kranium wrote:
i'm suggesting you are not completely unbiased, as you have made your position clear...
Norm, I don't think Chris is biased, he has made several well reasoned posts although I have only some 3/4 of his posts concerning this issue.
If anything at all, I would argue that you are little more biased than what you wrote of Chris since you have posted some things that are obvious exaggerations to the extent such as the one where you claimed that 2 engines are nearly 40-50% identical.
Ofcourse, I'm not being critical, just writing what I honestly think in order to try to refute your point that looks bit provocative.
Hi Swami-
yes i see your point. i wasn't trying to exaggerate...i posted immediately afterward that it probably wasn't accurate, and only an esitmate.
as far as possible bias by Chris, i simply meant that he has made his position clear...(and is therefore no longer impartial.)
My position is that an accused person is innocent until proven guilty.
What's yours?
Mine is the same, but the evidence has become substantial. We have the gun that killed someone. We have fingerprints on the gun. We have gunshot residue on the suspect. We have established motive. We have established opportunity. The suspect was seen entering and leaving the building during the time the victim was killed. Gunshots were heard from inside the building while the suspect was there. The suspect had victim's blood on his clothes. All we lack is an eye-witness. But the case _still_ looks pretty bad and people have been convicted on far less.
Yes, Bob, you have a kalashnikov full of metaphors,
chrisw wrote:Vas has this to say, together with a proposal.
Rybka is and was always completely original. He is aware there are rumours about Fruit for a long time but never really understood where they are coming from. He is more than happy to try and address concrete questions or curious similarities with Fruit.
So, may I suggest you use this thread to put together a concrete and formal concern list to send to Vas. He can answer here, or here by email or on Rybkaforum or wherever.
Basically, prepare a list of concerns/question which have some suitable evidential backing, and append the evidence/data. We'll email it to him.
Christophe/Zach - do you want to put together a (small) team to do this?
I am game, but this entire process will happen in the open. I'll post the list here, and Vas can answer here or on the Rybka forum. Though I would also call Norman biased of course, I think it's fair to say that you've been decidedly pro-Rybka stance throughout the discussion, so I think it's reasonable to ask to deal with Vas directly and in the public.
Well, now this thread is already getting long, so I'll start off a new thread with just questions/similarities etc.
Zach,
well put. I am sure you will be fair. I agree with you that Norman might not be unbiased. I am not sure about other experts either, who have been anti-Vas or Rybka for a long time (long before this issue), whether they will be biased or not.
On the other hand, as a human it's difficult to be totally unbiased about some issues. For examle, I remember you calling Vas's statement (that Rybka 3 MP efficiency might be better than other programs) about MP efficiency BS, even though you at that time did not measure and compare the MP efficiency with Zappa, .
There are some things we _do_ know without having to "compare". We _know_ what the overhead issues are because of imperfect move ordering, and no one can overcome those no matter how talented. I am still waiting for someone to publish some one-thread, two-thread, four-thread and eight-thread time-to-depth numbers for Rybka. We've had hundreds of those published for Crafty, and others.
Then we don't need "claims" just a "calculator".
I can even provide a test set that several of us can run with 1, 2, 4 and 8 threads, if you'd like and then we can test that "better scaling is better than others" statement factually rather than as marketing-speak.
The only way to test is by games at unequal time control.
Fixed depth for rybka means nothing because I remember reading that rybka mp can beat rybka single processor in games at the same fixed depth.
I can only guess that if it is correct then it is because
The mp version does less pruning at the same fixed depth.
chrisw wrote:Vas has this to say, together with a proposal.
Rybka is and was always completely original. He is aware there are rumours about Fruit for a long time but never really understood where they are coming from. He is more than happy to try and address concrete questions or curious similarities with Fruit.
So, may I suggest you use this thread to put together a concrete and formal concern list to send to Vas. He can answer here, or here by email or on Rybkaforum or wherever.
Basically, prepare a list of concerns/question which have some suitable evidential backing, and append the evidence/data. We'll email it to him.
Christophe/Zach - do you want to put together a (small) team to do this?
Thanks for your efforts regarding Rybka. I think Rybka is clean.
Vas does not really have to say anything or show anyone his code.
Best to you,
Gerold.
Hi Gerold,
i agree, he should not and does not have to show anyone his code.
he seems interested in answering the questions, though...
people are simply requesting an explanation as to why hundreds of (identical/equivilent) lines of code appear in both fruit 2.1 and strelka 2.0.,
etc.
i think that's a fair question...
Unfortunately, based on a post of his (link given above) you are not likely to get substantive answers. Just "it is likely that if we use the same algorithm, we will produce the same code." Even though decades of academic experience spread over every university offering a computer science or programming degree says otherwise.
I see. So there's no point in putting concrete concerns with evidence to him because, according to you, you won't get a substantive answer. Your decades trll you so.
Are you actually able to generate a list of concrete concerns? For example some code similarities between Rybka and Fruit?
If not, the case against Rybka is just so much hot air.
I have two so far. common code between strelka and Rybka 1 beta, already published. Which follows common code between strelka and Fruit, also already published. So that is part 1.
Second part is the common code between Rybka 1 and fruit, which was _also_ published here. Three programs with lots of code in common, defying any possible laws or probability on independent generation of identical code.
Yes, one might duplicate single lines. How many ways are there to do a read or write? But if you assign 30 students the task of producing a random number generator, with no outside access to the internet so that they do everything on their own, you are not going to get duplicate algorithms, much less duplicate code.
There is nothing wrong with erring on the side of conservatism. But you can't take the Uri approach and require the existance of some circumstance that has a probability of 10e-75 to support an argument. A monkey _could_ produce the same code as is in Crafty. But what is the probability and is that a realistic assumption for the argument? I doubt anyone would say "yes".
The third program Strelka doesn't help you for reasons already set out.
If you think there is common code between Rybka1 and Fruit then lets see it. Vas says he will address it. Let's see your code examples and then we'll see his comments.
bob wrote:Your quote is not exactly correct. Should be:
"Innocent until proven guilty beyond a _reasonable_ doubt". One can _never_ prove something with 100% certainty as there is an escape at many different points, although some might depend on interdimensional or time travel or something else that might well be possible, but has not been done to date.
So "beyond a reasonable doubt" is different.
Just as long as the proof doesn't rest on sinking or floating that shouldn't be a problem
Uri Blass wrote:I can only guess that if it is correct then it is because
The mp version does less pruning at the same fixed depth.
I suspect this might be true as well. Which means that the scaling of Rybka is impossible to measure. Which means that everyone will believe that it is superior.
Uri Blass wrote:I can only guess that if it is correct then it is because
The mp version does less pruning at the same fixed depth.
I suspect this might be true as well. Which means that the scaling of Rybka is impossible to measure. Which means that everyone will believe that it is superior.
I think that it is possible to get an estimate by games with unequal time control and no pondering.
Simply play Rybka single core against Rybka mp(4 cores) when Rybka single core get 7 minutes against 2 minutes of rybka 4 cores.
If you get something close to 50% after some thousands of games then 3.5 is a good estimate for 4 cores.
