Author: Ulysses Omycron
Date: 23:08:59 01/02/05
Go up one level in this thread
>Here is an example
>
>3r2k1/p4ppp/5n2/3p4/3R4/1P3KP1/P1r1NP2/R7 b - - 0 27
>
>Analyze the position to depth 15 in 3 best mode and report the time that you
>need to do it.
1... g5 2. Nc1 g4+ 3. Kg2 Ne4 4. Nd3 Rd2 5. Rf1 Rxa2 6. Nb4 Rd2 7. Rxd2 Kf8 8.
Nxd5 {[%eval -266,15]}
It doesn't show the time, but it's around 45 seconds... If you stop it as it
gets the best move at 15 Plyes... Otherwise you are right the time needed is a
lot more, but you get the other 2 moves analysed in detail, so if it turns out
that one of them is the best, 3 best mode may find it while 1 won't. Yet again,
I don't care about multiple lines, I care about DPA (Even when DPA uses them).
>
>Later analyze the same position in one best mode and report the time that you
>need it.
Analysis by Shredder 8:
[Snip]
1...g5 2.Rc1 Rxa2 3.Nc3 Rb2 4.b4 g4+ 5.Kg2 Rc8 6.Ne2 Rxc1
-+ (-3.06) Depth: 15/32 00:00:14 2640kN
That's Ply 15 in 14 seconds.
Ok, you are right, it takes more time, but you get g5 all the way, you may tryed
ply 14 or ply 13... It doesn't matter how much time it takes it matters to find
the best move.
I wonder if I did something wrong, but Shredder 8 gives very different results
if you imput 3 best moves (1... h5 2. Rad1 Rxa2 3. Nc3 Ra3 4. Nxd5 Rxb3+ 5. Kf4
Rxd5 6. Rxd5 Nxd5+ 7. Rxd5 Rb4+ 8. Ke5 f6+ 9. Kf5 Rb7 {[%eval -219,14]}) 1...
h6 2. Nf4 g5 3. Nd3 h5 4. Nb4 Rc3+ 5. Ke2 a5 6. Nd3 Ne4 7. Rh1 Rc2+ 8. Ke3
{[%eval -190,14]}) than the Only best move, The 1 move is faster and better, but
if the test suite says one move is the best and the computer won't play it, it's
useful to know what score it gives (Without having to move it yourself).
>
>Take care that the program forget previous analysis
>
By your logic, we can feed the Hash doing a 1 move search then the 3 move search
will be faster. Making the program forget previous analysis hurts the time
needed for it, explain further why would you like the search to take more time
than normal (Since I can fill the hash so DPA runs like a rocket) besides
proving that 1 move is better
>I found that Fritz need some minuts in 3 best mode to get depth 15 when it needs
>only one minute in one best mode.
Maybe Fritz is much slower than Shredder 8, but the time comparison remains true
>>best may save time, but may be risky (That's why I don't use the total time
>>option... Err... I think in some point I contradicted myself).
>
>It is more common to see the 3th best becoming best but seeing one of the worst
>move become the best is not very rare(at least in test suites)
>
>Here is one example:
>
>Fritz8 in 9 best mode cannot see the move Rxh2 at depth 11 as one of the best 9
>moves at depth 11 or depth 12
>
>r2q3r/2pkb1p1/p2p1n2/4p1p1/Pp2P1P1/1QP5/1P1P2PP/RNB2RK1 b - - bm Rxh2
>
>
>Analysis at depth 11:
>
>New game,
>r2q3r/2pkb1p1/p2p1n2/4p1p1/Pp2P1P1/1QP5/1P1P2PP/RNB2RK1 b - - 0 1
>
>Analysis by Fritz 8:
>
>1. ² (0.44): 1...Nxg4 2.h3 Nf6 3.d3 g4 4.hxg4 bxc3 5.Qxc3 Qb8 6.Be3 Nxg4 7.Bf2
>2. ² (0.50): 1...Qg8 2.Qd1 Qh7 3.h3 Qxe4 4.d3 Qb7 5.cxb4 Qxb4 6.Rf2 Kc8 7.Bxg5
>3. ² (0.50): 1...Qb8 2.d4 bxc3 3.Qxc3 Nxg4
>4. ± (0.72): 1...bxc3 2.Nxc3 Qb8 3.Qd1 Qa7+ 4.Kh1 Rh4 5.g3 Rh3 6.Qf3 Rah8 7.Rf2
>Kc8 8.d3
>5. ± (0.84): 1...Kc8 2.d4 d5 3.exd5 exd4 4.Bxg5 Nxg4 5.Bf4 g5 6.Bg3 Ne3
>6. ± (0.94): 1...Nxe4 2.cxb4 Qg8 3.Qf3 Nf6 4.d3 Qd5 5.Qe2 Kc8 6.Be3 Rb8
>7. ± (0.94): 1...d5 2.exd5 Kc8 3.Rf5 g6 4.Rxg5 bxc3 5.Rxg6 Bc5+ 6.Kh1 cxd2
>7.Nxd2 Nxd5
>8. ± (1.06): 1...Rb8 2.Qc4 Nxg4 3.h3 Nf6 4.d4 g4 5.dxe5 dxe5 6.Rd1+ Bd6 7.Bg5
>gxh3 8.Qxa6
>9. ± (1.19): 1...Qe8
>
>(, MyTown 03.01.2005)
>
>
>Analysis of 9 best moves at depth 12
>New game,
>r2q3r/2pkb1p1/p2p1n2/4p1p1/Pp2P1P1/1QP5/1P1P2PP/RNB2RK1 b - - 0 1
>
>Analysis by Fritz 8:
>
>1. = (0.22): 1...Nxg4 2.h3 Nf6 3.d3 g4 4.hxg4 bxc3 5.Qxc3 Qb8 6.Rf3 Nxg4 7.d4
>Kc8
>2. ² (0.50): 1...Qg8 2.Qd1 Qh7 3.h3 Qxe4 4.d3 Qb7 5.Bxg5 Kc8 6.Qb3 Qd5 7.Qxd5
>Nxd5 8.Bxe7
>3. ² (0.59): 1...Qb8 2.Qd1 Nxe4 3.d4 exd4 4.cxd4 d5 5.Nd2 c5 6.h3 Nxd2 7.Bxd2
>Kc8 8.Rc1
>4. ± (0.72): 1...bxc3 2.Nxc3 Qb8 3.Qd1 Qa7+ 4.Kh1 Rh4 5.g3 Rh3 6.Qf3 Rah8 7.Rf2
>Kc6 8.d3 Nxg4
