Endgame Test Position g6!

[Martin Hertz]

Found here: https://www.schachcomputer.info/forum/s ... hp?p=70344

Surely not that easy for the most engines, to find g6 in the standard playing mode:

[d]4b3/7p/4k3/B4pPP/5K2/P4P2/8/8 w - -

[Guenther]

Found here: https://www.schachcomputer.info/forum/s ... hp?p=70344

Surely not that easy for the most engines, to find g6 in the standard playing mode:

[d]4b3/7p/4k3/B4pPP/5K2/P4P2/8/8 w - -

Strange I see it immediately as Human, do they (programs) really have problems with it?

[Ras]

Strange I see it immediately as Human, do they (programs) really have problems with it?

DeepShredder13x64 with 1 GB hash and 6 cores takes 10 minutes to come up with g6, at depth 68/80 (full/selective).

[Martin Hertz]

Strange I see it immediately as Human, do they (programs) really have problems with it?

Yes, they want to play h6 instead and need very high depth to see g6.

[MikeB]

Found here: https://www.schachcomputer.info/forum/s ... hp?p=70344

Surely not that easy for the most engines, to find g6 in the standard playing mode:

[d]4b3/7p/4k3/B4pPP/5K2/P4P2/8/8 w - -

Strange I see it immediately as Human, do they (programs) really have problems with it?[/quote

SF-McB 3.0 sees it in 24 seconds on my machine:

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 48	+3.60!	1.17G	0:36.23	g6! 
 48	+3.08!	992.8M	0:31.06	g6! 
 48	+2.69!	897.3M	0:28.19	g6! 
 48	+2.39!	828.6M	0:26.01	g6! 
 48	+2.16!	789.0M	0:24.76	g6! 
 47	+1.94 	731.1M	0:22.92	h6 Bh5 a4 Bg6 Bb6 Kd5 a5 Kc6 Ke5 Kb5 f4 Ka6 Kf6 Kb5 Bc7 Kc6 Be5 Kb5 Bc3 Kc6 Bd2 Kb5 Bb4 Kc6 Be1 Kb5 Ke6 Ka6 Bd2 Kb5 Ke5 Kc6 Bc3 Kb5 Ke6 Ka6 Ke7 Kb7 Bd2 Ka6 Kd6 Bf7 Ke5 Bg6 Be1 Kb5 Bc3 Kc6 Kf6 Kc5 a6 Kb6 a7 Kxa7 
 46	+1.94 	510.1M	0:15.91	h6 Bh5 a4 Bg6 Bb6 Kd5 a5 Kc6 Ke5 Kb5 Kf6 Ka6 Ke6 Kb5 Bd8 Ka6 Bc7 Kb7 Be5 Ka6 Bc3 Ka7 f4 Kb7 Bb4 Ka6 Bd2 Kb5 Kf6 Kc5 Be3+ Kb5 Bb6 Ka6 Bc7 Kb5 Ke5 Ka6 Ke6 Kb7 Bd6 Kc6 Ke5 Kb5 Bb4 Ka6 Bd2 Ka7 Bc3 Ka6 Kf6 Kb5 

and Sf-McB 3.1a (dev) sees it 10 seconds on my machine:

