Naum 1.91

[Naum]

This version is not the same as version that played first 2 rounds in WBEC.

This is also not the version that scored well against Fruit.
I decided to pick beta that scores very well against attacking engines, but plays slightly worse (although still better then 1.9 I think) against the positional engines.

It will be interesting to see how much more (if any) ELO points I get with this version, since I didn't have time to do some comparison matches for both 1.9 and 1.91.

And as mentioned on WBEC, if you are running Naum on a very slow machine (dual P3 for instance) and using 5pc TBs, it is better to disable TBs for Naum altogether. I think 5pc TBs on a slow machine (or slow hard drive) do more damage then good for Naum.

Alex

[Graham Banks]

Thanks Alex!

[John Balchi]

Thanks for the new update! I don't know if you are interested at all, but out of my own curiousity I'm doing this anyway. Already I have started a test with the engines. I am doing some Naum 1.9 vs Naum 1.91 matches and larger tourney testing them on a few different engines. I will also do the tests on different time controls, ranging from very fast, to medium. I'll report my own results here as I get them. They will not be extensive though, since I want to get semi-fast results and I don't have the resources to run hundreds of games on long time controls. But for anyone interested, I'll post the results here just the same, it gives me something to do

[John Balchi]

Well the first small results are in, so I may as well post them before I head off to bed and set up another match.

The first test was a very fast test with Naum 19.1 vs Naum 19, running as uci engines, with books disabled, 3-4-5 man egtb enabled, 64mb hash size, using Arena 1.99 beta2 on an Amd Atholn XP 2500 with 2.25 GB RAM, on a 2 min/+2 second blitz time control.

Naum 19.1 vs Naum 19 (29.5-20.5) Win by Naum 19.1
True win/loss excluding ties (12-3), Naum 19.1 is the clear winner here on a fast blitz.

I wil probably do 2 more matches, each at higher time controls, then test them against more engines.

[Naum]

Thanks for the tests John.

It's good to know that 1.91 beats 1.9. I didn't have time to do that test.

Alex

[John Balchi]

Well the 2nd results for Naum 1.91 and 1.9 are in. On the slightly longer time controls it's not looking so good for 1.91, but then these tests aren't long or thorough enough for great results, it's just something to do.

Same conditions as before, but 5 min/+3 second blitz time.

Wins
Naum 1.91 vs Naum 1.9 (21-29) win by Naum 1.9
Excluding ties (12-20)

It looks like so far between the 2 tests I haven't really proven much heh. But then they are only 50 game matches and 2 versions playing each other isn't usually the best way for tests either.

[Naum]

Well, it is possible that the change I did is good for blitz games only. Time will tell.
Are you using start positions (noomen, nunn) or a special book?

Results will be more random if engines don't play the same position with both colors.

[John Balchi]

I'm not using any books for the engines. I am just starting tests against other engines now, all books and learning (if it applies) disabled.

[Uri Blass]

I'm not using any books for the engines. I am just starting tests against other engines now, all books and learning (if it applies) disabled.

In these conditions you may get the same game again and again and it seems very boring test.

If you do not use original books you should try fixed positions otherwise your result mean nothing.

It is clear that not all the games are the same based on the results but you may have many doubles.

Uri