Re: Please Analyze this game between Easy Level & Chrisf

Sanny wrote:

> Nice Analysis !!!

It will be a nice analysis if someone else checks it and agrees.
> 
>> Amazingly this position has previously occurred in 4 serious
>> matches arising out of odd Sicilian patterns. The moves are
>> transposed and all of them were hopelessly lost at this
>> point. It must have arisen in
> 
> So you have saved Game till 14 th Move? Have you played the same game
> earlier?

Apparently yes although I haven't looked for it yet. I expect some 
variant of poisoned pawn Sicilian ended up going that way. It looked 
vaguely familiar.
> 
>>> Is there any move wrongly played by Advance Level?
>>> Was the game won because of any weakness in GetClub engine?
>> Yes. It doesn't have sufficient selective look ahead down partially
>> forced lines (situations where one move is significantly better than
>> all the others). To be fair even the top engines struggled to see that
>> Qc7 doesn't really work at sensible match play time allocations and it
> 
> So GetClub Lost because of 7... Qc7 Move.

I think so yes. However, I ran Shredder10 without human guidance 
overnight to see if it could find the right line for itself. This is 
definitely one of those situations where an opening book is essential 
for the computer to cope with deep set piece opening traps in sharp 
lines. It cannot see deep enough under normal time controls.
> 
> How much point loss is shown for this move compared to 7.... Nd5. What
> would have been the score if Nd5 is played and what is the score given
> for 7... Qc7 by Shedder?

I ran it overnight after restarting the program and going to the 
position after whites move 7.bxB all score in centipawns

Initially I ran infinite analysis with 6 best lines displayed.
At ply 12 (30s) it had
Nd5  Qc7  Qa5  O-O  Ng8  Nc6   d5
107  157  303  341  345  356  356
(79) (142)
This morning I broadened the lines displayed to 18 to show the score of 
the ill fated Qc7 (which vanished from the top 6 overnight). It is 
interesting to note how from the same initial ocnditions (program 
restarted with a clean cache) the scores at the same ply are different. 
The additional exploration of the bad moves low down the list polluting 
the cache alter the score for the best moves by much more than I had 
expected!

It hadn't found the critical line when I gave up last night after 22 
minutes at ply 17 it slightly preferred Qc7 152 (143) to Nd5 159 (154) 
rated equal to within the noise (figures in brackets from second run).
This morning it was at ply 21 after 9 hours (scores here for ply 20)

Nd5  Ng8  Ne4  O-O  Nc6  a6
162  317  327  355  384 384

Qc7 was nowhere to be seen and a quick probe suggests it scores 420 or 
so with this depth of search. I don't know when Shredder finally grocked 
this position, probably at 2-4hours. But it is clear that it could never 
do so under tournament conditions whereas a human might. Especially one 
familiar with the motifs of this type of opening trap.

The position is an interesting test of engine strength. Anyone care to 
post the results of infinite analysis with their favourite engine on 
this position:

rnbqk2r/pp1p1ppp/4pn2/4P3/3N4/2P5/P1P2PPP/R1BQKB1R b KQkq - 0 7

Shredder must be going down some pretty wierd swapoff continuation lines 
in selective extensions 16/53 is about the most extreme ply/maxdepth 
ratios I have ever seen. 16/32 would be much more typical.

Regards,
Martin Brown
** Posted from http://www.teranews.com
**