Winboard Forum

Hi,

I just want to let people know that you shouldn't put current version of Naum in long running tournaments, since I am going to make another release soon.

I had a big surprise when I came back from work tonight and saw that Naum managed to beat Fruit 2.1 in a 100 game match. Previous results weren't very close.

The most of the strength increase comes from fixing the pawn shelter bug I knew about for about 2 years. Naum wasn't taking into account the position of its king when evaluating its pawn shelter. So if white king is on g5 and pawns are on f2-g2-h2, Naum would think it's completely safe

Don't get too excited. I don't think ELO increase would be as big as match against Fruit indicates, but I hope to get around 30 additional ELO points from this change.

Regards,
Alex

Hi Alex ,

good news. We are looking forward to the new version.

Although the League will be finished with Naum 1.9, we will test Naum 1.91 with many in CEGT.

Best Regards
Heinz

Hi Heinz,

I will probably do a release after I finish testing all the changes (in a week or two).

Bad news is that pawn shelter code had a bug, and once I fixed this bug Fruit crashed me again in a new match. Crazy!

It seems that I stumbled upon something good by pure chance. Of course, I am going to leave this bug in Naum for now and try to find out why it is so good against Fruit.
Results of the bugy pawn shelter eval are also good against other opponents, but not as good as against Fruit.

Regards,
Alex

What exactly is the error? Maybe you have discovered a new principle of chess!

Heh, I wish I did...

It basicaly gives a high penalty for king and pawn shelter position that I previosly thought is pretty safe. I did some thinking about it and now I think there is some logic to it. I still think the penalty is too high. Will try to experiment with the smaller values.
Of course, I don't want to give all the details to my 'opponents', but if I confirm this I will probably let Fabien know. I benefited so much from Fruit's code, so it would be fair to give something back.

Alex

Naum wrote:Hi Heinz,

I will probably do a release after I finish testing all the changes (in a week or two).

Bad news is that pawn shelter code had a bug, and once I fixed this bug Fruit crashed me again in a new match. Crazy!

It seems that I stumbled upon something good by pure chance. Of course, I am going to leave this bug in Naum for now and try to find out why it is so good against Fruit.
Results of the bugy pawn shelter eval are also good against other opponents, but not as good as against Fruit.

Regards,
Alex

Hi Alex,

I don't follow what you mean by the expression "Fruit crashed me again in a new match." I understand "crash" to mean that the application or system "froze" or "stopped working". Perhaps you meant "trashed" which would mean "defeated decisively". Perhaps it was a typo?

regards,
Norm

He probably meant "crushed". I don't think that it's good idea to pay attention to every typo here.

More interesting would be to know what Alex mean by "crush". 50%? Less?

Dariusz Orzechowski wrote:He probably meant "crushed". I don't think that it's good idea to pay attention to every typo here.

More interesting would be to know what Alex mean by "crush". 50%? Less?

crushed probably, anyway crushed means the same as trashed. 99.9% of the time, a typo is self-evident and can clearly be recognized and corrected visually. However in this case, it sent me in a different direction.

Just a typo of course. I ment 'crushed'.

By crushed I mean that score was 20:47 with 33 draws.
Version with a 'bug' in a pawn shelter had 33:31 with 36 draws.
Games were at 40/3' on AMD3000+

I did a bullet match last night at 40/1' with the 'bugy' version and some additional modifications and the score was 28:38, so I guess I still have to feed Naum more spinach if I want to beat Fruit.