Page 1 of 1
CM8000 vs. Bret Discussion
Posted: Mon Oct 18, 2004 2:55 pm
by tellymetwise
Hey all,
The game
CM8000 vs. Bret has started.
With This thread I invite people to discus this game.
The opening so far:
1.e4 c5 2.Ng1f3 d6 3.Bf1b5 Bc8d7 known as:
B52: Sicilian, Canal-Sokolsky (Rossolimo) Attack
The move 3. Bb5+ is known as "Bashing the Sicilian", though it avoids the Najdorf or Dragon lines, which are somewhat more complicated to compute.
For CM8000 it too will be more profitable to have less pieces on the board, artificialy increasing its calculation depth.
The end results for (known) games with the same opening yield a 54% chance of a draw, and an equal change for both white or black to win the game.
Posted: Tue Oct 19, 2004 10:32 pm
by gmiller
I figured a computer would do better with more pieces but less pawns on the board which makes for a tatical game. The computer can exhaustivley search several moves ahead, but after that a human's ability for form a long term plan takes the advantage so long as no tactical shots some into play.
Assuming Bret isn't going to read this thread until after the game is over, I'll post the computer's score and expected lines here throughout the game.
Posted: Tue Oct 19, 2004 10:38 pm
by gmiller
Just an FYI: The computer is searching about 700,000,000 positions every hour! In the past two hours it has searched 1,398,000,000 positions.
Posted: Wed Oct 20, 2004 8:28 am
by tellymetwise
CM8000 vs Bret
B52: Sicilian, Canal-Sokolsky (Rossolimo) Attack
1.e4 c5 2.Nf3 d6 3.Bb5+ Bd7 4.Bxd7+ Nxd7 5.d4
White has opted* for the easy open game, either an open center or an open queenside. (*Even though through statistics only.)
This will indeed lead to less pawns on the board, where the minor pieces (bishops) and rooks will get easier into play.
Software players indeed have a lot of difficulty ascerning the correct value of a pawn. Its value only being one point, but its position can be priceless.
Greg, CM8000 is using up memory at the rate of 700.000 positions per hour? How much is that? 64 sqaures + piece info + move info = 1Gb an hour?
Posted: Wed Oct 20, 2004 2:53 pm
by gmiller
It only evaluates 700,000,000 positions, it doesn't store all of them. It only remembers the interesting ones, and even then it only needs to remember which piece moved, not the entire position. I don't remember exactly how much RAM it used, not much, I think it was less than 1Mb. But that's just Chessmaster, other engines like Fritz you can tell it how much RAM to use which helps realize when it's evaluating a position it's already seen before.
Posted: Sat Oct 23, 2004 2:39 pm
by sirthursday
gmiller wrote: But that's just Chessmaster, other engines like Fritz you can tell it how much RAM to use which helps realize when it's evaluating a position it's already seen before.
You can do this on Chessmaster as well... at least, you can on CM9000...I don't know about CM8000.
Posted: Thu Oct 28, 2004 8:29 am
by tellymetwise
With 10.Qb4 (
View Board) CM8000 has now clearly shown its intention of creating an passed a-file pawn, with the attack focused on the queen-side of the board.
Though white might have rushed it too much. Black now has several options open to defend but also has a nice counter attack with a combination making his b7 pawn somewhat too hot to touch right now.
White's other option besides 10.Qd4b4 would have been the more passive Qc4 or even Qd3 keeping a more tight control, by delaying the queenside rush.
The game itself won't suffer from it as pieces might start flighing over the board in the upcomming clash between CM8000 and Bret, which sure will be more fun then a slow paced tactical play.
Posted: Thu Oct 28, 2004 4:44 pm
by gmiller
CM's previous best move was Qa4, and Qb4 only beat it out after about 8 hours of searching. The current score is 0.18.
Posted: Thu Oct 28, 2004 6:35 pm
by tellymetwise
How many moves ahead did CM get, Greg?
Posted: Thu Oct 28, 2004 9:18 pm
by gmiller
It completley evaluated every move to a depth of 9/15, whatever that means:?: . The lines it predicts go out to 9 moves
Posted: Fri Oct 29, 2004 1:13 pm
by tellymetwise
Beats me
Posted: Fri Oct 29, 2004 6:32 pm
by tellymetwise
I believe it means it checked upto the 9th best move at 15th ply (half move)
(or upto the 15th best move at 9th ply
)
Posted: Fri Nov 05, 2004 5:56 am
by tellymetwise
Posted: Sun Nov 07, 2004 9:17 am
by simbathelion
13) ... Qxc2 14) Rd2 Qxe4 15) Rd4 Qe6 16) Ng5 Qe5 17) f4
This leads to the black queen being trapped
. That's why black can't play 13) ... Qxc2
Best for black would probably be 13) ... Nxd5 14) exd5 Ra5, as Bret has successfully found
.
Posted: Sun Nov 07, 2004 12:24 pm
by tellymetwise
You've got that right again
Posted: Sun Nov 07, 2004 12:40 pm
by tellymetwise
Posted: Sun Nov 07, 2004 8:47 pm
by gmiller
Chessmaster's current score for this possition (after 10 hours of thinking) is -0.03. It's expected line is:
15.Qb4 a3 16.bxa3 b6 17.Bd4 Rxd5 18.c4 Ra5 19.Bxg7 Kxg7 20.Rfe1 Nf6
Posted: Wed Nov 10, 2004 7:08 pm
by tellymetwise
Posted: Tue Nov 16, 2004 7:51 pm
by tellymetwise
Position after move 19... Nxd5
View Board
With this move, Nxd5, Bret gave away some of the initiative gained. CM8000 has enough now to play for the draw, CM8000 only option. The most simple way is repetitive check by both rooks on the 7th rank. The ardious route is the slowly exchanging pieces ending in a pawn endgame.
20.Qxb7 Qxb7 21.Rxb7
View Board, attacking the weak e5 pawn.
black has two reasonable option to go from here,
21... Nxe3 22.fxe3 Rxa3 23.Rxe7 etc,
View Board, leading to a 2 rook endgame.
21... e5 22.Rd3 Nxe3 23.fxe3 (Rxe3?? Rxa3 Rxa3 Rxa3) 23... Rxa3 24.Rxd6
View Board, and CM8000 can gain the 7th rank.
Bret could have kept pressure on the game by playing 19... Qxc2 forcing CM8000 to keep up with him, but most probably this would have only been the scenic route, for the "tourists", towards draw.
at move 19.
View Board
19... Qxc2 20.Qxb7 Qxa2 21.Rd2 Qxa3 22.Bd4 (Qxe7? Nxd5!) 22... Re8 etc.
View Board
Posted: Thu Nov 18, 2004 5:21 pm
by gmiller
Chessmaster accepted a draw immediatley. All of the scores and lines Chessmaster examined are here:
Here