Engine testing

[Patrice Duhamel]

Hello,

In previous version of my engine, I was only testing with epd files or few games in Arena.

Now, I'm trying to use Little Blitzer, or Cute Chess cli to run more games.

What time control settings should I use for this ?
Where to find good opening book in PGN or EPD format to use with Little Blitzer or Cute Chess cli ?

[Laskos]

Hello,

In previous version of my engine, I was only testing with epd files or few games in Arena.

Now, I'm trying to use Little Blitzer, or Cute Chess cli to run more games.

What time control settings should I use for this ?

If you need many games in a limited time, then ultra-short TC, like 1s + 0.05s. Be careful what you are measuring there at these TC.

Where to find good opening book in PGN or EPD format to use with Little Blitzer or Cute Chess cli ?

I know of only some few:

EPD from Bob Hyatt (4,000 positions), it's on a FTP site
ftp://ftp.cis.uab.edu/pub/hyatt/tests/

PGN from Frank Quisinsky (5,000 or so games to 10 moves, if I remember). You should find it here, if it's the same site I took it from:
http://www.amateurschach.de/

Kai

[Patrice Duhamel]

Thank you,

It's dangerous to test with "ultra short" time controls ?

I want to run as much game as I can, but I'm not sure if I should trust the results with too short times.

[JuLieN]

Thank you,

It's dangerous to test with "ultra short" time controls ?

I want to run as much game as I can, but I'm not sure if I should trust the results with too short times.

No: just remember that our "ultra short time controls" were the "ultra long time controls" of our computers 20 years ago, and we then saw no problem with it.

For instance, back then I was using an Amiga 500 running at 0.8 MIPS.
My current system is an iMac running around 50.000 MFLOPS (let's say MIPS to simplify)... so it is 62.500 times faster.

So "100ms/move" would be 1 hour, 44 minutes and 10 seconds per move on the Amiga 500.

[kbhearn]

People talk about certain features like king safety being dangerous to test at ultrafast time controls (perhaps because such large swing items can require more depth to either find how to convert or find the exceptions where the eval term is wrong)