>5. ± (0.75): 1...Kc8 2.d4 d5 3.exd5 exd4 4.Bxg5 Nxg4 5.Bf4 g5 6.Qd1 gxf4 7.Qxg4+
>Kb7 8.cxd4
>6. ± (0.75): 1...d5 2.exd5 Kc8 3.d4 exd4 4.Bxg5 Nxg4 5.Bf4 g5 6.Qd1 gxf4 7.Qxg4+
>Kb7 8.cxd4
>7. ± (0.78): 1...Nxe4 2.cxb4 Qg8 3.Qd3 Nf6 4.h3 Qd5 5.Qf5+ Kc6
>8. ± (0.91): 1...Rb8 2.Qd1 Nxe4 3.d4 Qg8 4.Nd2 Qh7 5.Nxe4 Qxe4 6.Re1 Qh7 7.h3
>Kc8 8.Qe2 Kb7
>9. ± (1.06): 1...Qe8 2.h3 Qg8 3.Qd1 Kc8 4.cxb4 d5 5.d4 Bxb4 6.Be3 c5 7.exd5
>
>(, MyTown 03.01.2005)
>
Who was talking about multiple lines?... Oh yeah, chandler yergin, you proved
both us wrong about Multiple lines being worse than Only best line, If you burst
my bubble about DPA, I'm going to forget about computer chess at all and start
training myself in chess tourneys and so...
>
>The problem is that often it is not clear what is the best move so the question
>if you failed is going to be open question.
>
Do you mean we cannot discuss about how to know the best move on a given
position because the best move will remain unknown till god tell us what is the
best move?
I don't think so, there SHOULD be a best move, just think about tablebases,
isn't mate in 7 better than mate in 8 and so? so mate in 7 is the best move?
We just need to find something similar for a position without needing to analyse
all possible variations.
(In your example, g5 is the best move without a doubt, it treatens g4+ so Rc1 is
the only option, I made a PDA, and other moves (Like Nf4, h5, h6) are about
50-60 points weaker).
>If we could know what is the best move we could solve chess.
>Most positions are not positions when there is one winning move that you need to
>find.
>
I say the winning move from the starting position is e4 and I have yet to be
proven wrong.
>You may be better in finding moves when you know that there is something to find
>but worse than the computer in finding better moves when there is no decisive
>move to find.
Huh, I believe you though I don't understand you... I think all positions have a
decisive move to find...
>>level], but to know what is the best move available in a position after a Deep
>>Postion Analysis).
>
>I need to see some correspondence games when PDA beats not using PDA first.
>
We can have one by email, do you want White or Black, how many moves per day? Or
how many days per move?
(I bet you're gonna say you don't have time for this).
>>But it may turn out that d4 e6 turns out to be worse than e4 e6, and after
>>another DPA the petroff is more dangerous than e4 e6 making it to return to d4.
>>DPA is pretty dynamic, just don't use it once and say it's bad.
>
>The program also may analyze petroff later if it finds that 1.e4 e6 is not so
>bad for white.
>
>The point is that if 1.e4 e6 is better for black than 1.d4 than the exact value
>of petroff is not important.
>
>PDA does not know it and will continue to analyze petroff even if the score
>after 1.e4 e6 is better for black than the score after 1.d4
>
>Uri
In DPA you have total control of the moves being analysed, you may add a ? to a
move/variation and it will skip it, aside from everything else, it's very
flexible, and you can reorder the variations at will so you stop the DPA at any
point when everything you wanted analysed is so... But let's stop the opening
examples, you gave me a position didn't you, let's DPA it :):
[d]3r2k1/p4ppp/5n2/3p4/3R4/1P3KP1/P1r1NP2/R7 b - - 0 27
DPA says g5!! after checking the most important variations and it's worth -2.75
to -3.25.
To get a more exact score I'd need to make more DPAs, but it's very hard for g5
to fall, I'm 96% sure g5 is the best move.
Besides, the position you gave me it's a win for black (It seems), so black can
win even when not playing perfectly.
Some may think that when say "Solving chess" I mean to make the perfect move
each time, it's not. What it means it's you'll play the moves that avoid drawing
or losing, you have no other option than winning. With black it seems hard to
win, but the right to move seems to give white a decisive advantage (I'm talking
about the starting position! Not the example!).
I've just saw you replied to your own topic, let's see how I handle the other
position :)
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.