Robodini

[Don]

I'm just trying to determine whether Houdart has made an original contribution of importance or just combined the best of Stockfish and Robo plus some tuning.

Do you understand that there's no difference at all between the two? The only difference is on "knowledge parameters" for those that could benefit from the same, as chess engines authors in this case; philosophically (and so for ascertain the "revolutionary" status or not) there's no difference at all between the two. Re-read what I wrote about connections.

It doesn't matter at all the way you construct the connection (modus operandi, approach etc.) and how "large" it is, if there's a connection where there was none before by any other in any other way, then it's already a "revolutionary" idea, because before that idea others couldn't do the same and could not achieve the same nor think about a way to fill that gap.

If he combined ideas from Robbo and Stockfish in a certain peculiar way that nobody can understand or if he implemented something more and specific himself or something in between makes no difference at all on the status of the idea; it is in both cases a NEW creation, because it is a connection where before there was none.

I disagree. Combining ideas from other engines may be good engineering, but doesn't require any original thinking, just some hard work. Maybe nobody got around to trying some specific Stockfish ideas in Ippo code before Houdart did, either because it's a lot of work and testing or because they didn't judge the likelihood of success to be worth the trouble. He deserves credit for taking the time to find out which Stockfish ideas work in Ippo, but he would deserve a lot more credit if the gains came from some previously unknown idea. I'd like to know which is the case.

I've ranted on this before, but very few people really understand what is involved in producing a strong program. The average Joe has a very inaccurate "cognitive model" of the process based on conventional wisdom and their own intuition which is completely wrong.

The correct model is that it's 98% hard work and good engineering. I can give you anecdotal evidence of that.

How long has Robbo and Stockfish sources been available? Several years, right? How many program authors have been able to PASS these programs using these freely available ideas and given several years of time to do so since they came out? Now of those programs, how many are NOT based on the Robbo family sources? All I hear is a deafening silence because the answer is hardly any. Maybe 2 or 3.

The reason you don't see this is that producing a program the strength of Robbolito or Stockfish does not have that much to do with the ideas that have been freely available to literally HUNDREDS of chess authors for years. That is the assumption and myth that has been causing so much twisted reasoning.

Are ideas important? Of course they are. But every program author has access to the same major ideas and everything else is just background noise and stuff to get excited over. But the sad truth is that hardly anyone has been able to produce a strong program without starting from a complete product that is already at the top! How obvious can this be? Hello!!!!

[Uri Blass]

I see no proof for the claim that
producing a program the strength of Robbolito or Stockfish does not have that much to do with the ideas that have been freely available to literally HUNDREDS of chess authors for years.

How many programmers really read and understood all the ideas that are freely available?

I certainly did not read part of the ideas in other programs in the time that I devloped movei.

I believe that it is the same for most programmers that did not do it because of at least one of the following reasons:

1)They did not have time for that
2)They did not have an interest to spend time in order to understand all the ideas in the free source of Robbolito or Stockfish.

[Don]

I see no proof for the claim that
producing a program the strength of Robbolito or Stockfish does not have that much to do with the ideas that have been freely available to literally HUNDREDS of chess authors for years.

I never offered any proof. Where did you see that?

I called this "anecdotal evidence" because I am not trying to make up facts that cannot be 100% backed up. See my post on that.

It's entirely possible that nobody cares enough about writing a strong program to look at what is being done. I just don't personally believe that to be the case.

How many programmers really read and understood all the ideas that are freely available?

I certainly did not read part of the ideas in other programs in the time that I devloped movei.

I believe that it is the same for most programmers that did not do it because of at least one of the following reasons:

1)They did not have time for that
2)They did not have an interest to spend time in order to understand all the ideas in the free source of Robbolito or Stockfish.

I agree. A lot of programmers do this for fun and are not interested in looking at other peoples source code. Actually I don't like doing so myself.

However, several years have gone by and wouldn't you think that at least ONE really motivated programmers would care? I know a lot of professional people and most of them (if they are not burned out) want to keep up with the cutting edge of their profession. They keep up. So you seem to believe that I am the only one that tries to keep up (not counting Stockfish?) There are something like 1200 programs rated by CEGT and although some of them are just different versions of the same program, the majority are one-shot. We are talking about hundreds of programmers and only the ones that actually are being rated by CEGT. Many more are private and do not go out to the agencies.

So what is easier for you to believe? That all anyone has to do to produce a strong program is to pick out all the tricks from the sources and "plug" them into their own program and bingo! Or that it's actually hard to build a really strong program no matter what you do short of copying? I believe the latter and that is the real point I am trying to make.

My evidence is anecdotal to be sure, but it really stretches the limits of credulity to think nobody bothers to do this. The "conventional wisdom" is that this is all you have to do to produce a world class program, just start with your own crappy program and plug in all the "ideas" from Robbolito. That just really stretches the limits of credulity.

What is FAR MORE LIKELY is that a lot of programmers, even the fun hobby type have taken these ideas and made some improvements as a result but that it was not enough to automatically give them the 500 ELO boost that people believe Ippo is responsible for in computer chess. The big myth is that it's all about taking ideas and that hard work has very little to do with it.

And by the way, you can go over the Robbo or Stockfish sources in a day and anyone capable of writing a chess program can pick up the major ideas that make any real difference pretty quickly. If you have a 2000 ELO program it's not really going to help you that much because it is STILL going to require more work than most people are willing to put into it.

[geots]

(my guess for biggest improvement: space eval...missing entirely in Ippolit source code, just see Stockfish for a good one)

Bingo!

Well, to be honest, I don't believe it is the biggest improvement (except for Chess960)... Just that it is in there. And it is computed exactly the same way as in Stockfish btw.

