How many "official" Toga projects there is now?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
playjunior
Posts: 338
Joined: Thu Jun 21, 2007 10:53 pm

Re: How many "official" Toga projects there is now

Post by playjunior » Thu Jan 24, 2008 1:55 pm

Rolf,
Fruit is a very complex engine. Its source is open for a long time, yet small fraction of computer chess people understand the code.
Fruit advantages are, as far as I understand:
1. Stability
2. Fast and very good eval function, which is very well-tuned.
3. Great search function, with many different cutoff techniques, that actually work.

Note that a good search heuristics generally is dependent on your eval function. You change the eval-the old heuristics might become not working. You have to understand the eval, the search, and tune them together.

Now let's me try to explain what Thomas is saying.
The eval has been changed. The engine has become bitboard. If you change something like that, you introduce bugs, and you lose stability. You have to test and debug. New Fruit is rather stable, so Ryan has done a fine job there.
Then, he changed the eval, which means he had to somehow change the search also. So he has changed something in Fruit that so many have copied, (claiming it's nearly perfect-mind you, that's the eval function).

Huge amount of job has been done. The new engine is stable, plays better. Ryan has, in fact, made such big changes that he could not fully benefit from biggest advantages Fruit had-points (1)-(3). Go figure yourself how much effort would be needed to do this.
Thomas says-not less than you need for a new engine.

Hope this helps.

User avatar
Rolf
Posts: 6081
Joined: Fri Mar 10, 2006 10:14 pm
Location: Munster, Nuremberg, Princeton

Re: How many "official" Toga projects there is now

Post by Rolf » Thu Jan 24, 2008 2:30 pm

playjunior wrote:Rolf,
Fruit is a very complex engine. Its source is open for a long time, yet small fraction of computer chess people understand the code.
Fruit advantages are, as far as I understand:
1. Stability
2. Fast and very good eval function, which is very well-tuned.
3. Great search function, with many different cutoff techniques, that actually work.

Note that a good search heuristics generally is dependent on your eval function. You change the eval-the old heuristics might become not working. You have to understand the eval, the search, and tune them together.

Now let's me try to explain what Thomas is saying.
The eval has been changed. The engine has become bitboard. If you change something like that, you introduce bugs, and you lose stability. You have to test and debug. New Fruit is rather stable, so Ryan has done a fine job there.
Then, he changed the eval, which means he had to somehow change the search also. So he has changed something in Fruit that so many have copied, (claiming it's nearly perfect-mind you, that's the eval function).

Huge amount of job has been done. The new engine is stable, plays better. Ryan has, in fact, made such big changes that he could not fully benefit from biggest advantages Fruit had-points (1)-(3). Go figure yourself how much effort would be needed to do this.
Thomas says-not less than you need for a new engine.

Hope this helps.
Adranik, that's grat so far, but what about the question of honor as a man, you know what I mean? I mean we all learned it in school when we were young, isnt it worthier to write on your own than taking one of the already strongest and after 4 years you rewrite into bitb. and change what you have described? Vas is a different. He took some ideas from Fruit but NOT the whole code to just re-write it? This is correct, yes?

If Ryan is so eager to work on a directly high level, why not asking Vas to be his stabilizer and co-writer? But taking the whole code from Fruit? How can a reasonable man can presume that he's now the chief programmer?

I mean it's still a great performance, but where is the beef? IMO here someone tries to avoid the hard road all the others have because we could now expect that 300 programmers worked on the Fruit code - would that be funny?

You must know that I come out of the older generation when several paradigmata existed. Are they doing all the same now? Or asked this: is it now only about re-compiling and stealing, sorry, stabilizing etc?
-Popper and Lakatos are good but I'm stuck on Leibowitz

Thomas Gaksch

Re: How many "official" Toga projects there is now

Post by Thomas Gaksch » Thu Jan 24, 2008 2:52 pm

Rolf wrote: Or, please help me out, is everyrthing in computerchess programming ALWAYS only about transforminf or translating or tuning here and there, technically spoken?
NOT ALWAYS. But very much. And for example tuning of the eval is very important. And in case of Fruit there is nothing really new.
Rolf wrote: Are there never new IDEAS? Because these, if ever appearing, are already "stolen" minutes afterwards because everybody is always and continually compiling the codes of all serious competitors?
Of course there are a lot of new ideas. And a lot of people share these ideas with others in this forum or distribute their source code. Sometimes i have the feeling that distributing strong engine source code is not wanted by some people because this can produce other strong engines. Also the most new ideas do not help for playing strength. believe me or not it is not simple to increase the playing strength of an chess engine although it is only a clone.
I have high respect for Vas, because he had very good ideas else Rybka wouldn´t play so strong. i hope his secrets will not be fully disassembled. But i am also sure that some developers have tried it. not only the Strelka author.

Guetti

Re: How many "official" Toga projects there is now

Post by Guetti » Thu Jan 24, 2008 3:03 pm

Thomas Gaksch wrote:Congratulations Ryan,
this redesign was the right step to be sucessfull for the next years. That was really a lot of work for you. but a great decision to do it.
Now in practise Fruit isn´t anymore Fruit 2.1 and in my opinion Fruit is now your own baby. Why didn´t you rename it?


