Tough for Engines - Trapped Piece Detection & Zugzwang..

[Steve Maughan]

Mc Mad posted this endgame position on the general forum. It got lost in the thread and I think it's of interest to programmers so I'm re-posting here with some analysis:

[d] 5q2/1p1kp3/pP4p1/4B3/3N3K/1B6/P4P2/8 w - - 0 1

The position is really quite deep. The forced win of the Queen is 19 ply. Most of those moves would be regarded as quiet and most forms of selectivity would prune, or aggressively reduce. Most engines struggle, even when shown the first two moves.

From a engine / programming perspective I think the key position is after the following moves:

1. Be6+ Ke8 2. Nf5 gxf5 3. Kg5 f4 4. f3 Kd8

[D]3k1q2/1p2p3/pP2B3/4B1K1/5p2/5P2/P7/8 w - - 0 5

White is a knight down compared to the root position. It isn't zugzwang as black can play a5 with no immediate repercussions. The queen has a safe square to move to as it looks as if she can move to e8 (although it's mate in 1).So aggressive pruning is likely. I think this is difficult for engines to see past this position from the root.

I think only the engines with highly developed code for detecting trapped pieces and zugzwang are able to see it.

Any thoughts?

Steve

[gleperlier]

Nice idea for Monarch

[RoadWarrior]

5q2/1p1kp3/pP4p1/4B3/3N3K/1B6/P4P2/8 w - - 0 1

The position is really quite deep. The forced win of the Queen is 19 ply. Most of those moves would be regarded as quiet and most forms of selectivity would prune, or aggressively reduce. Most engines struggle, even when shown the first two moves.

From a engine / programming perspective I think the key position is after the following moves:

1. Be6+ Ke8 2. Nf5 gxf5 3. Kg5 f4 4. f3 Kd8

White is a knight down compared to the root position. It isn't zugzwang as black can play a5 with no immediate repercussions. The queen has a safe square to move to as it looks as if she can move to e8 (although it's mate in 1).So aggressive pruning is likely. I think this is difficult for engines to see past this position from the root.

I think only the engines with highly developed code for detecting trapped pieces and zugzwang are able to see it.

Any thoughts?

Steve

Moving on 2 ply from your initial position (i.e. Be6+, Ke8), Houdini 2.0 finds the win of the Black queen in 14 minutes. This is using 2 cores and 1 GB cache on a i7 2600K. I will now try it on your initial position.

[diep]

Steve, as a chessplayer i feel it's a great position.

However a small sidenote. With todays huge search depths you typically see that modern programs bust all sorts of positions. Is it relevant how fast someone wins?

So does it matter whether you win a position or does it matter that you win in the fastest possible time?

Todays software focuses upon searching something that is 'sufficient' deep enough. It seems to win game after game.

That's where this position also has a problem. If you do NOT play Be6 with the nice knight sacrafice after which there is a bunch of zugzwangs, then you still win this position with white of course, as 3 pieces is better than a queen and not a little. You got the luxury of 2 bishops here.

So in that sense for the modern software this position is not so relevant.