Vas Rajlich Video...

[Rolf]

The problem is that Rybka will not cater to its core users, which like to have the engine loaded in their home, physically on their computer.

All wrong. It sounds as if you or Mig were no real chess players. No matter of what level.

Let me explain why you both are wrong with almost all others too. Main reason the profan buttonpushing which isnt chess at all.

Let me rehash what chess is and well below master level in my case.

Look at this:

when I was young I played and analysed our games with my dad. Then I went by train to Switzerland and bought my first journal the Deutsche Schachzeitung at the station. Man, that was like Easter and Xmas on the same day. These little commentaries from the editors or in exclusive mode by the GM in person, that made my day. I tried to play like them - of course with little chance because my dad always followed other plans.

Later we went into a club. Man what a shock that was. Suddenly I lost confidence that I understood sometting in chess. They played the lines that have been commented in the paper but I hadnt analysed all this at home. Then I visited the team events and studied the play of the first boards.

I observed veritable masters, but had no idea how they always managed not to lose - what became a usual fate to me when I climbed the ladder of strength. In short there was always someone who knew more about it than what I could play.

Other than true talents I never had a IM or GM explain something to me.

And that continued with the machines. I played with something weaker while others operated their tuned stuff that costed several thousands and the best were standing in math institutes and couldnt be bought at all.

Now for the first time, after over a decade of software updates I own the Wch programm - BUT the gap was still there because I never had the horse power that were the main source of success for the different progs.

For my job I didnt need a pentium at all. And for chess I should get the always best hardware now? Never in my life. I mean I have a life besides chess.

Now listen closely. I am still addicted to chess if I could afford the best possible combination in soft and hardware for the analysis of my own games. I want to get it. But not on the basis of always new hardware investings.

So, for me I promise you that I will loan the analyses time online to get my own games analysed at a best possible level. Call me an idiot but I'm no GM, not even active anymore, but I would pay hundreds of € to get deeper insight in what I playedmyself when my dad was my opponent because I recall every idea of that time. Was it really that stupid or was it brilliant but lost due to a later oversight? And I have played almost 2000 games!

Now the moral of it all: I am not the only one who wanted to know a really deep comment on my own play. But I admit that mere button pushers and the pompous freaks who dont play chess at all, they have no history to get evaluated. They are just operators without a life in the past. Living deads.

But a true chessplayer like me has 1 million positions with all the ideas and dreams and memories, fears and hopes, bloodpressure changes, hunger and thirst. And rocknroll...

[Robert Flesher]

when he got started Fruit was not available as source.

Yes, thats true. When he started he wrote a program that was at the 2000 ELO border. It was completely different to the actual Rybka, for example it had MTD(f) search.

Then came Fruit.

Then came Rybka 1.0 Beta, that was _very_ similar to Fruit, and not to former Rybkas.

Exactly my point!
Without Fruit it appears his engine would still be a 2200 engine. Yet nobody even mentions Fabien as the true genius.

[Robert Flesher]

I disagree as this increase in Rybka's strength has been methodically linear and if anything, Rybka's real strength jump seems to be in Rybka 3 and not Rybka 1.0 beta.

It astonishes me how you are able to disregard facts completely, as if you lived in a dream and want the others dream with you.

Here is an excerpt from CCRL blitz:

Code: Select all

Rybka 3 64-bit          3176	
Rybka 2.3.2a 64-bit     3072	
Rybka 2.2n2 64-bit      3025	
Rybka 1.1 64-bit	     2989
Rybka 1.0 Beta 64-bit	2938

It is clear that Rybka 3 jumped 100 elos from Rybka 2, and roughly 240 elos relative to 1.0 beta. This jump took 3 years.
The beta is at least 700 elo stronger than the Rybka from 2004. The increase took 1 year.

You call this linear.

What is this Rybka engine that is 700 elo weaker than Rybka 1 called, and where can I obtain it?

Another who sees the truth, finally the Vas lovers are being met by logic.

[mcostalba]

when he got started Fruit was not available as source.

Yes, thats true. When he started he wrote a program that was at the 2000 ELO border. It was completely different to the actual Rybka, for example it had MTD(f) search.

Then came Fruit.

Then came Rybka 1.0 Beta, that was _very_ similar to Fruit, and not to former Rybkas.

Exactly my point!
Without Fruit it appears his engine would still be a 2200 engine. Yet nobody even mentions Fabien as the true genius.

Perhaps he missed to read a couple of papers before Rybka 1.0


It is just a joke, please don't get hungry, I apologize in advance. It is just I'm in a good mood this sunday.

[slobo]

when he got started Fruit was not available as source.

Yes, thats true. When he started he wrote a program that was at the 2000 ELO border. It was completely different to the actual Rybka, for example it had MTD(f) search.

Then came Fruit.

Then came Rybka 1.0 Beta, that was _very_ similar to Fruit, and not to former Rybkas.

Exactly my point!
Without Fruit it appears his engine would still be a 2200 engine. Yet nobody even mentions Fabien as the true genius.

Perhaps a 2300 engine, because we should not forget that Vas is a talented programer.

