released: micro-Max 4.8w

[Uri Blass]

Feature request for the Winboard interface part:
setboard

My guess is that it is impossible with an engine that is so small.

[Dann Corbit]

Feature request for the Winboard interface part:
setboard

My guess is that it is impossible with an engine that is so small.

It already has edit in the interface, so he could just use the FEN, write a parser in the Winboard interface part, and use calls to edit methods

[hgm]

I wonder what is the reason that micromax4.4 is going to play in
the following tournament and not micromax4.8
http://www.csvn.nl/index.php?lang=en

Actually I do play uMax 4.8 there. The reason that it says different there is that I never informed them of it. I was under the impression that on the registration form I had just entered ' micro-Max', without specifying any version number. But perhaps I remember wrong. When I registered 4.4 was the most recent version, so if I gave a version number I would have given that. I will have them correct it.

I also wonder what caused you to implement check extensions
because I remember that earlier you were against this idea inspite of the fact that I told you that it is productive for playing strength.

This is true. At some point I discovered, however, that the detrimental effect of the check extension was mainly caused in the end-game, especially at the short time control where I tested. If I only extended in the middle-game (according to the same criterion as where I drop the King-safetety terms and the King starts walking to the center of the board) it did cause a 58% result in self-play, where before it was within 1 sigma of 50%.

I later confirmed that positive effect in a gauntlet of Nunn matches against other opponents. I should add that I could only test meaningfully against these other opponents after having implemented the rep-draw recognition. Before that most improvements were masked because the opponents tricked uMax into a rep-draw as soon as it started winning.

[hgm]

Feature request for the Winboard interface part:
setboard

My guess is that it is impossible with an engine that is so small.

I do not count the length of the Winboard-interface driver as part of the length of the engine. The character counts I report are always for the stand-alone versions.

I see the wWinboard versions mainly as a way to offer a more user-friendly way to experience how the stand-alone version would play. This is not perfect, as the Winboard version manages the search time through the time and otim commands, while the stand-alone version has a hard-wired number of nodes per move to search compiled into it, and starts (and always finishes) iterations as long as this number is not yet reached.

In the future I hope to chance this procedure and write a separate WB2UM (WinBoard to Micro-Max) interface, that connects to the stand-allone versions through a pipe. For this it should be able for the stand-alone versions to understand a command that could set the number of nodes they are allowed to search on a move-by-move basis. This should be possile in an acceptable number of characters, e.g. by using an invalid from-square indicator (e.g. ' t ' ) of the input move as time command, and using the to-square as an indicator for the number of nodes to search.

[hgm]

It already has edit in the interface, so he could just use the FEN, write a parser in the Winboard interface part, and use calls to edit methods

Well, I have a FEN parser in Joker. (But no setboard either.)

Setboard belongs to the Wiinboard 2 protocol, while my Winboard drivers were written for protocol 1. Can I safely implement it without causing the existing interface no longer to work? Is it simply a matter of echoing "setboard=1" when recieving the "protover" command?

[modolief]

I'm trying to compile and run this code. I had to include stdio.h and stdlib.h and it compiles, but I don't understand how to actually use it on the command line. Below is an example run. I've modified the code to turn the kibbitzer on so it displays "1 searched, score=" etc. The depth is set to 3. But it doesn't seem to actually be doing anything, other than allowing me to input e2e4, e7e5, g1f3 to arrive at the 3rd diagram. It doesn't play moves. My (compilable) fork of the code is here: https://github.com/purplejacket/umax/bl ... /umax4_8.c . I've also tried this with versions 4.0 and 3.2, same result.

Code: Select all

$ ./umax4_8 
rnbqkbnr
++++++++
........
........
........
........
********
RNBQKBNR
e2e4
 0 ply,         1 searched, score=     0 by a8a8
 1 ply,         2 searched, score=     0 by a8a8
rnbqkbnr
++++++++
........
........
....*...
........
****.***
RNBQKBNR
e7e5
 0 ply,         1 searched, score=     0 by a8a8
 1 ply,         2 searched, score=     0 by a8a8
rnbqkbnr
++++.+++
........
....+...
....*...
........
****.***
RNBQKBNR
g1f3
 0 ply,         1 searched, score=     0 by a8a8
 1 ply,         2 searched, score=     0 by a8a8
rnbqkbnr
++++.+++
........
....+...
....*...
.....N..
****.***
RNBQKB.R
^C

[modolief]

Ok, I've answered my own question. I have to press enter twice! Once to enter the move, the second time to ask the computer to move

Any other comments on the command line experience of this program would be welcome. I'm playing around with using MicroMax as a teaching example of converting C code to webassembly using emscriptem.

Here's what I'm getting now:

Code: Select all

$ ./umax4_8 
rnbqkbnr
++++++++
........
........
........
........
********
RNBQKBNR
e2e4
 0 ply,         1 searched, score=     0 by a8a8
 1 ply,         2 searched, score=     0 by a8a8
rnbqkbnr
++++++++
........
........
....*...
........
****.***
RNBQKBNR

 0 ply,         1 searched, score=     0 by a8a8
 1 ply,         2 searched, score=     0 by a8a8
 2 ply,        13 searched, score=    15 by b8c6
 3 ply,       374 searched, score=     0 by b8c6
 4 ply,       772 searched, score=    15 by b8c6
 5 ply,     15164 searched, score=     2 by b8c6
 6 ply,     32643 searched, score=    15 by b8c6
 7 ply,    308953 searched, score=     5 by d7d5
 8 ply,   1354765 searched, score=    14 by b8c6
r.bqkbnr
++++++++
..n.....
........
....*...
........
****.***
RNBQKBNR
d2d4
 0 ply,         1 searched, score=     0 by b1a8
 1 ply,         2 searched, score=     0 by b1a8
r.bqkbnr
++++++++
..n.....
........
...**...
........
***..***
RNBQKBNR

 0 ply,         1 searched, score=    71 by c6d4
 1 ply,         2 searched, score=     0 by c6d4
 2 ply,         7 searched, score=    15 by d7d5
 3 ply,       143 searched, score=    23 by d7d5
 4 ply,       383 searched, score=    24 by d7d5
 5 ply,      4413 searched, score=    12 by d7d6
 6 ply,     17549 searched, score=    24 by d7d5
 7 ply,    174156 searched, score=    11 by d7d5
 8 ply,   1043860 searched, score=    12 by d7d5
r.bqkbnr
+++.++++
..n.....
...+....
...**...
........
***..***
RNBQKBNR

[Roland Chastain]

I don't understand how to actually use it on the command line.

Maybe use the CECP protocol?

[hgm]

It looks like this is the stand-alone version (4.8), not the CECP version (4.8w). The stand-alone version has a very simple interface: it accepts a move on a line, and thinks up a move of its own when you enter an empty line. After every input it prints an ascii board for the current position. That is all it does.

[mclane]

How is the size of the engine in bytes ?