Author: Vincent Diepeveen
Date: 15:48:18 12/24/00
Go up one level in this thread
On December 24, 2000 at 15:25:19, Uri Blass wrote:
>On December 24, 2000 at 15:11:45, Bruce Moreland wrote:
>
>>On December 24, 2000 at 03:02:26, Uri Blass wrote:
>>
>>>On December 23, 2000 at 20:31:04, Joshua Lee wrote:
>>>
>>>>So far Fritz solves Position 2 but gandalf now at 1:11:35 doesn't why it is
>>>>saying depth of 6 is beyond me. i will re-run this
>>>
>>>Gandalf has no chance to solve position number 2.
>>>
>>>This position is not for it and I am almost sure (based on analysis with it)
>>>that Gandalf is not going to solve it in 24 hours on your hardware.
>>>
>>>I did not give Gandalf 24 hours but I pushed the position forward and I found
>>>that Gandalf cannot see the problem in some hours even some plies after the
>>>initial position.
>>>
>>>Uri
>>
>>Position 2 is easier than position 1, in my experience.
>
>It depends on the program.
>
>For Gandalf position number 1 is easier.
>
>It can see a positive score and not only the right main line in a few hours for
>position number 1 when it cannot see a positive score in some hours even 2 or 3
>plies after Rxc5(It means that it probably has no chance to find Rxc5 because it
>has no reason to change its mind from another drawing move Bf6).
>
>Uri
You find nolot #1 Nxh6 in seconds if your king safety is giving real
optimistic scores for a few pieces attacking opponent king.
Position #2 i find at 9 ply already if i turn on just a few
mate-thread extensions otherwise it can take several plies more
and depending upon whether you do checks in qsearch or get a
draw score soon you find it a lot sooner or later.
In general position #2 Rxc5 is easiest of them all with mate-thread
extensions,
Position #3 Nxg5 diep never found, i checked it out why and
Diep misses quite a bit
of tenths of pawns to call this a 'hard to find' position,
i mean compensating for a queen is not easy if inside your
prog a queen is basically 11.5 pawns worth. Perhaps progs
which give a queen 9 pawns and a rook 5 pawns worth might
find it, but i would love to play those progs and get myself
a queen+passer and them 2 rooks.
Position #4 Nxe6
nowadays found by diep a bit sooner but still takes an hour
or so single cpu as it needs to see the queen get out of play
and blacks king get completely mated in center patterns.
Note the only problem to find this pos is that doing a different
move as Nxe6 is already getting a huge score, nearly positionally
up a pawn, otherwise Nxe6 would be found at quite a small depth.
Position #5 e5 is also in bs2830 testset if i remember well and it's
a position which diep never found also not when i ran it for
a few days. The line to find is just too deep for it before
its score is higher as root score.
Position #6 axb5 i always skip as i myself never could proof axb5
to be a correct sacrafice; correct meaning here that
alternatives are not better. axb5 is a kind of Shirov blitz
swindle which is more speculating at the fact that the
opponent is making big mistakes as speculating at the fact
that it is objectively best.
Position #7 Rxd8 was a big problem in the past, but with improved kingsafety
and especially improved endgame nowadays it's simply demanding
a bit of depth before it gets found. Nowadays pretty quick, but
a bit more as a minute of course.
Position #8 Bxh7 is a tactical shot where in the end programs need to
see Rxf7 Kxf7 Qh7+ king away and Qxc7 winning queen. If someone
by accident is doing that in qsearch you solve this at tournament
depths like 13 ply, but as far as i know no prog is doing
this in qsearch, neither do i in DIEP, so i do not solve it
with DIEP within an hour as it needs huge depth to see it.
Position #9 Ng5+ i never could proof that to be best move actually,
neither did Diep so far. So i have no clue why it is missing it
or what the problem lines for the engine is.
position #10 Rxf7 was never a problem for DIEP. Some versions need
a ply more to find it other versions a ply less.
basically if your evaluation knows pins and how to stop passers
you find it at tournament time then easily. Note that
queen=11.5 is quite helpful here too.
position #11 Rxh6 always was a problem for diep to find within a minute
because i used in past some dubious pruning and i hardly
extend passers, but i think some progs like schach and nimzo
find this at lightspeed.
Basically give away a queen, then push passers and with
a check in qsearch you mate opponent. Current version i don't
know whether it finds it real fast but it always found it
within the hour, with some luck nowadays at tournament level.
Vincent
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.