Thomas
It thought it was? Isn't it called Chess64 now or something?

Thomas Gaksch

Re: How many "official" Toga projects there is now

Post by Thomas Gaksch » Thu Jan 24, 2008 3:13 pm

Hello Andreas,
as far as i have understood is that chess64 is only developed by Fabien and Fruit only by Ryan anymore. But Ryan knows it best.

User avatar
Roman Hartmann
Posts: 295
Joined: Wed Mar 08, 2006 7:29 pm
Contact:

Re: How many "official" Toga projects there is now

Post by Roman Hartmann » Thu Jan 24, 2008 3:22 pm

Rolf, I try to make it simple as you're not familiar with the technical aspects of chess programming itself.

I'm planning to participate at the Berlin marathon this year. Unfortunately I don't have much time for training, so I think I will take the metro for the first part and run the last mile only. Running the last mile in a marathon is very hard as you can imagine and that's why I only run that last mile. Did I already mention that this last mile is very hard in a marathon?

I mean, let's face it, it's boring to run 40 Kilometers before you can actually start that last and important part of the race. Everybody can run 40 kilometers, but it's the last mile which actually counts. So running the last mile is actually the key part of it.

Roman

Uri Blass
Posts: 8611
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: How many "official" Toga projects there is now

Post by Uri Blass » Thu Jan 24, 2008 3:26 pm

Thomas Gaksch wrote:
Rolf wrote: Or, please help me out, is everyrthing in computerchess programming ALWAYS only about transforminf or translating or tuning here and there, technically spoken?
NOT ALWAYS. But very much. And for example tuning of the eval is very important. And in case of Fruit there is nothing really new.
It is still possible to write a different evaluation based on different factors.

For example mobility can be defined in a different way and not as number of squares that a piece controls.

It may be interesting if people do not try to improve fruit's evaluation
but try to do total replace of it by something better but of course you need also to write code that does fruit's evaluation for comparison in order to know that what you do is better.

Uri

Guetti

Re: How many "official" Toga projects there is now

Post by Guetti » Thu Jan 24, 2008 3:41 pm

Thomas Gaksch wrote:Hello Andreas,
as far as i have understood is that chess64 is only developed by Fabien and Fruit only by Ryan anymore. But Ryan knows it best.
Ah, ok. I probably associated Chess64 somehow with 64-bit.

Then:
Chess64 (Fabien) is a rewritten, multithreaded Fruit?
Fruit (Ryan) is a modified Fruit using bitboards and a rewritten eval?
Toga is a modified Fruit with tuned evaluation and SMP?

And the version playing in CCT10 is the version of Ryan? Sorry, I'm a bit lost with all the versions.

User avatar
Rolf
Posts: 6081
Joined: Fri Mar 10, 2006 10:14 pm
Location: Munster, Nuremberg, Princeton

Re: How many "official" Toga projects there is now

Post by Rolf » Thu Jan 24, 2008 4:29 pm

Roman Hartmann wrote:Rolf, I try to make it simple as you're not familiar with the technical aspects of chess programming itself.

I'm planning to participate at the Berlin marathon this year. Unfortunately I don't have much time for training, so I think I will take the metro for the first part and run the last mile only. Running the last mile in a marathon is very hard as you can imagine and that's why I only run that last mile. Did I already mention that this last mile is very hard in a marathon?

I mean, let's face it, it's boring to run 40 Kilometers before you can actually start that last and important part of the race. Everybody can run 40 kilometers, but it's the last mile which actually counts. So running the last mile is actually the key part of it.

Roman

I got the message! And dont forget that on the final kilometer they have all the videocams. While in the sub it's dark and boring. Look, if you gave me your dark sun glasses 100 meters before the Finish, I could also run the 40 km marathon of Berlin! That would be great for my development as a professional runner. BTW did you know that in 1996 I already won the IronMan on Hawaii and got the IGM title? The members of usenet were kind of sceptical and they doubted that I could really be an International GrandMaster in chess... You knew that, in the world of the one-eyed, the blind one is King.

Oder wie Dittsche zu sagen pflegt: Das is ne Weltidee, mein Ingo! <g>
-Popper and Lakatos are good but I'm stuck on Leibowitz

Tony

Re: How many "official" Toga projects there is now

Post by Tony » Thu Jan 24, 2008 5:17 pm

Roman Hartmann wrote:Rolf, I try to make it simple as you're not familiar with the technical aspects of chess programming itself.

I'm planning to participate at the Berlin marathon this year. Unfortunately I don't have much time for training, so I think I will take the metro for the first part and run the last mile only. Running the last mile in a marathon is very hard as you can imagine and that's why I only run that last mile. Did I already mention that this last mile is very hard in a marathon?

I mean, let's face it, it's boring to run 40 Kilometers before you can actually start that last and important part of the race. Everybody can run 40 kilometers, but it's the last mile which actually counts. So running the last mile is actually the key part of it.

Roman
:)

You forgot to mention how stupid the other people are for running these first 40 km.

And the others keep complaining. They are just jalous on how fast you run this marathon and they only wished they could run it so good as well.

Tony

Post Reply