Why computers need tablebases
Plug this position into your computer chess program, no tablebases allowed (white to move):
This problem (likely a composition) came up in CTB, and I plugged it into Fritz (I check all its KP endgame solutions as the program has many errors on these problems).
It's a draw, but it took Fritz over 10 minutes to find the solution. Then, it took it another 7 minutes to find the solution after it made black's best move. This is all quite a bit longer than I allow him to analyze a single position in the games I blundercheck with Fritz. Heck, that's longer than I let him analyze an entire game! Beware the horizon effect, J'adoube!
I'll put one of the two solutions in the comments on Saturday. Anyone have Rybka? How long before she finds the solution?
Did I say I was supposed to get a good night's sleep tonight. 4am still doing tactics. Oh well...
This problem (likely a composition) came up in CTB, and I plugged it into Fritz (I check all its KP endgame solutions as the program has many errors on these problems).
It's a draw, but it took Fritz over 10 minutes to find the solution. Then, it took it another 7 minutes to find the solution after it made black's best move. This is all quite a bit longer than I allow him to analyze a single position in the games I blundercheck with Fritz. Heck, that's longer than I let him analyze an entire game! Beware the horizon effect, J'adoube!
I'll put one of the two solutions in the comments on Saturday. Anyone have Rybka? How long before she finds the solution?
Did I say I was supposed to get a good night's sleep tonight. 4am still doing tactics. Oh well...
posted by Blue Devil Knight at 9/01/2007 03:54:00 AM
12 Comments:
I tested and found the following: Rybka 2.3.2mp w/tables 3.5 sec, w/o tables 5.5 sec. That's one of the reasons I'm a fan of Rybka--it plays a mean endgame. When I play a weaker engine, such as Crafty, I often end up winning a drawn position or drawing a lost position.
With respect to the puzzle, that's a tricky one! First, if the kings race to promote Black will win because Kf4-Kxf5-Kxf6-Kf7-f4-f5-f6-f7-f8 is 9 tempi while Kxa6-Kb5-a5-a4-a3-a2-a1 is only 7 tempi, with a won QvP ending. Second, if White tries to block/capture the a-pawn then Black has time to gobble up the kingside pawns and win. So I considered the game totally lost for White. But Rybka finds 1.h6 2.f4 3.Kf2 4.Kg3 5.Kh4 6.Kh5 7.h4. Basically, it's a stalemate cage that forces a draw.
Sleep is overrated... ;)
Keep up the good work!
-Chris
Likeforests: holy cow I need to get Rybka! You can play f4 first too.
I just checked Rybka web site and since it seems a 3.0 is due out in early 2008 I'll wait on the purchase and keep Fritz 9.
I guess I should title it why MY computer needs a tablebase.
This is way beyond current tablebases, it will be a long, long time before 10 piece tables are published.
As to Rybka, it is cheap, it really is worth it. It doesn't just play a little better than fritz 9, it generally beats well over 90% of the time.
It plays MUCH MUCH better at fast times, which translates extremely well to analysis.
It really is better to have 2 different kinds of engines during analysis (you can do this with the add kibitzer setting). It lets you focus on those positions that are the most controversial, and gives you world class help in working through those. I suspect the discount will be substantial for upgrades.
Apologies for going off topic...
I've implemented a slightly modified version of your wife's script in order to randomize the "Accolades" at BCC Weblog. As BDK might say, "Sweet!"
How fast is your computer? I have two 2.2GhZ processors and 2GB RAM. I am beginning to suspect that has more to do with it.
Fritz 5.3.2 solved this in 2.5s w/ or w/o TBs and Crafty solved it in 1s w/o TBs, which is the opposite of how these engines usually perform. I have Rybka set to play "solid" chess and Fritz set to play "tricky" chess. That gives me a good set of opponents to play when I'm trying to learn an endgame. And it may be why Fritz found this trick faster.
Note: 1.f4? seems to lose to 1...h6!
BDK,
Feel free to disregard the following unsolicited motherly advice : 8)
Don't waste too much of your tactical momentum on this technical crap.
Ride the tactical wave Bro,
Live Free,
Jim
Tak: it's part of CTB, so it's tactics as far as I'm concerned. :)
Likeforests: wow. This has a 1GHz processor, I wonder why yours is so much faster. This old thing has been acting strange lately...Maybe Fritz 5.2 is actually better for this problem!
DG: very cool. I really like the random blogroll action.
HBF: I really should buy it when I get the free funds.
LF: yes, you are right f4 loses...
While waiting for funds to free up. You simply should get these engines. For Free...
The Rybka 1.0 beta, an extremely strong engine.
http://www.rybkachess.com/free/Rybka10FreeDownload.zip
And
Fruit 2.3.1 Another extremely strong engine.
http://www.superchessengine.com/fruit.htm
Both are free, both some of the strongest engines available, and are perfect for being the "other" engine inside of fritz.
Post a Comment
<< Home