Test versions / private versions wanted for testing purposes

[hgm]

Seems to me that "something is better than nothing" applies here. I would be really surprised if the strength you created by adding features in 8x8 play would not generalize to 8x10 play or Chess960.

Furthermore, I really don't see why the existing Winboard protocol, for instance, would not be able to handel 8x10 chess, after te trivial addition of allowing an i and j file in parsing of the move.

[Dann Corbit]

There is another open source chess project that deals with arbitrary boards:
http://www.chessv.com/

This projects also seems related:
http://sourceforge.net/projects/gcboard/

Richard seems to go about things in different ways from other programmers but I think that is good. New ideas might hatch in that way. He seems to have put a lot of effort into move generation. He makes helpful posts on that subject from time to time.

[Dann Corbit]

Hi Daniel,

once upon a time there have been testers of chess programs, today there (mostly) are merely owners of computers having some testing automatism run. Maybe, if those tested engines there would play tic-tac-toe instead of chess, those "testers" would not even notice.

As far as I could experience myself by manually testing (there is no common protocol in 10x8 Chess yet) SMIRF actually is top together with GothicVortex, which is a very specialized program compared to SMIRF, which in contrast is able to play a lot of variants including traditional Chess and Chess960. Thus I prosume SMIRF would have a strength of about 2500 Elo minimum.

Does anyone know, what happened to Ed Trice, the main author of GothicVortex? His website has not been updated for weeks and his online server for Gothic Chess is no longer answering.

Reinhard.

I suggest this thing as an opponent:
http://www.chessv.com/

Lots of variants:
Alapo
Almost Chess
Angels and Devils
Archchess
Berolina Chess
Bird's Chess
Cagliostro's Chess
Capablanca Chess
Capablanca Chess, Aberg variant
Capablanca Chess, Paulowich variant
Carrera's Chess
Chess 480
Chess with Augmented Knights
Chess with Different Armies
Chess with Ultima Pieces
Courier Chess
Cylindrical Chess
Diagonal Chess
Diamond Chess
Embassy Chess
Emperor's Game
Eurasian Chess
Exinction Chess
Fischer Random Chess
Great Chess
Great Shatranj
Grand Chess
Grotesque Chess
Janus Chess
Janus Kamil Chess
Kinglet
Ladorean Chess
Legan's Game
Lions and Unicorns Chess
Los Alamos Chess
Modern Kamil
Modern Shatranj
Odin's Rune Chess
Opti Chess (mirror I)
Opulent Chess
Orthodox Chess
Polymorph Chess
Roman Chess
Royal Court
Schoolbook Chess
Shatranj
Shatranj Kamil
Shatranj Kamil (64)
Sosarian Chess
Switching Chess
TenCubed Chess
Three Checks Chess
Ultima
Unicorn Chess
Unicorn Great Chess
Unicorn Grand Chess
Univers Chess

Once upon a time it seemed about the same strength as Smirf:

• Final Results, 2004 Gothic Chess Computer World Championship
  
  Place Program Points Wins Losses Draws
  1 Gothic Vortex 14 14 0 0
  2 TSCP Gothic 11.5 11 2 1
  3 Chess V 9.5 9 4 1
  4 S.M.I.R.F. 7.5 7 6 1
  5 Zillions 6.5 6 7 1
  6 EGM 0.1 3.5 3 10 1
  7 CapaGNU Modified 3 3 11 0
  8 Max Gothic 0.5 0 13 1

[smrf]

Final Results, 2004 Gothic Chess Computer World Championship

Place Program Points Wins Losses Draws
1 Gothic Vortex 14 14 0 0
2 TSCP Gothic 11.5 11 2 1
3 Chess V 9.5 9 4 1
4 S.M.I.R.F. 7.5 7 6 1
5 Zillions 6.5 6 7 1
6 EGM 0.1 3.5 3 10 1
7 CapaGNU Modified 3 3 11 0
8 Max Gothic 0.5 0 13 1

Well, that was when SMIRF (engine and GUI) just have been written to participate there, and just its basic concept has been realized. In the time until now Gothic Vortex, ChessV and SMIRF have been improved a lot. There is no common protocol for 10x8 chess yet, so all tests and games still have to be performed manually.

