Displaying Matt in Stockfish

[retep1]

Why did'nt Stockfish show the move-number #15 instead of 128,44?
I've syzyzgy and gaviota installed.

[hgm]

Because syzygy is DTZ, and not DTM, and Stockfish uses syzygy?

[royb]

I tried this position out with SF (Jan 01 2017 version) and after 29 plies, it shows M18 (mate in 18 which it then changes to M15 as it gets even deeper with M15 shown at depth 39.

I use SCID on Linux and I cannot figure out how to copy the engine analysis so I cannot show the output here in this post.

[hgm]

But did you use tablebases?

[royb]

I used syzygy bases, yes. But not traditional (or whatever you call the "old style" tablebases with distance to mate built into them).

[Colin-G]

Stockfish 8 32bit running in Scid with linux Mint 17
Stockfish with 5 man Syzygy shows mate score of 99.70
The TB info shown by Scid is mate in 15.
This is because Scid uses Nalimov tablebases in its set up.

[royb]

Interesting. I use SCID 4.6.4 and cannot find any Nalimov TBs anywhere in the directory structure where SCID 4.6.4 "sits" on my hard drive. I was thinking maybe SCID just automatically had Nalimov TBs built into it. I don't see it, perhaps I'm just missing it.

I definitely downloaded syzygy 5 men tables and configure Stockfish to use those, but I cannot see anywhere where SCID is configured to use Nalimov TBs (which have built into them the distance to mate). There is definitely an option to have SCID use such TBs, but nothing is configured there (under Option -> Tablebase Directory) on my system. I've got nothing at all specified there (i.e. it's blank).

Interesting is that I put in this position:

4k3/8/8/8/8/8/8/2NBK3 w - - 0 1

and then ran the same Stockfish engine and get not a mate announcement but this eval: +128.51 So, somehow, with just Kings and pawns, there is a mate announcement coming, but with B+N+K vs K, no mate announcement (as expected with only syzygy bases to help the engine out).

[hgm]

Well, in KBNK the mate is much farther away, and it is much less obvious how to make progress (where in Pawn endings you have promotion as an obvious sub-goal with an enormous score improvement). The branching ratio in KBNK is also much larger than in any Pawn ending, with far fewer transpositions.

[kgburcham]

[D] 4k3/8/8/8/8/8/8/2NBK3 w - -


Engine: Stockfish 070117 64 BMI2 (8192 MB)
by T. Romstad, M. Costalba, J. Kiiski, G.
Found 172 tablebases

55/72 0:03 +128.51 1.Kf2 (125.573.048) 32389 TB:15
55/72 0:03 +128.51 1.Kf2 (126.023.000) 32388 TB:15
55/72 0:10 +M30 1.Kf2 Kf8 2.Ke3 Ke7 3.Kf4 Kd6 4.Bb3 Kc5 5.Ke4 Kb4 6.Kd3 Ka3 7.Kc2 Kb4 8.Nd3+ Kb5 9.Kd2 Kb6 10.Ke3 Kc6 11.Kf4 Kd6 12.Kf5 Ke7 13.Be6 Kd6 14.Kf6 (334.569.732) 31350 TB:15
56/72 0:15 +M29 1.Kf2 Kf8 2.Ke3 Kg7 3.Kf4 Kg6 4.Bc2+ Kf6 5.Nd3 Kg6 6.Bb3 Kf6 7.Ne5 Kg7 8.Kf5 Kh7 9.Kf6 Kh8 10.Bc2 Kg8 11.Nf7 Kf8 12.Bh7 Ke8 13.Ne5 Kd8 14.Be4 (487.866.679) 32102 TB:15

[hgm]

Well, with tablebases it should have it immediate, of course. Because DTZ is in fact identical to DTM in KBNK. There is no way to zero the ply counter ad still have a won position than checkmating.