![[hr]](/mwf/v12/flags/hr.png "Croatia")
Ok, a month ago we learned that things were going slower than Lucas expected. Any news now?
![[mx]](/mwf/v12/flags/mx.png "Mexico")
![[us]](/mwf/v12/flags/us.png "United States")
I don't get the excitement. From what I've seen on this forum, it is basically assured that Rybka 5 will be the second straight version of Rybka which is essentially cripple-ware. I'm not paying for that again. Fool me once, shame on you.
Can you define 'cripple-ware'?
If it is not considerably stronger than 4.1, I would not bother. It's a bit like having a 3.4 ghz computer and a 3.6 ghz...the gain in ghz is really unimportant after a certain point. The same goes for a difference in say a 3100 "rating" and a 3130 "rating". It's basically pointless except to the early adapters and 'nerds' who con themselves into thinking it's so important to have.
If it is not considerably stronger than 4.1, I would not bother. It's a bit like having a 3.4 ghz computer and a 3.6 ghz...the gain in ghz is really unimportant after a certain point. The same goes for a difference in say a 3100 "rating" and a 3130 "rating". It's basically pointless except to the early adapters and 'nerds' who con themselves into thinking it's so important to have.
Cripple ware is software intentionally weakened before sale. In other words, Vas is no longer selling us the best version of Rybka that he can, he is rather intentionally weakening it to the minimum point that he thinks will sell. He wants to be the number one engine again, but not by too much. He wants to hold out the strongest version he has, either for the rental version, or for future updates.
While I don't believe Vas ever directly confirmed that this was how Rybka 4 was released, I think he practically said as much, and I doubt many on this forum would dispute my claim. I don't remember the exact post, but he posted something here recently that was highly suggestive that the same process will be employed in Rybka 5.
Personally, I find this unacceptable. On principle, I think it is wrong, and I would like to discourage cripple ware wherever possible by refusing to buy such products, whether they are software or anything else (it's an increasingly common phenomenon in many areas of retail). But practically, I find little purpose it paying a large amount of money again for a product that will only be marginally stronger than free competitors, and thus will likely again be overtaken by them within a period of months. I think it is a poor business decision on Vas' part, but I guess I don't really know.
While I don't believe Vas ever directly confirmed that this was how Rybka 4 was released, I think he practically said as much, and I doubt many on this forum would dispute my claim. I don't remember the exact post, but he posted something here recently that was highly suggestive that the same process will be employed in Rybka 5.
Personally, I find this unacceptable. On principle, I think it is wrong, and I would like to discourage cripple ware wherever possible by refusing to buy such products, whether they are software or anything else (it's an increasingly common phenomenon in many areas of retail). But practically, I find little purpose it paying a large amount of money again for a product that will only be marginally stronger than free competitors, and thus will likely again be overtaken by them within a period of months. I think it is a poor business decision on Vas' part, but I guess I don't really know.
![[de]](/mwf/v12/flags/de.png "Germany")
Here's a definition of crippleware: http://en.wikipedia.org/wiki/Damaged_good
I don't think this applies to Rybka.
I don't think this applies to Rybka.
> In other words, Vas is no longer selling us the best version of Rybka that he can, he is rather intentionally weakening it to the minimum point that he thinks will sell. He wants to be the number one engine again, but not by too much. He wants to hold out the strongest version he has, either for the rental version, or for future updates.
That is simply good business, I have no problem with that at all.
> In other words, Vas is no longer selling us the best version of Rybka that he can
Actually, he is, just buy Rybka Cluster time. It may be overpriced but just because most people can't afford it doesn't mean Vas isn't selling the best Rybka that he has.
> In other words, Vas is no longer selling us the best version of Rybka that he can, he is rather intentionally weakening it to the minimum point that he thinks will sell.
IMO this is nonsense. R4.1 was the best he had at the time of its release.
+1
Still, I would want to see how much of an improvement it is before I purchased Rybka 5. I tend to think he's gotten about all he can out of the linage and R5 probably would be fractionally stronger than R4.
Still, I would want to see how much of an improvement it is before I purchased Rybka 5. I tend to think he's gotten about all he can out of the linage and R5 probably would be fractionally stronger than R4.
People said the same with Rybka 2.3.2a. The 100 elo improvement took most people by surprise.
Would you like to make a little wager that R5 is 100 or more points stronger than R4.1? I'd say there is a 98% chance it will not be.