Which version of Houdini added space eval? We tried a version of it quite a while ago in Komodo but could detect no benefit. Perhaps it's worth re-visiting.

In your opinion, is the big jump in Houdini 3 due to anything really original and significant (i.e. double-digit elo gains) or just to Stockfish ideas, parameter tuning, and a few small items here and there?

Larry, you are way too smart to be asking a question like this. "Parameter tuning and a few small items here and there"............... And the same with Robbo ideas= Houdini. If that is all there was to it, do you not think we would have had Houdini's multiplying faster than rabbits. In a discussion back in 2012 right here that 5 or 6 programmers were involved in, Robert gave away his "secret/secrets". And they were all too busy talking to even listen. Which did not even surprise me. Which prompted me a short time later to put up an offer right here- 5000 dollars against 1000 of theirs. To any programmer who could gut RobboLito and any other Ippolits- take some- take all. And in a decided reasonable period of time- catch Houdini 2. Of course that shut the talk up- no takers. Are you shocked?

You told me in 2011 in a thread on the Rybka forum that little or no strength > was gained by Vas from anything he took from Fruit. You said that came from his "unbelievable ability." (I believe "unbelievable" was the word you used- meant the same thing.) You said what he took basically just cut corners and saved him some time. And that bears out- because after he studied Fruit- his first effort after was over 220 elo stronger than Fruit. His next was over 300 elo stronger! If Fruit held those kinds of secrets, why weren't other programmers all over it?! Answer: they were. That was when they got the rude awakening. Vas was on a different planet from them. You know the planet- the same one Robert has joined him on- where the 2 programmers now reside. It is s_i_m_p_l_e: There are no other programmers out there with anything close to the talents that Vas and Robert possess. Not even remotely close.



Best, with all due repect to you,

george

I don't know how much of Robert's gain over Ivanhoe was due to original ideas, but it seems that Richard Vida does know the answer to this question, at least roughly. If much of it came from adding Stockfish ideas to Ivanhoe, well, that is legitimate, but hardly puts Robert in the same category as Vas. Rybka was full of original ideas.

You got me- because I am one of Vas' strongest supporters. No argument there. But you have got to know that Houdini is more than Robbo-Stockfish-Ivanhoe ideas in any combination with a bit of tuning. All programmers have access to all those codes- and if it was a matter of "just a bit of tuning"- we would have so many Houdinis here (with different names of course) we would be tripping over them. I am afraid the treasured secrets in those 3 are exactly like the treasured secrets in Fruit- nonexistant.



Best,

george

[geots]

Some people have hard time grasping the fact Houdini is just an Ippo mod.
It's not a major rewrite and there are no new revolutionary ideas in it.
That's why it takes only weeks to disassemble.

Looking forward to the new Criitter

Are you talking about Houdini 3 or earlier versions? We established some time ago that what you say was true of Houdini 1.5 and perhaps also Houdini 2, but I have not heard anything until now about whether this was also true if Houdini 3. If so, do you believe that the 60 elo gain over Houdini 2 was accomplished just by parameter tuning? It is hard for me to imagine that this large a gain could have been made without at least one significant new idea. But I could be wrong.

Ok Larry- take 1.5 and 2.0. Account for the fact that none of the Ippos- RobboLito, Ivanhoe and Stockfish as well- can really even play on the same board with Houdini 1.5 and 2.0. If Newman thinks it was simple with no new ideas- he is delusional. Trust me, I know of many who have tried every way under the sun to do what Robert did- they all failed miserably. Because what he did is not that simple. Same as Vas and Fruit in this case. You know Vas was not within miles of being the only person that studied Fruit. So why wasn't their first try after the study over 220 elo stronger than Fruit. Even 1 programmer who could just get a third of Vas elo improvement. My offers are still on the table- 5 times what any programmer will put up that he can't catch Rybka thru Fruit, and 5 times what one will put up that he can't catch Houdini thru the Ippolits and Stockfish.

There is always talk about how Vas and Robert had an easy job in front of them. Newman says just an Ippo mod- no original ideas. Or revolutionary. Take your pick. Let's see how lucky he is finding a programmer to put money on the table to back it up. Won't happen, because, in the end- "money talks and bullshit walks."


gts

[GenoM]

I think Richard Vida only make it as a joke : to prove it's possible to disassembly Houdini in a week.


Found here :

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=26422

from Richard Vida Date 2013-01-16 16:22

Barnard wrote :

>he can dissssembly Houdini,that for a good programmer,only will need about a week,or even less (full time)

Labyrinth wrote :

> You mean construct a reasonable source code from a decompile? You're nuts.


wanna bet?

And the name Robodini is from Dr.Wael Deeb (in 2011) :

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=22811

from Dr.Wael Deeb Date 2011-09-01 15:29

You mean ROBODINI

I am extremely glad to contribute to the computer chess community

As said by our prime minister:
Congradjulations, Doc

[Dr.Wael Deeb]

I think Richard Vida only make it as a joke : to prove it's possible to disassembly Houdini in a week.


Found here :

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=26422

from Richard Vida Date 2013-01-16 16:22

Barnard wrote :

>he can dissssembly Houdini,that for a good programmer,only will need about a week,or even less (full time)

Labyrinth wrote :

> You mean construct a reasonable source code from a decompile? You're nuts.


wanna bet?

And the name Robodini is from Dr.Wael Deeb (in 2011) :

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=22811

from Dr.Wael Deeb Date 2011-09-01 15:29

You mean ROBODINI

I am extremely glad to contribute to the computer chess community

As said by our prime minister:
Congradjulations, Doc