Author: Slater Wold
Date: 22:31:22 09/18/01
Go up one level in this thread
On September 19, 2001 at 01:03:58, Uri Blass wrote:
>On September 18, 2001 at 20:14:15, Slater Wold wrote:
>
>>On September 18, 2001 at 13:44:33, Janosch Zwerensky wrote:
>>
>>>Hi everyone,
>>>
>>>I don't expect that currently any program can solve one of these problems given
>>>less than weeks of reflection time, but I'd like to know anyway what different
>>>progs say:
>>>
>>>[D] 6kr/8/rp1p1p1p/pPpPpPpP/P1P1P1P1/8/8/4K3 w - - 0 1
>>>
>>>It is very easy to see (for a human) that in this position 1. bxa6 is a bad
>>>idea. How long does your program take to understand this, too?
>>>
>>>Regards,
>>>Janosch.
>>
>>
>>Again, Crafty has solved this. And in good time:
>
>I see from another post that Dann Corbit's Crafty18.11 could not solve it in an
>hour inspite of the fact that it finished iteration 25.
>I wonder what is the reason for it.
>
>Is it possible that big hash tables helped so much or maybe there is a super
>linear improvment from parallel search in these positions?
>
>I suspect that a difference in hash tables is the reason because hash tables
>help mainly in relatively simple positions and positions when the pawn are
>blocked are relatively simple(there are a lot of repetitions in the search).
>
>Uri
Well, I believe there can be a few reasons. Hash being one of them. Also let's
look at the evals:
MINE
25 1:51 -15.00 1. bxa6 Kf7 2. Ke2 Ra8 3. Kf3 Rxa6
4. Ke3 Ra7 5. Kd3 Ke7 6. Ke3 Kd8 7.
Kd3 Kc8 8. Kc3 Kc7 <HT>
DANN'S
25-> 6:48 -8.79 1. bxa6 Rh7 2. Ke2 Ra7 3. Ke3 Rxa6
4. Kd3 b5 5. cxb5 Ra7 6. Kc4 Rc7 7.
Kd3 c4+ 8. Kc3 Kf7 9. Kc2 c3 10. b6
Rb7 <HT>
His Crafty hasn't (for whatever reason) seen this PV. Dann stated he was having
a huge fail low, and more than likely would have gotten it on the next ply.
SMP machines are weird. There is no rhyme or reason on where they decide to
branch. Also, I was using a HUGE amount of HASH, as I thought this position was
going to take longer than it did. I was using 192MB. Also, you can see the
huge difference in speed. It took me 1:51 to get to the same ply as Dann's at
6:48.
I will run *this* particular position, because of the pawns, again tonight. One
with 48MB hash, and one with 192MB of hash. We'll see which one solves it
first.
Slate
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.