Test you engine

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
Kempelen
Posts: 620
Joined: Fri Feb 08, 2008 9:44 am
Location: Madrid - Spain
Contact:

Test you engine

Post by Kempelen » Mon Mar 10, 2008 9:27 am

When you program and develop a new feature in your engine, how do you know it is a possitive one?

I mean, lets suppose you have a little possitional improvement of knowing when your king it is better placed in other square. How do you know is good enaught, and it is better than don't calculate it and improve in calculaton time???
I think in lot of examples. What I am asking for is a good way to measure my effors.
thx

User avatar
hgm
Posts: 24619
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Test you engine

Post by hgm » Mon Mar 10, 2008 9:47 am

I always play a standard gauntlet of 480 games (Nunn matches against 24 opponents) to get a rough impression if a change works or not. Even then the improvement should be substantial, as the statistical error over 480 games is still 2%.

Richard Allbert
Posts: 771
Joined: Wed Jul 19, 2006 7:58 am

Re: Test you engine

Post by Richard Allbert » Thu Jul 17, 2008 5:03 am

hgm wrote:I always play a standard gauntlet of 480 games (Nunn matches against 24 opponents) to get a rough impression if a change works or not. Even then the improvement should be substantial, as the statistical error over 480 games is still 2%.
Hi,

At what time control do you run the gauntlet?

Thanks

Richard

User avatar
hgm
Posts: 24619
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Test you engine

Post by hgm » Thu Jul 17, 2008 8:22 am

That depends. I usually start with 40/1', to check if I have not broken anything, without wasting too much CPU time if I have. Eventually I do 40/10'.

Post Reply