That's 1% because nothing is ever certain and 1% perhaps on the rare occassion, once can catch lightening in a bottle. I'd say it's closer to 99%.
That's 1% because nothing is ever certain and 1% perhaps on the rare occassion, once can catch lightening in a bottle. I'd say it's closer to 99%.
And I think your proposed bet would have looked exactly like that in the times of Rybka 2.3.2a. You'd have lost it.
And we are talking NOW...not then. You have no faith in your position?
Here is my wager: If Rybka 5 comes out and NOT 100 points stronger than Rybka 4.1, same hardware: YOU will not post on this site for a TWO MONTHS.
If is is 100 points stronger:I will not post for a full TWO MONTHS. No posting under an alias. Nada.
Do we have a deal?
Here is my wager: If Rybka 5 comes out and NOT 100 points stronger than Rybka 4.1, same hardware: YOU will not post on this site for a TWO MONTHS.
If is is 100 points stronger:I will not post for a full TWO MONTHS. No posting under an alias. Nada.
Do we have a deal?
I think I have much more to lose than you, as it's not only my loss, but the loss of people playing me in my games in the corr section, that would have to wait 2 months for my move. And other people asking for Rybka support, or other kinds, that I couldn't give.
What about this: if Rybka 5 isn't 100 elo stronger than Rybka 4.1, I abandon this account, and create a new one, so that I lose all my 23477 posts (and my golden crown) and have to begin from scratch.
You wouldn't gain anything from that but I wouldn't gain anything to see you stop posting for two months.
What about this: if Rybka 5 isn't 100 elo stronger than Rybka 4.1, I abandon this account, and create a new one, so that I lose all my 23477 posts (and my golden crown) and have to begin from scratch.
You wouldn't gain anything from that but I wouldn't gain anything to see you stop posting for two months.
Oh that's too much...no need to abandon your account or your Golden Crown or any of the other stuff. I wasn't thinking.
How about something simple and fun: The LOSER of the wager must always address The WINNER of the wager by starting "Uly, whose boots I am not fit to lick..." or "Leavenfish, whose boots I am not fit to lick..." An inside joke if you will.
This would be in any post where the Loser is either Responding to the Winner, Referring to the Winner or Addressing the Winner in any way.
For example, here, if I had lost, I would have started, "Uly, whose boots I am not fit to lick, oh, that is too much...no need, etc."
I post a fraction of what you post, mostly because you are intersted in far more things in the forum than I am though there is some overlap, so it would be fairly rare. The only hard part would likely be remember to do so.
How about something simple and fun: The LOSER of the wager must always address The WINNER of the wager by starting "Uly, whose boots I am not fit to lick..." or "Leavenfish, whose boots I am not fit to lick..." An inside joke if you will.
This would be in any post where the Loser is either Responding to the Winner, Referring to the Winner or Addressing the Winner in any way.
For example, here, if I had lost, I would have started, "Uly, whose boots I am not fit to lick, oh, that is too much...no need, etc."
I post a fraction of what you post, mostly because you are intersted in far more things in the forum than I am though there is some overlap, so it would be fairly rare. The only hard part would likely be remember to do so.
Yeah, let's just agree on the inside phrase that we'd use, something like, I'd address you as "Leavenfish, whose boots I am not fit to lick..." if you win, but if I win, you'd address me as "Uly, my ninja master", or something (I guess anything from here or here followed by "master" would work).
Also, inspired by the Simpsons episode in where Homer and Ned Flanders hold a similar bet, I think we should allow to the possibility of both losing the bet, you said that Rybka 5 would only be "fractionally stronger" than Rybka 4.1, how much is "fractionally"? 20 elo? Because then the bet should be: Rybka 5 stronger than Rybka 4.1 by only 20 elo or less -> You win the bet. Rybka 5 stronger than Rybka 4.1 by 100 elo or more -> I win the bet. Rybka 5 stronger than Rybka 4.1 by more than 20 elo, but less than 100 -> both are losers and have to address the other one like that.
I also think that this should take into account the ELO + - numbers for accuracy, in a way that would minimize the chances of both losing, and that we should agree in what rating list to use, I propose the CCRL:
http://computerchess.org.uk/ccrl/4040/rating_list_all.html
That currently has Rybka 3 better than Rybka 2.3.2a by 104 +29 -29 elo.
