Winboard Forum

Hello!
After being away for a long time I started to look at Alaric again (Terra is forever freezed).
What has happened lately?
Who are the 3 top amatueur engines these days?
Any new chess programming idea to look closer at?



/Peter

Hello Peter,

Welcome back! I was wondering where you had gone ...

Peter Fendrich wrote:Hello!
After being away for a long time I started to look at Alaric again (Terra is forever freezed).

How soon can we expect a public release of Alaric?

What has happened lately?

The biggest news is that Rybka has improved by something like 500-600 Elo points to secure a very clear first place on all rating lists. I suppose you have already noticed this.

Who are the 3 top amatueur engines these days?

Depends on whether you count the various Fruit-derivatives as a single program or several programs, on whether you consider performance on one or on two CPUs, and on whether the free demo version of Rybka counts as an amateur engine. Arguably the best resource we've got for comparing playing strengths is the CEGT rating list.

Any new chess programming idea to look closer at?

Depressingly little, as far as I can remember.

Tord

Hello Peter,

No answers to your queries - Tord is best here - but it is good to see you back.

As for Alaric, let the little beastie go will you!?



Later.

I think one trend worth noting is that bitbases have become more popular.

Robert

Hi Peter,

well it depends on how long you took the break

So called late move reductions, some based on history scores (history pruning), some based on other move properties or combining both seem to be quite successfull for most. The hardware development is also quite interesting, current dual-core- and future quad-code-cpus. Intel seems to bear down the P4 but a PPro reincarnation with the new woodcrest. Amd has announced the K8L with popcnt and lzcnt instructions and 128-bit SSE2 Alus for faster and more parallel fillstuff and dot-products. Multiplication has become quite cheap already with K8, 3 or 4 cycles direct path - so you may give antirotated bitboard approaches a try.

Cheers,
Gerd

Tord Romstad wrote:Hello Peter,

Welcome back! I was wondering where you had gone ...

Yes, sometimes I wonder myself...!

Peter Fendrich wrote:Hello!
After being away for a long time I started to look at Alaric again (Terra is forever freezed).

How soon can we expect a public release of Alaric?

I see no reason to wait but maybe I would like a few testers first before throwing it to the wolfs!

What has happened lately?

The biggest news is that Rybka has improved by something like 500-600 Elo points to secure a very clear first place on all rating lists. I suppose you have already noticed this.

Wow that Rybka is a real monster. Do you know anything about this program.

Who are the 3 top amatueur engines these days?

Depends on whether you count the various Fruit-derivatives as a single program or several programs, on whether you consider performance on one or on two CPUs, and on whether the free demo version of Rybka counts as an amateur engine. Arguably the best resource we've got for comparing playing strengths is the CEGT rating list.

Any new chess programming idea to look closer at?

Depressingly little, as far as I can remember.

Good, it is like Dallas or some other soap. You can miss 30 parts and come back later without anything really happened.

Tord

Roger Brown wrote:Hello Peter,

No answers to your queries - Tord is best here - but it is good to see you back.

As for Alaric, let the little beastie go will you!?



Later.

Well Roger....
You did some good testing of Terra, why not Alaric?
Just a humble question...
/Peter

Gerd Isenberg wrote:Hi Peter,

well it depends on how long you took the break

So called late move reductions, some based on history scores (history pruning), some based on other move properties or combining both seem to be quite successfull for most. The hardware development is also quite interesting, current dual-core- and future quad-code-cpus. Intel seems to bear down the P4 but a PPro reincarnation with the new woodcrest. Amd has announced the K8L with popcnt and lzcnt instructions and 128-bit SSE2 Alus for faster and more parallel fillstuff and dot-products. Multiplication has become quite cheap already with K8, 3 or 4 cycles direct path - so you may give antirotated bitboard approaches a try.

Cheers,
Gerd

Hehe, maybe 6-12 month but my memory is not that good either.
You mean the good old History pruning?
Hardware is not my strong part in case I have any stron part at all...
Thanks!
/Peter

Peter Fendrich wrote:

Tord Romstad wrote:How soon can we expect a public release of Alaric?

I see no reason to wait but maybe I would like a few testers first before throwing it to the wolfs!

