Hoffman endgame tool

[Michael Diosi]

Hi,

This is my first posting here as I just registered. I was reading the forum quite often though. O.K. Hoffman is a tool comparable (somehow) to Freezer. Here is the homepage:

http://www.freesoft.org/software/hoffman/

So what do you think ?

Michael

[Graham Banks]

Welcome to our forums Michael.

[Dhanish]

Hi Michael,

Thankyou for the link!

Do we need another tablebase? No, unless it is open source, and free to use by all.

Hoffman cannot make use of 6men bases, I think. Hence an engine which can utilize the 6men bases would be welcome. I had made a request to Franz Huber for incorporating this feature in ChestUCI, since it supports 6men bases. But he has difficulties with C programming.

Anyway, I downloaded Hoffman and had a try.

I copied all the xml files to the Hoffman directory and ran the batch file genalltb.bat.
36 htb files were generated.

Generation time (for all 4 men positions) on my Laptop: Hoffman 49 minutes; Nalimov - Generation 8 minutes, Compression 26s
So it seems generation of Nalimov tables is much faster.

File size (for all 4 men positions): Hoffman 15.92MB; Nalimov 15.23MB
File size appears similar.

I was not able to make the verify command work with the Nalimov files in another directory. How should the path be specified?
I tried
1) hoffman -v G:\4men kpk.htb
2) hoffman -v G:\4men\ kpk.htb
3) hoffman -v G:\\4men kpk.htb

I was getting comments like:
...
Can't find Nalimov tablebase
8/8/P7/2k5/8/8/8/K7 b - - (3441): Nalimov says illegal, but we don't
....

Note that Hoffman was installed in the default directory - C:\Program Files\Hoffman

After copying the Nalimov files to the Hoffman directory, the verify command works OK.

The kqk, krk, kbk, knk, files were OK, but when I tried to verify kpk, I got comments like:

4 piece Nalimov tablebases found
8/8/8/8/8/8/P7/K1k5 w - - (0): Nalimov says mate in 14 (65519), but we say draw
8/8/8/8/8/8/P7/K1k5 b - - (1): Nalimov says mated in 14 (14), but we say draw
...

So the kpk.htb is not OK. The file header is
<tablebase offset="0x0750">
<index type="compact" symmetry="2-way"/>
<format><dtm bits="8"/></format>
<piece color="white" type="king"/>
<piece color="black" type="king"/>
<piece color="white" type="pawn"/>
<futurebase filename="kk.htb"/>
<futurebase filename="kqk.htb"/>
<futurebase filename="krk.htb"/>
<futurebase filename="kbk.htb"/>
<futurebase filename="knk.htb"/>
<generation-controls>
<output filename="kpk.htb"/>
</generation-controls>
<tablebase-statistics>
<indices>202272</indices>
<PNTM-mated-positions>2348</PNTM-mated-positions>
<legal-positions>178794</legal-positions>
<stalemate-positions>11</stalemate-positions>
<white-wins-positions>0</white-wins-positions>
<black-wins-positions>0</black-wins-positions>
<forward-moves>1202258</forward-moves>
<futuremoves>62878</futuremoves>
<max-dtm>0</max-dtm>
<min-dtm>0</min-dtm>
</tablebase-statistics>
<generation-statistics>
<host>oem-4a301307828</host>
<program>Hoffman $Revision: 1.505 $ $Locker: $</program>
<args>hoffman -g kpk.xml </args>
<start-time>Thu Jun 19 13:52:11 2008 IST</start-time>
<completion-time>Thu Jun 19 13:52:12 2008 IST</completion-time>
<user-time>1.062s</user-time>
<system-time>0.000s</system-time>
<real-time>1.074s</real-time>
<page-faults>1327</page-faults>
<page-reclaims>0</page-reclaims>
<pass type="initialization" real-time="0.416s" user-time="0.406s"/>
<pass type="futurebase backprop" real-time="0.641s" user-time="0.640s"/>
<pass type="futuremove check" real-time="0.017s" user-time="0.016s"/>
</generation-statistics>
</tablebase>

The dtm statistics are incorrect. There appears to be some problem with this version of the generator?


Regards,
Dhanish