Also, inspired by the Simpsons episode in where Homer and Ned Flanders hold a similar bet, I think we should allow to the possibility of both losing the bet, you said that Rybka 5 would only be "fractionally stronger" than Rybka 4.1, how much is "fractionally"? 20 elo? Because then the bet should be: Rybka 5 stronger than Rybka 4.1 by only 20 elo or less -> You win the bet. Rybka 5 stronger than Rybka 4.1 by 100 elo or more -> I win the bet. Rybka 5 stronger than Rybka 4.1 by more than 20 elo, but less than 100 -> both are losers and have to address the other one like that.
I also think that this should take into account the ELO + - numbers for accuracy, in a way that would minimize the chances of both losing, and that we should agree in what rating list to use, I propose the CCRL:
http://computerchess.org.uk/ccrl/4040/rating_list_all.html
That currently has Rybka 3 better than Rybka 2.3.2a by 104 +29 -29 elo.
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
> IMO this is nonsense. R4.1 was the best he had at the time of its release.
But there was a BIG difference in strength between Rybka 4 & the Cluster 6 months later, and not simply because of hardware. (Evals were very different in endgames)
The RC software and R4 series software will have been different for many reasons, including of course, being able to work on a cluster of almost 300 processors. I suspect that many parts of the software engine will be as the R4 series though - it will not have been a complete re-write. Perhaps as much as 10% of the code was changed {new lines added, some modified, some deleted}. The question I would ask is, was the increase in RC strength entirely to do with hardware or did software play its part? If the latter then I wonder what the ratio would have been, 80:20 in favour of hardware, 60:40 perhaps? I suspect were you to offer a position known to be a problem for the R4 series, you may well have got the same {I don't understand this position} answer from RC. Did you try this?
Having an idea of this would mean the speculation around the R5 series would be a little more accurate - I for one hope Vas can deliver a significant improvement - the chess world is certainly looking for something and apart from this, the other big potential chess news this year could be Chess Explorer from the Hiarcs camp.
Having an idea of this would mean the speculation around the R5 series would be a little more accurate - I for one hope Vas can deliver a significant improvement - the chess world is certainly looking for something and apart from this, the other big potential chess news this year could be Chess Explorer from the Hiarcs camp.
> I suspect were you to offer a position known to be a problem for the R4 series, you may well have got the same {I don't understand this position} answer from RC. Did you try this?
Yes, a bit. Rook & pawn endgames seemed to be played much better by the cluster, and endgames in general were evaluated better.
A familiar occurrence at the time was R4.1 going into an endgame showing +1.5 against Houdini whilst the latter calmly showed +0.15.
About 30 moves later the game was duly drawn. But with the Cluster I noticed the reverse - it predicted a win and usually got it. SOME of that will be down to faster hardware (40 cores vs 4 cores) but a lot of it was the software IMO.
![[se]](/mwf/v12/flags/se.png "Sweden")
>"SOME of that will be down to faster hardware (40 cores vs 4 cores) but a lot of it was the software IMO."<
But Vas always claimed at least indirectly that the rybka cluster is the same as the commercial version.
Whatever difference there is, it's certainly of no fundamental change to the commercial one.
> But Vas always claimed at least indirectly that the rybka cluster is the same as the commercial version.
No, I really got the impression that that's not the case. I got the impression a lot of improvements were saved for the cluster. I tested the cluster well after the release of R4 so that's quite a bit of time for him to improve anyway.
Lukas may answer this though. But my impression was: they evaluated differently and the Cluster was a long way ahead of R4
>"I tested the cluster well after the release of R4 so that's quite a bit of time for him to improve anyway."<
Of course it has been more than a year since R4 so surely it's improved, but round about the times of the commercial releases and the cluster the difference is probably slim.
If you look at the cluster games in tournaments it plays the same as current versions of Rybka of those years.
If what you say were true and that there were two code streams {code streams were split at R4 let's say} then the life Vas had would now be even worse - trying to get one code stream update from Vas has been at least slow; how do you reconcile Vas now maintaining two code streams? More likely is that there are some parts of the Rybka code that is re-used and there will be other parts that have needed to be changed to accommodate the Cluster hardware. I'm not saying that this doesn't result in any software evaluation change but question whether the change is that significant in terms of chess evaluation - the hardware must be playing a big part otherwise why have it. the cost of building/maintaining/running such a piece of kit will not be cheap and if the significant part of the improvement were mainly software then this wouldn't make a great deal of sense.
