Stunning Chest!

[Jouni]

[D]5k2/8/8/1p5p/1P5P/8/4K3/8 w - -

Analysis by ChestUCI Ver.4.4:

1.Kd3 Ke7 2.Ke3 Ke6 3.Ke4 Kd6 4.Kd4 Kc6 5.Ke5 Kc7 6.Kf5 Kd7 7.Kg5 Ke6 8.Kxh5 Kf5 9.Kh6 Kf6 10.h5 Kf7 11.Kg5 Kg7 12.Kf5 Kh7 13.Ke5 Kh6 14.Kd5 Kxh5 15.Kc5 Kg5 16.Kxb5 Kf6
+- (#30) Depth: 30

Total Time (virt) = 396.700 sec (ca. 6.6 min).

Stunningly Chest verifies this mate in 30 in 6 minutes WITHOUT any tablebase use (automatic mode)! Can any playing program match this? I doubt.

Jouni

[hgm]

This is not really a mate-in-N problem, but basic end-game skill.
As it is possible to build the required EGTBs in a fraction of those 6 min, even if you wouldn't have them, there wouldn't really be an excuse if your engine could not solve this within, say, 1 minute.

[Uri Blass]

This is not really a mate-in-N problem, but basic end-game skill.
As it is possible to build the required EGTBs in a fraction of those 6 min, even if you wouldn't have them, there wouldn't really be an excuse if your engine could not solve this within, say, 1 minute.

How much time does your engine need to see the mate?

Does it build the required EGTBs?

It is possible in theory but I know no engine that does it and rybka failed to win against zappa KRP vs KNP thanks to the fact that it did not build the required EGTB.

Uri

[Mike S.]

Recently I noticed, that Toga 1.3.1 found a #22 against itself in a 4+2 blitz game (1.2 GHz)! So I thought that could be a good candidate for this. Engine does not use tbs. nor bitbases.

Analysis by Toga II 1.3.1 (3.4 GHz, 512 MB hash):

1.Kd1
= (0.02) Tiefe: 1/1 00:00:00
1.Kd2
² (0.26) Tiefe: 1/1 00:00:00
1.Kd3 Ke7 2.Ke3 Ke6 3.Ke4 Kf6 4.Kf4 Ke6 5.Kg5 Ke5 6.Kxh5 Kf5 7.Kh6 Kf6 8.h5 Kf7 9.Kg5 Kg7 10.Kf4 Kh6 11.Ke3 Kxh5 12.Kd4 Kg6 13.Kc5 Kf7 14.Kxb5 Ke6 15.Kc6 Ke7 16.Kc7 Ke6
² (0.38) Tiefe: 1/1 00:00:00
+- (10.60) Tiefe: 43/67 00:01:22 131mN

Here, Toga couldn't finish the 44th ply within 7 minutes.


The #22 position in the game was:
[D]8/k7/2p5/2P5/2P5/2P5/3K4/8 b - - 0 87
Analysis by Toga II 1.3.1:

87...Ka6
+- (1.46) Tiefe: 1/1 00:00:00
87...Kb7
± (1.34) Tiefe: 1/1 00:00:00

(...)

87...Kb7 88.Kd3 Kc7 89.Ke4 Kd7 90.Kf5 Ke7 91.Ke5 Kd7 92.Kf6 Kd8 93.Ke6 Kc7 94.Ke7 Kc8 95.Kd6 Kb7 96.Kd7 Ka6 97.Kxc6 Ka5 98.Kd6 Ka4 99.c6 Kb3 100.c7 Kc2 101.c8D Kd2 102.c5 Ke2 103.c6
+- (22.31) Tiefe: 32/65 00:00:14 20610kN
87...Kb7 88.Kd3 Kc7 89.Ke4 Kd7 90.Kf5 Ke7 91.Ke5 Kd7 92.Kf6 Kd8 93.Ke6 Kc7 94.Ke7 Kc8 95.Kd6 Kb7 96.Kd7 Ka6 97.Kxc6 Ka5 98.Kd6 Ka4 99.c6 Kb3 100.c7 Kc2 101.c8D Kd2 102.c5 Ke2 103.c6
+- (#23) Tiefe: 33/65 00:00:14 21027kN
87...Kb7 88.Kd3 Kc7 89.Ke4 Kd7 90.Kf5 Ke7 91.Ke5 Kd7 92.Kf6 Kd8 93.Ke6 Kc7 94.Ke7 Kc8 95.Kd6 Kb7 96.Kd7 Ka6 97.Kxc6 Ka5 98.Kd6 Ka4 99.c6 Kb3 100.c7 Kc2 101.c8D Kd2 102.c5 Ke2 103.c6
+- (#22) Tiefe: 34/65 00:00:15 22024kN
87...Kb7 88.Kd3 Kc7 89.Ke4 Kd7 90.Kf5 Ke7 91.Ke5 Kd7 92.Kf6 Kd8 93.Ke6 Kc7 94.Ke7 Kc8 95.Kd6 Kb7 96.Kd7 Ka6 97.Kxc6 Ka5 98.Kd6 Ka4 99.c6 Kb3 100.c7 Kc2 101.c8D Kd2 102.c5 Ke2 103.c6
+- (#22) Tiefe: 35/65 00:00:16 22721kN

This analysis was with 96 MB hash, like in the game but on faster cpu. According to 6-piece tbs., #21 is possible but I don't criticise the #22 announcement.

[hgm]

This is not really a mate-in-N problem, but basic end-game skill.
As it is possible to build the required EGTBs in a fraction of those 6 min, even if you wouldn't have them, there wouldn't really be an excuse if your engine could not solve this within, say, 1 minute.

How much time does your engine need to see the mate?

Does it build the required EGTBs?

It is possible in theory but I know no engine that does it and rybka failed to win against zappa KRP vs KNP thanks to the fact that it did not build the required EGTB.

Uri

As you know, my engine is less than a year old, and far from being finished yet. This is one of the capabilities that is on the list of things that have yet to be implemented.

Not very high on the list, though, as the capability to be able to tell that this is mate in 30, rather than mate in some other number of moves, brings you about zero Elo. (Joker is at 20 ply, +2 within a second on a 1GHz Athlon XP, and has the winning move even at 6 ply.) So I will give priority to things that will give large improvement. An evaluation would be nice, for instance, even if it was only through piece-square tables.

But a year from now Joker will certainly be able to use the EGTBs. I already have an EGTB builder that can do a 4-piece EGTB in 2 sec, but am still working on it to improve this to 0.2 sec. (4-piece is all that is needed here, as at most two Pawns can promote. So you would need KQQK and KQKQ, and a few P-slices of KQKP, KQPKP, etc. which are really 3-piece EGTBs and will only take a few msec.)

[Adam Hair]

I thought I was going to see a stunning chest