pichy wrote:
Now that you have Komodo 64-bit. Test it with this endgame position without tablebase to see if it can find the best move Be8 sacrificing the bishop
[D]2b2k1K/6p1/p1p2p1p/P2p1P1B/1P1P2P1/7P/8/8 w - - 0 1
This is not an interesting position. White does not have to play Be8 right away to win this, he can play it any time. White can kill a tempo with Kh7 and black has to make a committal move and white easily wins with Be8 later. In fact this is in some of the PV's you posted.
On the latest development version of Komodo which is 30 ELO stronger than Komodo 4 it does play Be8 in 26 ply in 36 seconds on my 3 year old Desktop machine but I don't believe that this proves anything.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
lech wrote:Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
But you have to do better than just to enumerate the problems in computer chess. You have show that this actually makes them play weaker than humans. Showing they play weaker in a few endings does not say anything about how the play in general, even in the endings.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
lech wrote:Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
But you have to do better than just to enumerate the problems in computer chess. You have show that this actually makes them play weaker than humans. Showing they play weaker in a few endings does not say anything about how the play in general, even in the endings.
I have not to do something better.
Some problems with zugzwangs on higher ply, cascades of zugzwangs and pawn endgames seems to be a sad fact. Maybe 1% ELO gain can solve it.
pichy wrote:
Now that you have Komodo 64-bit. Test it with this endgame position without tablebase to see if it can find the best move Be8 sacrificing the bishop
[D]2b2k1K/6p1/p1p2p1p/P2p1P1B/1P1P2P1/7P/8/8 w - - 0 1
This is not an interesting position. White does not have to play Be8 right away to win this, he can play it any time. White can kill a tempo with Kh7 and black has to make a committal move and white easily wins with Be8 later. In fact this is in some of the PV's you posted.
On the latest development version of Komodo which is 30 ELO stronger than Komodo 4 it does play Be8 in 26 ply in 36 seconds on my 3 year old Desktop machine but I don't believe that this proves anything.
What is proves is that your engine Komodo will eventually play the endgames without tablebase as good as the top GMs like Capablanca, Averkak, Smylov, and Current GMs Arorian and GM Kramnik
lech wrote:Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
But you have to do better than just to enumerate the problems in computer chess. You have show that this actually makes them play weaker than humans. Showing they play weaker in a few endings does not say anything about how the play in general, even in the endings.
There are several positions that I can use from the book the Greatest Chess Endings ever Played, in which none of the top engines could play them as good as those master played them.
PS: The topic here has been that most top engines are not as good as the top GM playing endings when you allowed them plenty of time to avoid blunders that engines will not commit.
lech wrote:Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
But you have to do better than just to enumerate the problems in computer chess. You have show that this actually makes them play weaker than humans. Showing they play weaker in a few endings does not say anything about how the play in general, even in the endings.
There are several positions that I can use from the book the Greatest Chess Endings ever Played, in which none of the top engines could play them as good as those master played them.
There are tons of position where the computer does not play the same move, for instance some move that is annotated as being a "great move" or "best" when in fact the computer is playing something that is just as good.
Anyway, give me the fen's of these "several positions" and what the "right" move is supposed to be and I will take a look - even use it to improve Komodo if there is something to be learned.
PS: The topic here has been that most top engines are not as good as the top GM playing endings when you allowed them plenty of time to avoid blunders that engines will not commit.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
Too many emotions.
I earlier wrote about some problems of engines (http://talkchess.com/forum/viewtopic.ph ... e3e#452323). Pawn endgames can be a solid problem too. In this case all endgames have a problem, since can lead to pawn endgames.
It is possible to correct it, enough.
Maybe authors of Komodo did it earlier.
lech wrote:Engines have problems with zugzwangs on higher ply, cascades of zugzwangs and pawn engames. A static evoluation of pawns (passed, doubled, backward) can't match to lots of pawn endgames. For this reason engines will not play well in many of them. Especially without tablebases.
An engin or a chess player is not able to play endgames well, if "don't understand" pawn endgames or has a problem with zugzwangs.
But you have to do better than just to enumerate the problems in computer chess. You have show that this actually makes them play weaker than humans. Showing they play weaker in a few endings does not say anything about how the play in general, even in the endings.
There are several positions that I can use from the book the Greatest Chess Endings ever Played, in which none of the top engines could play them as good as those master played them.
There are tons of position where the computer does not play the same move, for instance some move that is annotated as being a "great move" or "best" when in fact the computer is playing something that is just as good.
Anyway, give me the fen's of these "several positions" and what the "right" move is supposed to be and I will take a look - even use it to improve Komodo if there is something to be learned.
PS: The topic here has been that most top engines are not as good as the top GM playing endings when you allowed them plenty of time to avoid blunders that engines will not commit.
It's good you will evaluate it. There's always something to learn. The Endgame is very much a fertile ground of research for GM/s and for programmers of GM engines.
pichy wrote:
Now that you have Komodo 64-bit. Test it with this endgame position without tablebase to see if it can find the best move Be8 sacrificing the bishop
[D]2b2k1K/6p1/p1p2p1p/P2p1P1B/1P1P2P1/7P/8/8 w - - 0 1
On the latest development version of Komodo which is 30 ELO stronger than Komodo 4 it does play Be8 in 26 ply in 36 seconds on my 3 year old Desktop machine but I don't believe that this proves anything.
Have you?
When you release it finally i will be able to show what i already know:
Komodo is stronger than Houdini even at blitz 5'+3'' in a head to head match.