ChessV has an even more flexible concept than SMIRF, which in contrast is targeting only at rule supersets to traditional chess, a concept called FullChess. But ChessV still needs some work to become more stable. E.g. it iternally sometimes switches sides when being under pressure, that is, when a forced draw or mate is coming near. Nevertheless SMIRF has done a lot of successful training matches playing ChessV.

GothicVortex is a very specialized program, actually supporting only GothicChess as a single game. Its strong elements are a huge end game table support and a big opening library. Because of SMIRF still having no looking-up tables in its small 72K universal engine playing with Vortex mostly leads to a big initial advantage in time consumption for Vortex. But SMIRF still is able to compete, whereas Vortex Gold version actually exceeds selected average time limits when getting in trouble. Thus now SMIRF over all is slightly more successful playing Vortex.

The SMIRF GUI is able to import and export PGN files for all supported variants as well as related X-FEN strings. So it does with exporting of 8x8 or 10x8 Chess sceneshots.

To get an impression of the real playing strength of SMIRF and also the both other 10x8 chess protagonits it would be helping to have a seriously performed test of SMIRF playing traditional chess.

Regards, Reinhard.

[hgm]

To get an impression of the real playing strength of SMIRF and also the both other 10x8 chess protagonits it would be helping to have a seriously performed test of SMIRF playing traditional chess.

Well, this is exactly the point I tried to make. There are hundreds of engines out there to perform such tests, and to do a meaningful test you would indeed need a large number of opponents.

But you cannot expect them all to change their interface to accomodate SMIRF. That burden is on you. Neither can you expect testers to play hundreds of games by hand, when it is about a hundred times easier for you to make a version with an adapted interface...

[smrf]

Seems to me that "something is better than nothing" applies here. I would be really surprised if the strength you created by adding features in 8x8 play would not generalize to 8x10 play or Chess960.

I have to confess, that I do not understand the intention of that sentence.

Furthermore, I really don't see why the existing Winboard protocol, for instance, would not be able to handel 8x10 chess, after te trivial addition of allowing an i and j file in parsing of the move.

Well, you should have a look at SMIRF. That is, why I still hasitate to have a UCI version of SMIRF (as I had done with SmirfMateUCI for a short time).

There are some problems with common protocols:

a) no support for Chancellor and Archbishop (playing, promoting ...)
b) no support for different castling methods
c) no consistent support for random variants like Chess960 or CRC
d) no support for the compatible X-FEN extension
e) no support for necessary PGN files

Reinhard.

[smrf]

To get an impression of the real playing strength of SMIRF and also the both other 10x8 chess protagonits it would be helping to have a seriously performed test of SMIRF playing traditional chess.

Well, this is exactly the point I tried to make. There are hundreds of engines out there to perform such tests, and to do a meaningful test you would indeed need a large number of opponents.

But you cannot expect them all to change their interface to accomodate SMIRF. That burden is on you. Neither can you expect testers to play hundreds of games by hand, when it is about a hundred times easier for you to make a version with an adapted interface...

Watching the "testers'" scene, I do not expect this, despite of hoping, it would still be possible.

I simply stated, that there would be a need for testing. As in earlier days of computer chess and as in the 10x8 chess scene it is possible to have games performed manually. If there are not such games e.g. between SMIRF and traditional 8x8 chess playing engines, that would mean, that there is no interest in chess development, instead there is an interest to have the 999th automatically performed Blitz-event, digging for absolutely "surprising" 2 Elo point shifts.

Reinhard.

[Daniel Mehrmann]

Does anyone know, what happened to Ed Trice, the main author of GothicVortex? His website has not been updated for weeks and his online server for Gothic Chess is no longer answering.

Reinhard.

Hi Reinhard,

just a question: What makes your Go plans ?

I'm thinking Go is the ultimate challenge. Chess or variants of chess like chinese chess are working with the same concept. The strategy and flow of pieces are more or less the same, but Go is completly different.

