Author: Sune Larsson
Date: 15:46:51 12/27/01
Go up one level in this thread
On December 27, 2001 at 17:59:41, Dieter Buerssner wrote:
>On December 27, 2001 at 11:13:40, Sune Larsson wrote:
>
>> [D]8/3k2p1/4pp2/1p2PP2/1PpP2P1/8/4K3/8 w - - 0 35
>>
>> This position is from a game Rocha - Spraggett, 1999.
>> White is worse here and there are several ways for him to
>> lose this position. Hecht, in his commentaries, even thought
>> that the key move is the only way for white to make the draw.
>> This ending is tricky, so I'm not 100% sure that Hecht is correct.
>> Anyway, white has a simple and straight forward way to secure the draw.
>>
>> Theme - pawn ending: One protected passer can hold his own vs two connected.
>>
>>
>> Test: Programs with some of this knowledge should evaluate the position
>> after 1.d5! as 0.00 or close to it.
>
>This is an interesting sentence :-) I actually had looked at such positions, and
>found it not totally easy ... What I fear, is that such a rule can fail, or
>better formulated, that one can miss the restrictions under which it is valid.
>What I think of for example: Can the base of the protected pawn be attacked by
>another pawn. And of course the (at least to the human) obvious things, like:
>are both - the protected passer and the connected passers stopped right now? Are
>there more candidates for passers on the board?
>
>From an engine point of view, one could also ask, how useful such a rule would
>be. In this case, it seems Ke3 is also drawing - but I agree - not that
>convincing. However, when all other ideas would be losing, the engine might see
>this anyway by search. So, such a rule might be most useful when one trades into
>a pawn endgame. But now there might be not much search depth, so the rule would
>need to be pretty tight, to not heal one case and open another bigger trap. You
>are welcome, to make a suggestion, for all restrictions such a rule needs :-)
>
>I followed the lines you discussed with Miguel, and I allways get a draw
>position, as you said already. I did not allways get a draw score, but in KPPKPP
>with search depth 26 and only K moves in the PV, I found it pretty obvious. Also
>some lines yielding Q endings with one pawn up for black showed draw score after
>following the line.
>
>Here is what Yace thinks of the position:
/snip
> 819191465 1:20:37 -0.75 18. 35. Ke3 Ke7 36. Ke4 g6 37. exf6+ Kxf6 38. g5+
> Kxg5 39. fxe6 Kf6 40. d5 g5 41. Kd4 g4 42. Kc5
> c3 43. Kd6 c2 44. e7 c1=Q 45. e8=Q Qf4+ 46. Kd7
> Qxb4 47. Qe6+ Kg5 48. Qe5+ Kh4 49. d6 {-80}
Yes, this is another line I looked at. This 38.g5+ trick spoiled it all
for black. And it doesn't help black to first put his king on f7 and then
play g6, because now white has fxe6+ followed by d5+ and e6...
The queen ending above can't be won for black.
Sune
PS No general rule for these kind of endings today...;-)
>
>I wonder, how Crafty reaches such high search depth fast in this position
>(Pavel's post). Queens are seen pretty soon, which will increase the branching
>factor. Also, there are many pawn moves possible.
>
>Regards,
>Dieter
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.