Any resemblance to Vasik Rajlich (The Swordman) is purely coincidental!Milos wrote:For all of you Chuck Norris, Steven Seagal, ninja, jedi, mambo-jambo, marital arts guys, one nice vid
http://www.youtube.com/watch?v=bwGg_F7s7xg
How does Mr. Rajlich know the Ippolit series is Rybka 3?
Moderators: hgm, Rebel, chrisw
-
- Posts: 431
- Joined: Tue Dec 01, 2009 11:59 am
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
Gods are fragile things; they may be killed by a whiff of science or a dose of common sense.
-
- Posts: 227
- Joined: Tue Oct 05, 2010 5:29 pm
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
If I'd bothered using a search engine, I would have done better than that. You're not the only person who knows anything about martial arts. I could probably tell you things about it that you wouldn't know and couldn't find on the internet. I worked around these people for a good while: something else you'd realize if you knew who I was.geots wrote:]
Dont pee on my head and then try to tell me its rainwater. You googled and almost copied word for word what was on the different pages you brought up. You got to go one better than that. I was trying to make peace with you, and the mods can see that. But you have persisted in this course, so tell me where you are and when you want to meet halfway.
So much for the olive branch. I suppose that a leopard can't change his spots, and I doubt you're capable of being civil to anybody for longer than a day, if that.
This production is being brought to you by Rybka: "The engine made from scratch.™"
-
- Posts: 5106
- Joined: Tue Apr 29, 2008 4:27 pm
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
Robert,Houdini wrote:Larry,lkaufman wrote:But the eval is not much changed from Robbo or from R3 in terms of the output, even if the terms were changed from tables to formulae.
Let me repeat here what I've just posted in the "which program is best in endgames" thread, in reply to M. Ansadi making a similar claim for end games.
Below 10 end game examples from the recent 40 games TCEC Elite Match. I've selected positionally imbalanced, but not too tactical end game positions in which the evals remain more or less constant for several moves.
Stop treating us like we are idiots.
There is nobody who doubts that you made significant improvements to Robbolito and that you are working hard on the evaluation, especially the endings. We also know you scaled the evaluation differently. But showing us a few positions where the scores are a lot of different might convince naive people but it's not going to fool people with half a brain.
I can do this same type of analysis on Komodo vs an identical version of Komodo and it will return different scores and sometimes significantly different scores. In a non-trivial percent of the time it will return different moves. And I can pick out the ones I want to show you.
What is very clear is that your program plays VERY much like all the other Ippolito derivatives. Over time this will change, as you continue to work on the evaluation and diverge from the original code base.
You are doing the same exact thing the Ipolitto guys did. Their "proof" that the program is not a derivative of Rybka is some examples where it plays MUCH better and the fact that it plays stronger. If that is proof, then this same proof could be applied to Rybka 3, right?
If you really want to make your case, publish your source code. You are not going commercial so you have no reason not to, and didn't you say at one point that you intended to?
The 10 positions demonstrate rather convincingly how much Rybka's and Houdini's end game evals are different.
Can you tell us which of the two evals you consider to be based on your evaluation terms?
Robert
-
- Posts: 1471
- Joined: Tue Mar 16, 2010 12:00 am
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
The point is exactly that I'm not just "showing a few positions" that I would have cherry-picked from a very large number of obscure games.Don wrote:But showing us a few positions where the scores are a lot of different might convince naive people but it's not going to fool people with half a brain.
I'm showing 10 of the 40 games played this week in the very public TCEC match. In fact, I'm showing nearly every game in which a not obviously equal end game has occurred.
Now take a look at the positions, compare the evals, you could even play the full 40 games from the PGN file that is available at the TCEC web site. After that, consider Larry Kaufman's claim that "he recognizes the Rybka 3 eval in Houdini", and LOL.
Please note also that the burden of proof lies with Larry who's repeatedly made the same strong claim without providing any evidence whatsoever.
Robert
-
- Posts: 5106
- Joined: Tue Apr 29, 2008 4:27 pm
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
I provided strong circumstantial evidence when I created the similarity tester. The only program that matched Robbolito more than your own program is Firebird.Houdini wrote:The point is exactly that I'm not just "showing a few positions" that I would have cherry-picked from a very large number of obscure games.Don wrote:But showing us a few positions where the scores are a lot of different might convince naive people but it's not going to fool people with half a brain.
I'm showing 10 of the 40 games played this week in the very public TCEC match. In fact, I'm showing nearly every game in which a not obviously equal end game has occurred.
Now take a look at the positions, compare the evals, you could even play the full 40 games from the PGN file that is available at the TCEC web site. After that, consider Larry Kaufman's claim that "he recognizes the Rybka 3 eval in Houdini", and LOL.
