Martin Thoresen wrote:
This is the first time EVER in TCEC history that a Stage is played without books. Yes, that's right, in Stage 2 the engines are on their own from the very first move.
Does anyone think this might not be fair in that some programs may be designed to play from a position AFTER enough opening moves, and were not equipped to play the first few very intelligently whilst others are?
I think of it the opposite way. I think it is an appropriate reward to those engines whose authors have actually worked at making sure there engines are good at making opening moves.
Martin Thoresen wrote:
This is the first time EVER in TCEC history that a Stage is played without books. Yes, that's right, in Stage 2 the engines are on their own from the very first move.
Does anyone think this might not be fair in that some programs may be designed to play from a position AFTER enough opening moves, and were not equipped to play the first few very intelligently whilst others are?
I think of it the opposite way. I think it is an appropriate reward to those engines whose authors have actually worked at making sure there engines are good at making opening moves.
-Sam
I agree that those engines might be rewarded best in Fisherandom chess, but with regular chess, some programmers might feel that the right opening book is designed to take care of any important line played by the other side, and therefore didn't give any specific opening instructions, besides general play. But the first few moves might require a very different list of instructions than after the first few moves.
Or is this not true?
And if it is, then doesn't this make it great for fisherandom?
Martin Thoresen wrote:
This is the first time EVER in TCEC history that a Stage is played without books. Yes, that's right, in Stage 2 the engines are on their own from the very first move.
Does anyone think this might not be fair in that some programs may be designed to play from a position AFTER enough opening moves, and were not equipped to play the first few very intelligently whilst others are?
I think of it the opposite way. I think it is an appropriate reward to those engines whose authors have actually worked at making sure there engines are good at making opening moves.
-Sam
If you make a tournament of 100 games all starting from a particular rare endgame position, which requires particular knowledge to deliver mate, then we would find that all those engines who were given that knowedge will win every game, and all those were not, will NOT win any game.
So we would be proving nothing! Just ask the programmer.
I was wondering if that might be comparable here, but to a smaller extent.
Martin Thoresen wrote:
This is the first time EVER in TCEC history that a Stage is played without books. Yes, that's right, in Stage 2 the engines are on their own from the very first move.
Does anyone think this might not be fair in that some programs may be designed to play from a position AFTER enough opening moves, and were not equipped to play the first few very intelligently whilst others are?
I think of it the opposite way. I think it is an appropriate reward to those engines whose authors have actually worked at making sure there engines are good at making opening moves.
-Sam
If you make a tournament of 100 games all starting from a particular rare endgame position, which requires particular knowledge to deliver mate, then we would find that all those engines who were given that knowedge will win every game, and all those were not, will NOT win any game.
So we would be proving nothing! Just ask the programmer.
I was wondering if that might be comparable here, but to a smaller extent.
I think you are making a good analogy. I agree it is comparable in that engines tested from the start position and positions close to the start position (after 1 or 2 or 3 moves) will probably have a slight advantage over engines that only start testing later in the opening. The difference in my mind, is that a rare endgame rarely comes up in a game, while the start position always comes up in a game. So it seems to me that if a tournament is supposed to test an engine's ability to play chess from start to finish, it is nice to actually start at the start.
Of course, there are issues with doing this over and over and getting repeat games or repeat decisive portions of games, but that is a bit tangential to the topic we are discussing .
Game 1 of stage 2 was very nice.
The games between the big 3 are 13, 62, and 81.
I'll be very much looking forward to see their openings and how it continues.
Are there engines which would play the opening moves better than these "big 3"?
Game 2, I'm strongly skeptical if the end position was a won ending. I think it was very likely a draw.
(It says win rule. But I still think it might be the wrong verdict.)
editing now:
maybe end position was lost for white, but it was still a drawn ending a few moves back, perhaps. White could have blockaded any win with his king and bishop.
S.Taylor wrote:Game 1 of stage 2 was very nice.
The games between the big 3 are 13, 62, and 81.
I'll be very much looking forward to see their openings and how it continues.
Are there engines which would play the opening moves better than these "big 3"?
Chiron already beat houdini with black in game 10 with 1.d4 d5 2.Nf3 a6
It is funny that houdini cannot punish black for 2...a6 that is against the chess principles.
Houdini blundered in the following position by 76.b5 but the main question is what are houdini's mistakes that allowed chiron even to get equality.
Maybe after 1.d4 d5 2.Nf3 a6 3.c4 e6 4.cxd5 exd5 5.Bg5 is wrong and better is 5.Nc3
Here is the position when houdini blundered and lost.
[D]1q2r2k/Rp5p/2p3r1/Q2p1pn1/PP1P4/3NP1Pp/2R2P2/7K w - - 13 76
It seems that it will be possible to prove for all programs that they are not deterministic even only based on the pgn of this tournament
If I am not wrong
All the program except Houdini already played different openings when the opponent played the same moves and probably with more games we will see it also for houdini.
I wonder what happens if you play 1000 games with no opening book between the same 2 programs.
What is the probability that 2 random games are going to be equal in the first n moves for different values of n?