Author: Dan Homan
Date: 13:30:13 02/22/99
Go up one level in this thread
On February 22, 1999 at 13:44:42, Bruce Moreland wrote: > >On February 22, 1999 at 08:14:05, Dan Homan wrote: > >>On February 22, 1999 at 04:09:49, Will Singleton wrote: >> >>>New Feature >>> >>>I thought I'd post a position each week, taken from member submissions >>>from the current week's play. Recently on the CCC board we've had a few >>>positions posted, and these have helped me track down some problems with >>>my code. Probably just chance, but I think these things are helpful >>>for programmers in that they can see how their results differ from others, >>>and figure out why. >>> >>>So members, please send me interesting games that your programs play (or >>>just the positions), and perhaps we can all benefit from the analyses. >>> >>>To start the ball rolling, here's a position from the game Amateur-LambChop, >>>played on ICC yesterday. For some reason, Amateur declined the easy move, >>>which would have won quickly, and played a tricky move that led to a loss. >>>On subsequent analysis, it took Amateur 9 ply to see the right move. During >>>the actual game (blitz), it only searched about 7 ply. I'm thinking this >>>might be some bug or other weirdness. Does your program have any trouble >>>finding the right move quickly? >>> >>>7k/2pn1Q2/2r4p/p1qpp3/Pp1b3N/1P1P2PP/2P1R2B/7K w - - >> >>Nice problem. I'll have to look at it in more detail later, but here >>are EXchess' results >> >>EXchess likes Qxd7 with a score of +3.3 until ply 6 where it switches >>to Nf5. The score Nf5 climbs to +4.1 in ply 7 and 8. In ply 9, Nf5 >>fails low with a score of 0.6 and EXchess switches back to Qxd7 with a >>score of 3.8. Qxd7 fails high at ply 10.... I haven't run it deeper >>yet. Looks very interesting. > >This is another interesting FEN. > >7k/2pn1Q2/2r4p/p1qppN2/Pp1b4/1P1P2PP/2P1R2B/7K b - - 2 51 > >This is the position after 51. Nf5 in the game. The problem here is to figure >out how long until your program thinks that white's advantage is much reduced. > >bruce EXchess sees the big score change at ply 8. Here are the results from a rather slow machine (approx. P133 speed) which is running at half speed due to a large astronomical calculation: 12 seconds to fail high, 20 seconds to resolve the score change. (~ 1-2 secs on a PII 400). 3. -3.43 0 2259 1. ... Qf8 2. Qxd7 Qc5 3. -3.43 0 2394 1. ... Qf8 2. Qxd7 Qc5 4. -- 4. -3.70 0 5537 1. ... Qf8 2. Qxd7 Rc5 3. Qe6 4. -3.70 0 6364 1. ... Qf8 2. Qxd7 Rc5 3. Qe6 5. -- 5. -4.01 0 16315 1. ... Qf8 2. Qxd7 Rc5 3. Qe6 Kh7 5. -4.01 1 19748 1. ... Qf8 2. Qxd7 Rc5 3. Qe6 Kh7 6. -- 6. -4.40 3 45397 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Ra6 4. Nxd4 6. -4.40 3 55103 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Ra6 4. Nxd4 7. -4.31 6 84612 1. ... Qf8 2. Qxd7 Rg6 3. Nxd4 Qf1+ 4. Bg1 exd4 5. Re8+ Rg8 6. Rxg8+ Kxg8 7. Qxd5+ Kh7 7. -4.31 8 111986 1. ... Qf8 2. Qxd7 Rg6 3. Nxd4 Qf1+ 4. Bg1 exd4 5. Re8+ Rg8 6. Rxg8+ Kxg8 7. Qxd5+ Kh7 8. ++ 12 166119 1. ... Qf8 8. -0.76 20 253475 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Qxf5 4. Bg1 Rf6 5. Bxd4 8. -0.76 22 291931 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Qxf5 4. Bg1 Rf6 5. Bxd4 9. -0.79 34 442201 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Qxf5 4. Rg2 Qf8 5. Bg1 Bxg1 9. -0.79 41 524971 1. ... Qf8 2. Qxd7 Rd6 3. Qxc7 Qxf5 4. Rg2 Qf8 5. Bg1 Bxg1 I always find it interesting when EXchess shows a long line early on with the wrong score/move but the right score/move has a much shorter pv. I tend to chaulk this up to overcoming null-move failures or a terminated pv due to a hash table hit. - Dan
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.