Please note also that the burden of proof lies with Larry who's repeatedly made the same strong claim without providing any evidence whatsoever.
Larry has a strong interest in the playing styles of various programs and he KNOWS how they all play. He constantly pesters me about working with him on the playing style of Komodo, to make it play more moves like humans play.
He also knows that Houdini plays like all the other clones. Houdini is STRONGER than the other derivatives, but that's not the point at all.
I also know that building a strong evaluation function is an extremely difficult art and takes years and you never get it right. You did not just come out of nowhere with a program like Houdini that plays like Robbo without starting with Robbolito's evaluation function. You can try to sell this to someone else, but not me.
Robert
-
- Posts: 1471
- Joined: Tue Mar 16, 2010 12:00 am
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
Don, why don't you let Larry answer the question? After all, he is the one making the claim I'm challenging.
Your point of view on the subject is well known, there's no point in repeating it endlessly.
Robert
Your point of view on the subject is well known, there's no point in repeating it endlessly.
Robert
-
- Posts: 4190
- Joined: Wed Nov 25, 2009 1:47 am
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
Would you please stop treating us as idiots.Don wrote:I provided strong circumstantial evidence when I created the similarity tester. The only program that matched Robbolito more than your own program is Firebird.
Larry has a strong interest in the playing styles of various programs and he KNOWS how they all play. He constantly pesters me about working with him on the playing style of Komodo, to make it play more moves like humans play.
Your similarity tester is a joke. Larry is a guy that doesn't even understand alpha-beta. Talking about engine playing style is a joke.
Actually most of your posts are a joke.
You are simply jealous since your Komodo is crap compared to Ippolit and even more compared to Houdini. Occam's razor tells us that undoubtedly.
-
- Posts: 5106
- Joined: Tue Apr 29, 2008 4:27 pm
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
I'm not stopping him from answering. He can answer whenever he wants or he can choose not to.Houdini wrote:Don, why don't you let Larry answer the question? After all, he is the one making the claim I'm challenging.
I have mostly refrained from talking about Houdini as you have kept a low profile, but you seem to be getting bolder now so I am less inhibited too.
Your point of view on the subject is well known, there's no point in repeating it endlessly.
Robert
-
- Posts: 16465
- Joined: Wed Aug 01, 2007 4:16 am
- Location: Canada
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
More ad hominen attacks on both Don and Larry! Wow, you're really something aren't you....venom and libel is your trade and I think it's insane you're allowed to keep this up shitmeister.Milos wrote:Would you please stop treating us as idiots.Don wrote:I provided strong circumstantial evidence when I created the similarity tester. The only program that matched Robbolito more than your own program is Firebird.
Larry has a strong interest in the playing styles of various programs and he KNOWS how they all play. He constantly pesters me about working with him on the playing style of Komodo, to make it play more moves like humans play.
Your similarity tester is a joke. Larry is a guy that doesn't even understand alpha-beta. Talking about engine playing style is a joke.
Actually most of your posts are a joke.
You are simply jealous since your Komodo is crap compared to Ippolit and even more compared to Houdini. Occam's razor tells us that undoubtedly.
Terry McCracken
-
- Posts: 2331
- Joined: Mon Apr 09, 2007 5:36 pm
Re: How does Mr. Rajlich know the Ippolit series is Rybka 3?
ad hominem, McCracken, not ad hominen.Terry McCracken wrote:More ad hominen attacks on both Don and Larry! Wow, you're really something aren't you....venom and libel is your trade and I think it's insane you're allowed to keep this up shitmeister.Milos wrote:Would you please stop treating us as idiots.Don wrote:I provided strong circumstantial evidence when I created the similarity tester. The only program that matched Robbolito more than your own program is Firebird.
Larry has a strong interest in the playing styles of various programs and he KNOWS how they all play. He constantly pesters me about working with him on the playing style of Komodo, to make it play more moves like humans play.
Your similarity tester is a joke. Larry is a guy that doesn't even understand alpha-beta. Talking about engine playing style is a joke.
Actually most of your posts are a joke.
You are simply jealous since your Komodo is crap compared to Ippolit and even more compared to Houdini. Occam's razor tells us that undoubtedly.
Besides, you should know that telling a truth is not the same as an ad hominem attack. You should first ask him if he could prove his words before your start your ad hominem attacks.
Perhaps you are still enough young to learn new things.
"Well, I´m just a soul whose intentions are good,
Oh Lord, please don´t let me be misunderstood."
Oh Lord, please don´t let me be misunderstood."