I've started to playing this game. This is my first stage. I want to learn it first. I want to understand the strategy, plan and ideas. I'm impressed so far. How deeper i'm going into the Go world, i realize how complex and difficult is this game. I'm starting to feel the power of this game move by move and game by game.

I bought 2 books, a board and pieces. I readed the books learning some basic stuff. I'm playing on a internet go server currently. Next week i'm going to a real Go club in Hannover and there will be start a Go tournament in 2 months...

I had some basic ideas for my Go program. But now, right now, i',m see the power, the flow and the energy of this game and i'm starting to rethink about my ideas.
However, it makes fun and i'm think i found my personal challenge to write a Go program. I feel chess was just a preparation for my Go program...



Best,
Daniel

[smrf]

Does anyone know, what happened to Ed Trice, the main author of GothicVortex? His website has not been updated for weeks and his online server for Gothic Chess is no longer answering.

just a question: What makes your Go plans ?

Hi Daniel,

Go is always present in my head. But still there are some very basic questions to be solved before starting any Go programming. Those all are related to the kind of influences a single stone has on a Go board.

My approach to chess programming has started in a comparable way. As you might know I developed a simple model on the average exchange values of different chess pieces. From that I could design a related evaluation concept. SMIRF still is based on the first ad hoc and very slowly working evaluation routine. Thus I am sure, that a completely rewritten SMIRF would improve significantly.

I'm thinking Go is the ultimate challenge. Chess or variants of chess like chinese chess are working with the same concept. The strategy and flow of pieces are more or less the same, but Go is completly different.

In the meantime I have made concepts for a performant Go representation including moves and removes. Moreover the 10x8 SMIRF concept already is handling about 35% more moves a node as in traditional chess. Thus it could be seen as a challenge of intermediate level on the way to handle Go programming. In 10x8 chess you also could experience the raising of nearly independent battle fields, which is typical for Go.

I've started to playing this game. This is my first stage. I want to learn it first. I want to understand the strategy, plan and ideas. I'm impressed so far. How deeper i'm going into the Go world, i realize how complex and difficult is this game. I'm starting to feel the power of this game move by move and game by game.

I bought 2 books, a board and pieces. I readed the books learning some basic stuff. I'm playing on a internet go server currently. Next week i'm going to a real Go club in Hannover and there will be start a Go tournament in 2 months...

I had some basic ideas for my Go program. But now, right now, i',m see the power, the flow and the energy of this game and i'm starting to rethink about my ideas.
However, it makes fun and i'm think i found my personal challenge to write a Go program. I feel chess was just a preparation for my Go program...

My personal experience is, that it is important to know a game by personal experience but to keep distant from practise and from trying to implement personal experiences during programming. Because that has to be done from a CPU's view. A human being is not able to decribe its modus operandi sufficiently exact, thus I regard it for to be a bad programming strategy to try to simulate human behaviour.

Go is still in my mind, but I probably will first rewrite SMIRF, to get my head free from still unimplemented strategies for my chess programming approach.

There still a to be recomended Go GUI seems to be missing, e.g. concerning pondering and flexible protocols. But I have not been watching the Go programming scene for a while. If you would be starting with that matter, may be it would be a good idea, to exchange ideas especially on a to be recommended flexible GUI.

Best regards, Reinhard.

[hgm]

Well, you should have a look at SMIRF. That is, why I still hasitate to have a UCI version of SMIRF (as I had done with SmirfMateUCI for a short time).

There are some problems with common protocols:

a) no support for Chancellor and Archbishop (playing, promoting ...)
b) no support for different castling methods
c) no consistent support for random variants like Chess960 or CRC
d) no support for the compatible X-FEN extension
e) no support for necessary PGN files

It seems to me that you don't need any fundamental change in the existing protocols for any of this. I am only familiar with Winboard protocol I, but that protocol fully provides in a move syntax that allows you to promote to a piece of your choice. Castling moves can be easily represented in normal algebraic notation. FENs I so far did not need at all to play my engines under Winboard, as far as I know they are not part o the protocol. PGN files are not communicated between teh GUI and the engine at all. So none of this seems a real problem.