Page 11 of 16
Re: IvanHoe compiles testing
Posted: Sat Jan 01, 2011 8:54 pm
by ernest
Tough to follow!!!!!!
Re: IvanHoe compiles testing
Posted: Sun Jan 02, 2011 1:28 am
by kingliveson
Here are 2 matches between latest IvanHoe and Houdini; extra long time control and 2 GB hash with IvanHoe using RobboBases.
I like to test this position which really favors Black to see how each engine responds. It's a tennis type of position where Black is supposed to hold serve. Houdini's Bishop pair attack against 2 Rooks and a exposed King was nice.
Game 1:
Game 2:
Look at the difference in speed.
Re: IvanHoe compiles testing
Posted: Sun Jan 02, 2011 11:29 am
by Robert Houdart
kingliveson wrote:Houdini's Bishop pair attack against 2 Rooks and a exposed King was nice.
...
Look at the difference in speed.
The second game is beautiful indeed.
I'm not that much impressed by the node speeds (Crafty has even higher node speeds...), I'm struck by White being completely unaware of what is coming even as late as move 33. Rg2
{-0.26/26 66}, whereas for Black the game was nearly winning as early as move 27... Re8
{-0.93/25 273}.
Re: IvanHoe compiles testing
Posted: Mon Jan 03, 2011 2:53 am
by kingliveson
Rybka vs IvanHoe, Blitz 1m
1 IvanHoe 9.47b x64 +1 +118/=266/-116 50.20% 251.0/500
2 Deep Rybka 4 SSE42 x64 -1 +116/=266/-118 49.80% 249.0/500
† 250 Positions (
test suite)
† EGTB Off
† Ponder Off
† 128 MB Hash
† Phenom II 940
† 4 Cores
† 1+0 Time Control
Game data available: http://chess.cygnitec.com/pgn/
Re: IvanHoe compiles testing
Posted: Sun Jan 09, 2011 5:25 am
by kingliveson
256 MB hash, 4 cores, 3-4-5 tb on, ponder off, 10 mins, 5 positions...
Re: IvanHoe 0B.01.08
Posted: Mon Jan 10, 2011 12:21 am
by kingliveson
Rounds 1-10:
Round 10 is very interesting because it was lost, but IvanHoe pulled magic and drew. Not quite, but this is a game where tablebases advantage is clear.
Re: IvanHoe compiles testing
Posted: Mon Jan 10, 2011 6:04 am
by BB+
70... Kg6? only draws, while 70... Kf7! is the only move to win.
The 6-piece RobboBlockedTripleBases (or whatever these are called) are quite useful here.
70... Kf7 71. Nc7 Ra3!, and I guess the Black king on f7 is closer to the action. 72. Nxa6 Rxa5, and it's a root 6-piece position (with Blocked bases).
Re: IvanHoe compiles testing
Posted: Mon Jan 10, 2011 8:14 am
by BB+
Game 1: two bishops and an open-ish centre beat an off-side knight. Game 2 was a Father-special, with 200+ moves of shuffling. Game 3 shows that you can still lose in 20 moves.
The White knight on d5 was really strong in Game 4, and the centre control just divides the Black army in two it seems. Game 5 shows that you can still lose in 15 moves.
I kept thinking Black might save Game 6 (with a perpetual check, after the 2B vs R part had passed into queens and pawns). Game7 looked to be one those "pawn-up but no win" games that some GMs love to defend. White's break-thru almost worked in Game8, except his own king was too airy. Game 9 was a classic "Kramnik-style" win, where you have to get a microscope to figure out where exactly Black lost the thread. For Game 10, the exchange was worth more than the pawns, particularly after the queen exchange relieved Black of the "redundant" majors. I'm still not sure why Black didn't simply play 64... Rxg4.
Re: IvanHoe compiles testing
Posted: Mon Jan 10, 2011 1:14 pm
by kingliveson
BB+ wrote:Game 1: two bishops and an open-ish centre beat an off-side knight. Game 2 was a Father-special, with 200+ moves of shuffling. Game 3 shows that you can still lose in 20 moves.
The White knight on d5 was really strong in Game 4, and the centre control just divides the Black army in two it seems. Game 5 shows that you can still lose in 15 moves.
I kept thinking Black might save Game 6 (with a perpetual check, after the 2B vs R part had passed into queens and pawns). Game7 looked to be one those "pawn-up but no win" games that some GMs love to defend. White's break-thru almost worked in Game8, except his own king was too airy. Game 9 was a classic "Kramnik-style" win, where you have to get a microscope to figure out where exactly Black lost the thread. For Game 10, the exchange was worth more than the pawns, particularly after the queen exchange relieved Black of the "redundant" majors. I'm still not sure why Black didn't simply play 64... Rxg4.
Thanks for these analysis.
Rounds 11-20:
Re: IvanHoe compiles testing
Posted: Mon Jan 10, 2011 8:52 pm
by kingliveson
Rounds 21-30:
Tournament had to be stopped.