Mate in 115

[Dave Mitchell]

Which is just what you would expect, if that install of CM 10, did not yet include the table base files and have CM set up to use them.

[zullil]

It seems that you have the 5-piece tablebases but not the 6-piece tablebases.

[Jimbo I]

About Allan's position, I'm getting conflicting numbers. Most of the engines using Nalimov TBs return -M 115. Except that Tornado 4.4 returns -M 116.

And with Gaviota TBs, Umko 1.1 returns -M115, but two versions of Gaviota (including 0.80) return -M 102.

Any ideas for the discrepancies? Problems with how a few engines use the TBs? And maybe someone could confirm my results? (All engines were 32-bit in Arena 2.0.1.)

[Houdini]

Houdini 1.5 with Gaviota TB says it's mate in 115.

[ernest]

According to the tablebases, the position is indeed Mate-in-115.

This shortest mate in 115 is indeed legal (White pushes his Pawn after less than 50 moves)
It would be interesting to search, starting from this given position, if White can hold his Pawn still during 50 moves, allowing a non-optimal (shorter) Mate, but in fact obtaining a legal Draw.

[michiguel]

About Allan's position, I'm getting conflicting numbers. Most of the engines using Nalimov TBs return -M 115. Except that Tornado 4.4 returns -M 116.

And with Gaviota TBs, Umko 1.1 returns -M115, but two versions of Gaviota (including 0.80) return -M 102.

Any ideas for the discrepancies? Problems with how a few engines use the TBs? And maybe someone could confirm my results? (All engines were 32-bit in Arena 2.0.1.)

The problem is that I messed up the output of the engine in analyze mode when I was trying to limit it to avoid lag in Arena... arrrrrrrrgggggggggghhhhh. Thanks for reporting it. I need to fix this because it is embarrasing.

The probing and the TBs are fine (console command tb) =

Code: Select all

+-----------------+
| . . . . . . . . | [Black]
| k . . . . . . . |
| . . . . . . . . |
| . . . . K n . . |    Castling: 
| . . . . . . . . |    ep: -
| P . . . . . . . |
| . . . . . . . . |
| . n . . . . . . |
+-----------------+

tb
 Black wins in  115 moves
 ----------------------------------
 Ne7      wins in 115 (229 plies)
 Nc3     DRAW
 Nd2     DRAW
 Nd4     DRAW
 Ne3     DRAW
 Nd6     DRAW
 Nxa3    DRAW
 Ng3     DRAW
 Ng7     DRAW
 Nh4     DRAW
 Nh6     DRAW
 Kb6     DRAW
 Kb7     DRAW
 Ka6     DRAW
 Kb8     DRAW
 Ka8     DRAW

Miguel

[michiguel]

About Allan's position, I'm getting conflicting numbers. Most of the engines using Nalimov TBs return -M 115. Except that Tornado 4.4 returns -M 116.

And with Gaviota TBs, Umko 1.1 returns -M115, but two versions of Gaviota (including 0.80) return -M 102.

Any ideas for the discrepancies? Problems with how a few engines use the TBs? And maybe someone could confirm my results? (All engines were 32-bit in Arena 2.0.1.)

The problem is that I messed up the output of the engine in analyze mode when I was trying to limit it to avoid lag in Arena... arrrrrrrrgggggggggghhhhh. Thanks for reporting it. I need to fix this because it is embarrasing.

I fixed the problem here and it was a very silly bug. Historically, Gaviota understood mate scores until 200 plies. That was plenty because the maximum ply search was 64. Of course, until I introduce the TBs... So, it could never report mates longer than a 100 + number of plies from the root. I needed to change one number.

Miguel

The probing and the TBs are fine (console command tb) =

Code: Select all

+-----------------+
| . . . . . . . . | [Black]
| k . . . . . . . |
| . . . . . . . . |
| . . . . K n . . |    Castling: 
| . . . . . . . . |    ep: -
| P . . . . . . . |
| . . . . . . . . |
| . n . . . . . . |
+-----------------+

tb
 Black wins in  115 moves
 ----------------------------------
 Ne7      wins in 115 (229 plies)
 Nc3     DRAW
 Nd2     DRAW
 Nd4     DRAW
 Ne3     DRAW
 Nd6     DRAW
 Nxa3    DRAW
 Ng3     DRAW
 Ng7     DRAW
 Nh4     DRAW
 Nh6     DRAW
 Kb6     DRAW
 Kb7     DRAW
 Ka6     DRAW
 Kb8     DRAW
 Ka8     DRAW

Miguel

[ernest]

Found the position on the web page below.
http://en.wikipedia.org/wiki/Two_knights_endgame.

Are you sure?
I find no such diagram in that link.

[rabbits]

Ernest: I could not find it either when I clicked on to the link I had pasted there-which is wierd- yet when I typed in, Chess two knights versus pawn checkmate, into "Google" the same article for Wikipedia came up on the screen with the position on it. Try that method and see if you have any success. The article is to do with the pawn being beyond the Troitsky line
Allan
I've pasted the address again.
http://en.wikipedia.org/wiki/Two_knights_endgame

[ernest]

http://en.wikipedia.org/wiki/Two_knights_endgame

Thanks: this link works!
Found the problem: the dot at the end of the previous http://en.wikipedia.org/wiki/Two_knights_endgame.