Super settings for Hiarcs 12....

[Dr.Wael Deeb]

http://schachcomputerwelt.foren-city.de ... iment.html

Can someone enlight us regarding this issue,what are these setting and are they realy that strong

[Harvey Williamson]

http://schachcomputerwelt.foren-city.de ... iment.html

Can someone enlight us regarding this issue,what are these setting and are they realy that strong

This thread maybe the best place to ask in http://hiarcs.net/forums/viewtopic.php?t=1166

[mclane]

http://schachcomputerwelt.foren-city.de ... iment.html

Can someone enlight us regarding this issue,what are these setting and are they realy that strong

hi - i am testing this in the moment.
i did some first testings. the results were promising.

[Tony Thomas]

Thorsten, are you ready to post the settings yet?

[mclane]

in the moment i am trying to find out which of the parameters are in charge for the strength increase.
that will take a time.

feel all free to do your own changes. IMO hiarcs is a platform for styling experiments.

[Dr.Wael Deeb]

in the moment i am trying to find out which of the parameters are in charge for the strength increase.
that will take a time.

feel all free to do your own changes. IMO hiarcs is a platform for styling experiments.

Thanks guys for the answers

[fallstalk]

once I found engine parameters on my H12sp from ChessBase I began experimenting... I am starting to win or draw against Rybka 2.3.2a (or whatever it is...) and they (the competition) usually have some steroid injected hardware compared to my lil ole Gateway desktop... HIARCS 12 plays like an M1 tank

[Seb1]

Hi there, I bought Hiarcs 12 the first day it came out, and since then I have tested every possible setting, I have played 1500 games on playchess with H12. (only playing 3 minute games).

I really love the way H12 plays when it wins, some nice games with win in less than 35 moves against the top computers/bookmakers there.

The problem is that I have not been able to find any settings that make it possible for me to match the rating I get when I use Rybka2.3.2a. ( have been up to 2960 with rybka, and can stay at 2770-2830 stabile with my book and open formula)

after 1500 games and with the best settings I managed to come up with I finaly ended up with 2600 rating with H12 (Hiarcs book, tested my book also, and have the same preformance more or less(did have a peak up at 2850 with H12 and my own book).

I desided to go offline and test H12 more. So I read up here, and created my own tournament with Rybka2.3.2a vs Hiarcs 12 mp.

I have a q6600 (oc'ed to 3,3Ghz) 2048ram.
For the tests I use the Nunn2 test possisions I have permanent brain off.
4 cores to each engine 256 hash.

I have played almost 1500 games (time=3 & inc=0)
I have set rybka to default.
For H12 I have tested EVERY prossible settings twisted and tuned them. played 10 games, tuned more, and tested again.

So far H12 does score 15-20% at the best.
The games where H12 seems to be able to win, I seem to end up in time troubble, and lose on time or because of the time trouble.

There must be some thing I do wrong with default H12 settings it does not get better score, I'd love to get my H12 strong enough to compete against rybka

So if there is any settings that makes H12 stronger I'd love to test them.

Sorry for the bad english.

Regards
Seb

[PortCitySlim]

after 1500 games and with the best settings I managed to come up with I finaly ended up with 2600 rating with H12

I run an AMD X2 TL-64 2.2ghz and have got my rating up to 2616 on 3min games. If your running a quad core you should try changing FP10-FP11, thats what I did and after 40+ games with that setting I improved my playchess elo 150 points. Everyone says that the default settings are the best but personally I like Forward Pruning set to 11 instead of 10, in the future I plan to test the two settings against each other at 3min time controls to see for myself if its for real.

[Seb1]

I have tested several 100 games with FP 11.

still same result.

So there must be something else

[mclane]

??

i wonder why you are not able to produce a style that works.
in my posting

you can see how i did it.

http://schachcomputerwelt.foren-city.de ... ment.html

i began with 4 or 5 different styles. let them gauntlet rbyka.
took the style with the highest chances and came to my
best style.

this was not very complicate.

to make sure no opening book was in charge i took 91 positions,
swapped sides per game and got 182 games where hiarcs won against rybka.

this was done on intel core2quad with 4 cpu's for each program, permanent brain off.

now you post you did thousands of games and you tried out all kind of styles?!

do you have an idea how many combinations of styles you can make with hiarcs parameters. your statement that you tried them all looks a little "strange".

there are plenty of combinations of parameters.

maybe you did not try the most interesting.

try first with DIFFERENT styles. lets say 5 completely different ideas of styling.
see which one gets the highest scores.
than make clones of this style, but with little changes.
again look which clone is the best.

[Seb1]

Yes I did play 1500 games on playchess with H12, there I only tested :
active/aggressive/solid settings and mixed with H12 book on/off
and FP 9-10-11.

I also did some toying with sharpen PV on/off and Optimistic Search on/off.

I then went offline, and have tested some 1000 ++ ish games. vs rybka 2.3.2a.

here I have tested with different settings, but I have played too few games to say that I have tested EVERY thing good enough.

So I have done 10 games with different settings, and when score is worse than 25% score for H12, I change some settings and test again.

I will have to say that with H12 book on (and default settings), and random book for rybka I get good results for H12. But then again with rybka and my book, H12 get bad score again.

So there must be something I have missed.


I would love to understand your forum mclane, but I dont undertand german.

with "style" what do you mean?

might I ask what Playing Style (solid/active/aggressive) you find best against rybka?

Seb1

[mclane]

style is a combination of parameters that can be adjusted in
hiarcs to do a personality.

playing style such as active / aggressive or solid is only ONE little choice you can adjust.

[Eelco de Groot]

I think a possible explanation is that on Playchess you play against Rybka(Rybkas) with a highly optimized killer book, and it is well known that there are openings that Rybka does not like and would rather not play. Thorsten has tested mostly with general Arena books, Hiarcs with its own book. There is a difference. The only other thing I can think of is maybe the processor but that does not seem very likely, and probably there only very few styles that really work. Also the tournament results of Rybka are just generally worse than head to head matches (not for every program of course that would be mathematically impossible).

Eelco

[fallstalk]

yes, very interesting... but I can assure you that last night I drew a G120 against a PlayChess veteran (who spends most of his time tweaking his book in the Engine Room and also used some sort of Quad 4 Hyper Speed hardware against my lil ole Gateway 500x... *my secret weapon was HIARCS 12 SP.... that's Human Intelligence Auto Response Chess System 12 sp. Mark, Harvey, et al, If you can get me HIARCS 13 by July 31 I'll pay you double. If you can get me HIARCS 14 with state of the art gui by Christmas then you can just name your price ...