recommend by stockfish out of 68, and GM Caruana only selected 16 moves recommended by stockfish, but ended up winning that particular game, it only showed us that Good moves made by computers are NOT necessarily the only best moves which could win that particular game in the fastest way. Sometimes computer moves are only good against other computers, but due to the horizon effect it could end up being the worse move
Sorry I am giving up, I don't know what you are saying .
Ferdinand sa pinaka tuktok ng tao laban sa tao paligsahan Stockfish o Komodo , ngunit higit sa lahat Stockfish ay ginagamit upang pag-aralan at magrekomenda ng susunod na pinakamahusay na gumagalaw upang malaman ang tagamasid o chess mga tagahanga sa mga advanced na ang susunod na pinakamahusay na ilipat, ang tanging tool na kailangan ay upang magagawang subaybayan at idagdag ang lahat ng mga gumagalaw agad sa pamamagitan ng mga nangungunang mga tao at upang magawang magbigay ng ilang uri ng bilang ng gayon sa pagtatapos ng paligsahan maaari naming ihambing kung sa katunayan tuktok GM na napili ang pinaka gumagalaw iminumungkahi ng mga Stockfish bilang ng susunod na pinakamahusay na paglipat talaga ay nagtatapos up pagiging ang nanalo http://www.uschesschamps.com/2015-us-ch ... ation/live
For example if Ivanchuk played versus GM So Wesley in TaTA Steel 15 and GM So Wesley played 13 moves which were recommended by Houdini during that particular game out of a total of 26 moves, but Ivanchuk only played 7 moves recommended by Houdini out of the entire 26 moves in that game, more likely GM So Wesley will end up winning that game since GM Wesley So selected more winning moves then Ivanchuk. http://www.chessdom.com/tata-steel-chess-2015-live/
George wrote:Ferdinand can you create another tool that keep track of how many times during any tournament each top players select stockfish prefered best move and provide a statistical percentage?
It is not too difficult to create such tool, but my concern is who will do the analysis? As what I have observed, generally the player that has higher ranks in the tourney has a high rate of matching the best moves of the engine.
A tool here just need a little tweaking and can already show a simple stat. http://talkchess.com/forum/viewtopic.php?t=56010
But this need a custom epd which is not included in the shared tool. So what I can do is convert the tournaments pgn file to the custom epd then someone will do the analysis and eventually producing the stat.
Ferdinand there is no need to make any analysis if a module or a tool of a program can keep track of the moves made by all the players and instantly add only those moves that each player made that are equal to the moves that Stockfish or Komodo recommended as the best
How can we compare the move made by a human when we will not run the engine? There are softwares that can keep track of live pgn, if this is what you mean? But still the engine has to be run.
Ferdinand in most top human versus human tournaments Stockfish or Komodo, but mainly Stockfish is being used to analyze and recommend the next best moves so the observer or chess fans know in advanced which is the next best move, the only tool needed is to be able to track and add all the moves instantly by the top humans and to be able to provide some type of count so at the end of the tournament we can compare if indeed the top GM that selected the most moves recommended by Stockfish as the next best move really ends up being the winner
PS: For example in any given game where GM Carlsen played versus GM Caruana if GM Carlsen played 22 moves recommend by stockfish out of 68, and GM Caruana only selected 16 moves recommended by stockfish, but ended up winning that particular game, it only showed us that Good moves made by computers are NOT necessarily the only best moves which could win that particular game in the fastest way. Sometimes computer moves are only good against other computers, but due to the horizon effect it could end up being the worse move
Sorry I am giving up, I don't know what you are saying .
Ferdinand sa pinaka tuktok ng tao laban sa tao paligsahan Stockfish o Komodo , ngunit higit sa lahat Stockfish ay ginagamit upang pag-aralan at magrekomenda ng susunod na pinakamahusay na gumagalaw upang malaman ang tagamasid o chess mga tagahanga sa mga advanced na ang susunod na pinakamahusay na ilipat, ang tanging tool na kailangan ay upang magagawang subaybayan at idagdag ang lahat ng mga gumagalaw agad sa pamamagitan ng mga nangungunang mga tao at upang magawang magbigay ng ilang uri ng bilang ng gayon sa pagtatapos ng paligsahan maaari naming ihambing kung sa katunayan tuktok GM na napili ang pinaka gumagalaw iminumungkahi ng mga Stockfish bilang ng susunod na pinakamahusay na paglipat talaga ay nagtatapos up pagiging ang nanalo http://www.uschesschamps.com/2015-us-ch ... ation/live
For example if Ivanchuk played versus GM So Wesley in TaTA Steel 15 and GM So Wesley played 13 moves which were recommended by Houdini during that particular game out of a total of 26 moves, but Ivanchuk only played 7 moves recommended by Houdini out of the entire 26 moves in that game, more likely GM So Wesley will end up winning that game since GM Wesley So selected more winning moves then Ivanchuk.
Or in the game GM So Wesley Vs Van Wely you can see that GM So Wesley played more moves recommended by Stockfish than Van wely did, therefore GM So Wesley won since he played more stronger moves than GM Van Wely. http://www.chessdom.com/tata-steel-chess-2015-live/