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 43	+7.51!	713.4M	0:21.56	g6! 
 43	+6.15!	609.9M	0:18.60	g6! 
 43	+5.08!	497.3M	0:15.35	g6! 
 43	+4.24!	432.3M	0:13.42	g6! 
 43	+3.58!	388.8M	0:12.02	g6! 
 43	+3.07!	361.1M	0:11.03	g6! 
 43	+2.67!	332.9M	0:10.18	g6! 
 43	+2.37!	315.8M	0:09.68	g6! 
 43	+2.15!	308.3M	0:09.44	g6! 
 42	+1.92 	243.7M	0:07.51	h6 Bh5 Bb4 Kd5 a4 Bg6 a5 Kc6 Be1 Kb5 Bc3 Kc6 Ke5 Kb5 f4 Kc6 Bd4 Kb5 Bb6 Kc6 Be3 Kb5 Bd2 Kc6 Bb4 Kb5 Kf6 Ka6 Bc3 Kb5 Bd2 Kc5 Ke7 Kb5 Ke6 Kc6 Ke5 Kb5 Bb4 Ka6 Bc3 Kb5 Kf6 Ka6 Bd2 Kb5 
 41	+1.92 	184.9M	0:05.78	h6 Bh5 Bb4 Kd5 a4 Bg6 a5 Kc6 Be1 Kb5 Bc3 Kc6 Ke5 Kb5 f4 Ka6 Bd2 Ka7 Be1 Ka6 Kf6 Kb7 Bd2 Ka6 Ke7 Kb5 Bc3 Bh5 Ke6 Bg6 Kf6 Ka6 Bd2 Kb5 Ke7 Kc5 Bc3 Kb5 Ke6 Kc6 Ke5 Kb5 Bd2 Ka6 Bb4 
 40	+1.92 	145.3M	0:04.64	h6 Bh5 Bb4 Kd5 a4 Bg6 a5 Kc6 Be1 Kb5 Bc3 Kc6 Ke5 Kb5 f4 Ka6 Bd2 Kb5 Be1 Kc5 Bc3 Kb5 Bd2 Kc5 Kf6 Kc6 Be3 Kb7 Ke5 Ka6 Bb6 Kb5 Bd8 Ka6 Ke6 Kb7 Bb6 Kc6 Be3 Kb5 Bd2 Kc5 Bc3 Kb5 a6 Kxa6 

it appears normal SF has problems (t depth 110 and still searching) with this positions as with all the other engines I tested ...

[ernest]

I wonder if the best programs have the notion of "good bishop" for the promotion on column a and h. Here, it is essential !

[kasinp]

Found here: https://www.schachcomputer.info/forum/s ... hp?p=70344

Surely not that easy for the most engines, to find g6 in the standard playing mode:

[d]4b3/7p/4k3/B4pPP/5K2/P4P2/8/8 w - -

Thanks for posting, great position. I think two distant passers (key motive in this) is still an underestimated evaluation term in opposite-color bishop endings.

Tried many engines, and the old trusty Zappa Mexico gets it, immediately too... that said the evaluation is not convincing deep into the search.

PK

[peter]

Hi!

Found here: https://www.schachcomputer.info/forum/s ... hp?p=70344

Surely not that easy for the most engines, to find g6 in the standard playing mode:

[d]4b3/7p/4k3/B4pPP/5K2/P4P2/8/8 w - -

As for H6, it depends on SMP- luck much obviously.

On 24 threads I had 2 tries under 1 minute (25 and 50") with full Syzygys, one with 50" without tbs, and with and without tbs I broke off after 3 minutes each once.

Andscacs seems to differ much too as for TTS here, again 1 try in about 50", best one 13", third one after 2 minutes still no g6.

CorChess was 3 times under 1 minute without tbs, from 30 t0 50",

[zenpawn]

I was proud of RookieMonster for choosing g6 at depth 5. Unfortunately, it eventually changes its mind to h6 at depth 17-24 (at which point I stopped it).

[MikeGL]

Critter 32 bit gets it before 3mins on my 9 year-old laptop, running at 2.0 GHz on 2 cores

Clearing hash before each run:
1st run: 00:57 seconds
2nd run: 02:02
3rd run: 00:40 seconds


Engine: Critter1.6 32bit
2 cores @ 2.0 Ghz
Hash: 256MB

02:02

Code: Select all

 24/46	02:02	 126.782.728	1.038.862	+1.70	g6 hxg6 h6 Kf7 Bc3 Kg8 Bd2 Kh7 Ke5 Bc6 Bg5 Be8 Kd6 Bb5 f4 Be8 Kc5 Kg8 Kb4 Bc6 a4 Bf3 a5 Be2 Kc5 Kf8 h7 Kg7 Kb6 Kxh7 a6 Kg7 a7 Bf3 Kc7 Kf7

00:40s

Code: Select all

 21/40	00:40	  41.965.338	1.026.875	+1.52	g6 hxg6 h6 Kf7 Ke5 Bc6 f4 g5 fxg5 Kg6 Bc3 Be8 Kf4 Bc6 Be5 Be8 Bd4 Ba4 Bb6 Bb3 Bd8 Ba4 Bf6 Bc2 Bc3 Ba4 Ba1 Be8 Bd4