[slobo]

I am willing to keep an open mind about remote Rybka.

I think whether this is a viable model or not comes down to pricing. If you want to use Rybka 24/7 then $750 a year may be a lot. It is probably not a lot if you are a chess professional, but if you're on playchess all the time with an engine just for fun, then your hobby just got more expensive. So it remains to be seen how large the market is for the remote service, at the proposed pricing (which could still change I assume).

Personally I use Rybka mainly for game and position analysis and if I could plug into the remote version from ChessBase I'd probably be fine with that, and I wouldn't run it 24/7. So that may be an ok model for me.

It is an interesting point that, while not the primary motivation for Vas apparently, this model also prevents reverse engineering and piracy.

--Jon

If you really use it 24/7, you have to take into account up to date hardware. Renting the fastest quad for a year, to run the last version of the top engine, and access it from everywhere, is not absurd. Hardware updates, electricity + software (if you buy a computer many people end up buying antivirus etc.).

For a professional, buying it per hour (0.50 Euros) may make sense. The main concept is sharing hardware.

Miguel

The concept of sharing software is much more atractive, because it is much cheaper. You are probably aware of the importance of the economic factor.

[Milos]

Perhaps a 2300 engine, because we should not forget that Vas is a talented programer.

Talented MIT programmer, please don't forget that, otherwise our Rofl will be sad...

[Robert Flesher]

Well if he had just picked up Crafty and used it as the basis of his code then he would have started with a 2600+ ELO engine to start with. He obviously decided to go a different way. I mean my God, do you really think Rybka is only strong due to Fruit ??? I think any reasonably gifted programmer can create a chess engine that will equal the strength of the strongest freely available source code by simply using the ideas gleaned from the code ... maybe he would raise some eyebrows if he managed to increase the playing strength even slightly higher than that. What I think is truly "astonishing" is moving this engine into the realms of what was once thought impossible. I can't help but feel there is a tinge of jealously by other programmers when it comes to Vas ... after all how can someone fresh from MIT simply go into a field that hundreds of extremely talented people have spent thousands of hourse on and trump them so quickly. Of course you can blame it on Fruit code, but be honest with yourself ... how much do you think Fruit helped Rybka ?? Vas says it is 20 ELO ... and you probably would laugh at that. But then again you would probably have had the same laugh if someone would have told you that he could improve Rybka 1.0 beta by a couple of hundred ELO points within 2 years.

I do not agree at all. I think a strong player like Vas can tweak Fruit's search and evaluation and get HUGE gains. Weaker players/authors will not see the problems an international master will see, and with the help or several GM/IM's.

I am only a 2000-2100 player at best, however, I will not forget what the now deceased IM Bryon Nickoloff said about Shredder after a few drinks and playing several matchs with me. " Shredder is a true genius, compared to your other engines, but it is clear the author lacks the understanding of many positions. With my help this thing would be unstoppable". Now remember Bryon was only an IM, but this was due to poor health habits and alcohal. However, Paul Keres once said " Bryon Nickoloff has the potential to be a world champion, such is his genius"

Strong player + talented programmer + Fruit source = Rybka (the genius !?).

I am not stating Rybka is a clone, far from it. But, Fabiens work certainly should be mentioned. It was proven that Rybka Beta 1.0 had alot of Fruit ideas.

[jdart]

The concept of sharing software is much more atractive, because it is much cheaper. You are probably aware of the importance of the economic factor.

If the "cheap" remote service was really better than a typical home quad machine, or even up to the level of an 8-core as Vas seemed to be saying, then yes, it is more economical to buy the service vs. buying a machine + software (my home-assembled quads were around $750-$800 apiece, just for hardware). Although even then you have to factor in that computer chess enthusiasts have already bought hardware, so they're not starting this evaluation from scratch.

But I'm a bit dubious about the cost/performance of a remote service. I tried Amazon's EC2 compute service recently and their "high-CPU" configuration with 8 virtual processors was a lot slower than my dedicated quad. It's also not typical for virtualized environments to provide a lot of RAM per instance, although Amazon can do this.

In addition, unlike most EC2 users, Rybka users would be consuming 100% of the allocated CPU during analysis or game play. So it's challenging I think to provide a service that matches dedicated on-site hardware in performance, with cost that is lower than the hardware itself.

Most users who are getting cost savings from virtualization are doing it because their existing environments are under-used and over-provisioned (so they have more hardware than they need) - they can share it without loss of performance. But for chess, I ping my local CPUs at 400% usage (4 cores x 100%) all the time, for hours. So to share my box with someone else the performance will suffer.

[Harvey Williamson]

If the "cheap" remote service was really better than a typical home quad machine, or even up to the level of an 8-core as Vas seemed to be saying, then yes, it is more economical to buy the service vs. buying a machine + software (my home-assembled quads were around $750-$800 apiece, just for hardware). Although even then you have to factor in that computer chess enthusiasts have already bought hardware, so they're not starting this evaluation from scratch.

It also presumes you are only interested in running 1 engine at a fast speed. I want to run many engines at their full speed and compare them. Impossible with the Rajlich model.