> If what you say were true and that there were two code streams {code streams were split at R4 let's say} then the life Vas had would now be even worse - trying to get one code stream update from Vas has been at least slow; how do you reconcile Vas now maintaining two code streams? More likely is that there are some parts of the Rybka code that is re-used and there will be other parts that have needed to be changed to accommodate the Cluster hardware. I'm not saying that this doesn't result in any software evaluation change but question whether the change is that significant in terms of chess evaluation - the hardware must be playing a big part otherwise why have it. the cost of building/maintaining/running such a piece of kit will not be cheap and if the significant part of the improvement were mainly software then this wouldn't make a great deal of sense.
I really can't answer this, only Vas or Lukas can and Lukas is your best bet I think!
All I can do is say what I saw.But could there be another way of viewing the issue? What if the Cluster version is the "Real Rybka". This contains all of Vas' improvements. He holds nothing back because no one will ever be able to clone this software since it will only be available for rental. Let's say it is 100 elo stronger than Rybka 4.1 on equal hardware (which is my guess BTW).
Any user who wants to rent this - remember the long-term plan of Rental Rybka is to be able to rent even small amounts of hardware - will get a very strong chess entity.
BUT, in order to make a living Vas has to release a UCI version as well. So he cherry picks from the Cluster version and leaves some of his better improvements out, because he knows his code will be copied soon. This results in a potentially good but perhaps buggy UCI program with long delays in between releases.
I think that's what we're seeing in Rybka 4-5.
What you describe is possible but is it very likely? Writing software the way you describe it being used would mean writing it in a very 'modular' way and certainly from the various discussions I had with Vas would say that Vas came across as a very 'informal' programmer who came up with a great 'testing' idea that led to making small adjustments to the code to improve performance. Someone like this is very unlikely able to write software in a very organised 'de-coupled' way that would mean he could 'cherry-pick' the bits he wants in one solution and 'reserve' other bits in a different solution. I am not even sure if we look at probably the most organised chess programmer {in the public domain} on the planet today {SMK} that even he wants to do this. If you remember he did for a while maintain two chess engines {as well as loads of other stuff} and suspect one of the reasons for abandoning this was the drain on his time for very little benefit. In the end his cut-down version of Shredder was just not purchased.
So I'm with Mr Vida on this one; I don't believe it is very likely that Vas can simply keep a significant amount of chess software improvement back for the Cluster. If i'm wrong then do consider why RC was born in the first place; it was primarily driven to protect the IP of the stuff Vas produced so why would he now abandon this approach?
So I'm with Mr Vida on this one; I don't believe it is very likely that Vas can simply keep a significant amount of chess software improvement back for the Cluster. If i'm wrong then do consider why RC was born in the first place; it was primarily driven to protect the IP of the stuff Vas produced so why would he now abandon this approach?
Most of your assumptions are absolutely wrong. For example this one:
Vas started the cluster project because I had a bunch of computers and he was fascinated by the idea of taking advantage of all of them at a time. He even had the idea of adding remote computers for additional strength, but he dropped that idea.
>If i'm wrong then do consider why RC was born in the first place; it was primarily driven to protect the IP of the stuff Vas produced
Vas started the cluster project because I had a bunch of computers and he was fascinated by the idea of taking advantage of all of them at a time. He even had the idea of adding remote computers for additional strength, but he dropped that idea.
Well that is very interesting Lukas.
From this I guess we can deduce then that the new UCI Rybka engine will contain all the strength {from a software perspective} that the RC version has at the point of release - true?
From this I guess we can deduce then that the new UCI Rybka engine will contain all the strength {from a software perspective} that the RC version has at the point of release - true?
Of course not.
even if the answer to your question is 'yes',they must answer 'no',to people that rent the RC continue renting it,thinking the RC has a software advantage over the released rybka engine (even if that is true or not)
so,in general terms,your question has no sense,because even if the ''true'' answer is 'yes' (or 'not',indifferent),the answer you will receive,will be,as you saw,''of course not''
so,in general terms,your question has no sense,because even if the ''true'' answer is 'yes' (or 'not',indifferent),the answer you will receive,will be,as you saw,''of course not''
Very harsh.
sorry,i didnt understood you,what you mean?(im not jocking,im Spanish an my English is very limitated)
It's always a good idea not to talk about things you don't understand.
i dont need to be a programmer to understand you and vas are making money with the cluster,and if the answer to razor were 'yes',you need to answer 'no' to dont losse people continue renting RC and continue making money
also is always a good idea evade answer questions asked you many times like if rybka fakes the plies depth to show -3 depth at it search and why do you think vas implemented it
also is always a good idea evade answer questions asked you many times like if rybka fakes the plies depth to show -3 depth at it search and why do you think vas implemented it
HI Salva,you are very unwise to doubt anything Lukas says,i have known Lukas for a few years now,we used to play team Chess together.