Good! I hope you remember to release a version for Mac OS X.

What has happened lately?

The biggest news is that Rybka has improved by something like 500-600 Elo points to secure a very clear first place on all rating lists. I suppose you have already noticed this.

Wow that Rybka is a real monster. Do you know anything about this program.

Unfortunately, not many technical details are known. I am fairly sure Rybka was active already at the time you were busy with Terra, but at that time Rybka was just an average amateur engine, and you could easily miss it among the countless other engines of similar strength. Nevertheless, I had very high expectations for the program, based on the obviously very high chess and programming skills and the ambitions on the author (Vasik Rajlich, who at that time was rather active in the technical discussions on the CCC). I was very dissappointed when the author and his program just disappeared (very much like you and Alaric, when I think about it), and I feared that Vas had started spending time on something fun or useful instead of computer chess, or something similarly terrible.

Near the end of last year, it turned out that Vas had just been hiding deep underground for a while, and never stopped working on his program. All of a sudden, he suprised us all by releasing by far the strongest chess program anyone had seen.

Any new chess programming idea to look closer at?

Depressingly little, as far as I can remember.

Good, it is like Dallas or some other soap. You can miss 30 parts and come back later without anything really happened.

I've never owned a TV and therefore can't confirm your opinion about Dallas, but it is certainly true about computer chess.

Tord

Tord Romstad wrote:

Peter Fendrich wrote:Good, it is like Dallas or some other soap. You can miss 30 parts and come back later without anything really happened.

I've never owned a TV and therefore can't confirm your opinion about Dallas, but it is certainly true about computer chess.

Tord

I am not sure I would subscribe to this point of view: Since I started YABRL 3 years ago the absolute top has moved up by 150 points, the top amateurs probably by 200 points. This is an impressive development, but apparently it has been achieved more by technical evolution rather than a revolution.

Robert

Robert Allgeuer wrote:
I am not sure I would subscribe to this point of view: Since I started YABRL 3 years ago the absolute top has moved up by 150 points, the top amateurs probably by 200 points. This is an impressive development, but apparently it has been achieved more by technical evolution rather than a revolution.

Robert

How much of that do you think is from faster hardware?
I mean 3 years must mean at least double speed.
/Peter

Peter Fendrich wrote:

Robert Allgeuer wrote:
I am not sure I would subscribe to this point of view: Since I started YABRL 3 years ago the absolute top has moved up by 150 points, the top amateurs probably by 200 points. This is an impressive development, but apparently it has been achieved more by technical evolution rather than a revolution.

Robert

How much of that do you think is from faster hardware?
I mean 3 years must mean at least double speed.
/Peter

Hi Peter,

Welcome back!
AFAIK Robert hasn't changed his hardware during that 3
years, thus nothing of the improvement mentioned is
because of hardware speed. It's still the same Athlon 1.1Ghz.
IMHO the improvement is still a post-Fruit (and younger
very strong open source programs, as Glaurung and Scorpio)
consequence.

Best regards,
Guenther

Yes, hardware is unchanged and plays no role in these improvements. Before the advent of List 5.12 and Ruffian 1.0.1 essentially no free engine was above 2600, in fact the best free engines were around 2550 to 2580 (Crafty, SoS, Aristarch, Little Goliath, Yace etc.). Now Toga and the free Rybka are >2800, Spike and free Fruit are >2700, ProDeo and Glaurung are close to 2700 and a large number of free engines are >2600.
The top was probably around 2750 then (Shredder 7.04 and 8, both not tested in YABRL). Now commercial Rybka is close to 2900 and several further engines are also >=2800 (Hiarcs, Shredder, Toga and Fruit).
I share the view that Fruit had a catalysing effect in this development. In any case - even without revolutionary new techniques - the progression due to continuous improvement of the software is remarkable.

Robert

Good to see you back Peter. I remember that you used to visit WBEC site regularly. Your engine has been sliding down the list slightly every edition. I thought about Terry few days ago, as it is in 21st place on my blitz 2nd division list and would soon have to move it down to 3rd division due to the new strong engines. Well, good to see that you are back in to programic, I wish you the best of luck with Alaric, I liked the previous name better as it sounded like Terror.