My First Project : Toga2012

[Tennison]

Yes it 'd be a complete nice solution !!!

Thanks a lot ...

[michiguel]

Yes it 'd be a complete nice solution !!!

Thanks a lot ...

Ok. I will put a brief note in your original message. Sorry to edit it, but it will be the quickest solution.

Miguel

[Tennison]

File removed from my web space ...

[michiguel]

File removed from my web space ...

Thanks for understanding and best luck in your project!!

Miguel

[Tennison]

Here is the results of the 3 min games

Nice results ...

Code: Select all

Test 5 : 3'
====================

Fritz 11 GUI / TB5 on / HS7movesplus.ctg

Toga 2012 beta 1.00 x64   +28/-22/=50   53.00%
Toga II 1.4 beta5c        +22/-28/=50   47.00%   

[JuLieN]

Here is the results of the 3 min games

Nice results ...

Code: Select all

Test 5 : 3'
====================

Fritz 11 GUI / TB5 on / HS7movesplus.ctg

Toga 2012 beta 1.00 x64   +28/-22/=50   53.00%
Toga II 1.4 beta5c        +22/-28/=50   47.00%   

Hello Ben, and good luck with your project!

But be careful with those results: 100 games is nothing, when testing. I often saw a new version of my engine leading a test match by 20 or more won games after 100 games, and fail miserably if the test continued to 2000 games.

My suggestion: play at least 2000 games at a very fast pace (my tests, for example, are done at 0 second for the whole game plus 100 milliseconds by move).

[Tennison]

I know 100 is not enough but there you can see 500 results :

Test 1+ Test 2 + Test 3 + Test 4 + Test 5 =500

The result is :

Code: Select all

Toga2012 :              252,5
Toga II 1.4 beta 5c :   247,5 

So this is enough for me to have an idea : the compile is not COMPLETELY buggy ...


When I try quicker than the tests I show abose the engines (original one and my compile) loses lot of games on time ...

[ethanara]

I am quite sure the reason for this (losing on time) is that you use Arena or other GUI.
I would recommend you (if it doesn't matter) to use something like Cutechess-cli, which is much faster because there is no graphics. I am quite sure you can use 4 seconds with 100 ms increment without any problems.
Regards
Ethan

[JuLieN]

I am quite sure the reason for this (losing on time) is that you use Arena or other GUI.
I would recommend you (if it doesn't matter) to use something like Cutechess-cli, which is much faster because there is no graphics. I am quite sure you can use 4 seconds with 100 ms increment without any problems.
Regards
Ethan

I do my test under Arena with 0s + 100ms increment and have no loss on time, despite my engine is far from being the most reactive one in the world. So the problem comes from somewhere else. (Arena probably compensates for the lag the graphics would produce).

[bob]

Nice

On the other hand if you get a compiled version for linux maybe you'll find more testers for your project

EDIT: well, for what Miguel says reltaed to GPL code it doesn't seem to be that nice, but in any case good luck with your project! Besides, I guess the more open you make your project the more you can get feedback from the community.

As I said, we should welcome Ben in his endeavor, but there is no other option than removing the link. It is a direct violation of the license. When he is ready with the source code, we should be happy to allow the distribution of his modifications.

Miguel

I understand well but without some testers I can't be sure the slightly modified source don't weaken the original project ...

I just need some testers for 2 or 3 days (just to finish cleaning the files).

It's why I'll give the source in 2 or 3 days ...

Is this a so big problem 2 or 3 days at max ???

As Miguel said, "yes it is." The GPL is not open for debate/discussion, it is quite clear. If you distribute a binary, you have to provide the source. At the same time. not a day, or a week, or a year later.