Lukas has never misled anyone on the forum,if he does not give you all the answer's you want try sometimes to read between the lines.
Lukas will never attempt to mislead you he may ignore a question.Respect him in not replying.
IMO it is common knowledge that Rybka depths readings are misleading,there are hundreds of posts on this matter.
The accepted opinion is that you add 3 to whatever Rybka says.However once again this can be misleading it depends upon position.
just my 2 cents.
Lukas has never misled anyone on the forum,if he does not give you all the answer's you want try sometimes to read between the lines.
Lukas will never attempt to mislead you he may ignore a question.Respect him in not replying.
IMO it is common knowledge that Rybka depths readings are misleading,there are hundreds of posts on this matter.
The accepted opinion is that you add 3 to whatever Rybka says.However once again this can be misleading it depends upon position.
just my 2 cents.
Hi Paul
im not unwise to doubt anything that Lukas said,in fact,i dont need that he answers me to know that rybka doesnt count properly the depth,and you must add plies to be a 'real' depth
just Lukas was rude with me with his answer telling me is better ''doesnt speak about i dont understand'',when i just told that razor asked a question,and the answer,in all cases,will be a 'not';i just told that
p.s. usually is a fault of education dont answer a question when is asked that lot of times;i ALWAYS answer a question,even if is only to tell ''im not gonna answer you''
edit:the post at you wrote me,you found any error in my reasoning that made can you think i 'dont understand'?...well,the same reasoning made told Lukas i ''dont understand''
im not unwise to doubt anything that Lukas said,in fact,i dont need that he answers me to know that rybka doesnt count properly the depth,and you must add plies to be a 'real' depth
just Lukas was rude with me with his answer telling me is better ''doesnt speak about i dont understand'',when i just told that razor asked a question,and the answer,in all cases,will be a 'not';i just told that
p.s. usually is a fault of education dont answer a question when is asked that lot of times;i ALWAYS answer a question,even if is only to tell ''im not gonna answer you''
edit:the post at you wrote me,you found any error in my reasoning that made can you think i 'dont understand'?...well,the same reasoning made told Lukas i ''dont understand''
p.s. usually is a fault of education dont answer a question (Barnard).
By saying this you are been very insulting to Lukas.
I know you dont mean it this way but thats how a number of people could read it.
By saying this you are been very insulting to Lukas.
I know you dont mean it this way but thats how a number of people could read it.
i never wouldnt insult Lukas or anyone;when i want insult someone,i directly insult him,i dont use rethorical forms or something similar
i just told that when a person ask a question,is deserved an answer,at least to tell that person ''im not gonna answer you'',so that person knows wont be answered,and wont be waiting an answer...
just leave the other person wating for a response without telling him nothing,and knowing the other person is waiting for your response,is a fault of a education
and you must agree with me that it is a lack of respect to another person leave him waiting for your response,and more even if you know he is waiting for your response
i just told that when a person ask a question,is deserved an answer,at least to tell that person ''im not gonna answer you'',so that person knows wont be answered,and wont be waiting an answer...
just leave the other person wating for a response without telling him nothing,and knowing the other person is waiting for your response,is a fault of a education
and you must agree with me that it is a lack of respect to another person leave him waiting for your response,and more even if you know he is waiting for your response
The "no" is because they are going to reserve some strength to keep the cluster software stronger, i.e. probably making the commercial release an earlier version of the strongest thing they have available. (I could be completely wrong on this--it is possible that development of Rybka Cluster software and development of the commercial Rybka version are on completely separate paths, in which case, Lukas' "of course not" would have nothing at all hidden within.)
This doesn't contradict the fact that the original purpose of making Rybka Cluster was as a software/hardware project due to Vas' interest in testing out Lukas' hardware. It's just that now it can be used commercially and profits can be made.
This doesn't contradict the fact that the original purpose of making Rybka Cluster was as a software/hardware project due to Vas' interest in testing out Lukas' hardware. It's just that now it can be used commercially and profits can be made.
interesting reasoning
from the point i know,other software developers also 'reserve' some strength/features private(here for RC,others to help known players)
from the point i know,other software developers also 'reserve' some strength/features private(here for RC,others to help known players)
I don't see why he would not use the full strength of the software at the time of release. The simple reason is that the cluster is stronger simpley because its the cluster with more cores to work with. There is no need to nerf down the software just for the end users at home to use. They cannot compete with the cluster analysis regardless. Doesn't really make sense to me. But i guess it doesnt really matter as long as the new version is much better than the old version right!? After all were buying the new version since we want to have a strong better tool to use.
Please look at the reply from Lukas; he was very clear in his response to me.
When R5 {or whatever the version is} is released, the most important thing is that it plays stronger chess; there would be very little point in buying the new release if it does NOT provide a stronger chess problem solving capability over R4.1 - at least that is my view.
When R5 {or whatever the version is} is released, the most important thing is that it plays stronger chess; there would be very little point in buying the new release if it does NOT provide a stronger chess problem solving capability over R4.1 - at least that is my view.
>There is no need to nerf down the software
yes,there is need to do that,to atract people with a lot of money to rent rybka cluster promissing them rybka cluster software has better evaluations (or analisis options,is just an example) than the software released to the general public,so people with that lot of money will continue renting the cluster,and they will continue making money
This seems speculative... Imagine this from the viewpoint of the distributors ChessBase and Convekta. The competition is so tough one can't seriously offer them something second best.
What would make sense to me is though, that the cluster version has certain adaptations to optimize it for that specific hardware environment (and which do not provide improvement for "normal" computers). - But that is speculative either.
What would make sense to me is though, that the cluster version has certain adaptations to optimize it for that specific hardware environment (and which do not provide improvement for "normal" computers). - But that is speculative either.
this isnt speculative...from words of Lukas and Vas,the software that clusters runs has different and unique features than the rybka engine released to the general public
and the second point,they dont rent for free,they rent for money,so my post isnt speculative,all that i said is true
from the rybka page about the cluster: ''THE RYBKA CLUSTER FEATURES EXCLUSIVE RYBKA VERSIONS NOT AVAILABLE TO THE PUBLIC''
so you are wrong,i was not especulative,i was telling the truth
and the second point,they dont rent for free,they rent for money,so my post isnt speculative,all that i said is true
from the rybka page about the cluster: ''THE RYBKA CLUSTER FEATURES EXCLUSIVE RYBKA VERSIONS NOT AVAILABLE TO THE PUBLIC''
so you are wrong,i was not especulative,i was telling the truth
this is true but........ basically with the strong competition. He will still need to exceed other engines other wise he wont make any money on the software period. It may in fact have different versions on the cluster but, as to weather it really is that much better over the standard release is questionable. People will rent the cluster simpley because more cores is more power. and more power is better quality overall anyway. Which is why you rent the cluster in the first place otherwise there would be no point in renting it.
they can,ad yes,that is just speculating,release software that is just enought to be 10-20 elo points stronger than the competitor,and continue reserving his better improvements for the cluster so people will continue renting and they will continue making money
and that is another speculation,maybe nothing about that is right,and they havent a 'special' rybka for the cluster (just a better implementation for multiprocessing,but WITHOUT any real improvement about chess knowledge or difference between rybka released to general public),and they tell they have a 'special' rybka at the cluster,but will release his better rybka to the general public with,an example,rybka 5,so they can reach top 1 (or near top 1 again),and they will continue telling people that the rybka running on cluster has better improvements than the released to the public,even if that isnt true,so people will continue renting the cluster,and they,releasing their better rybka,reach the top 1 again
of course,all of that is speculation
and that is another speculation,maybe nothing about that is right,and they havent a 'special' rybka for the cluster (just a better implementation for multiprocessing,but WITHOUT any real improvement about chess knowledge or difference between rybka released to general public),and they tell they have a 'special' rybka at the cluster,but will release his better rybka to the general public with,an example,rybka 5,so they can reach top 1 (or near top 1 again),and they will continue telling people that the rybka running on cluster has better improvements than the released to the public,even if that isnt true,so people will continue renting the cluster,and they,releasing their better rybka,reach the top 1 again
of course,all of that is speculation
> just a better implementation for multiprocessing
That makes sense. In addition to it, I can imagine that the evaluation function(s) could indeed be a bit more and/or bigger if that turned out to improve the overall strength on the cluster, while being too slow on typical comps.
Nevertheless, have we seen any facts or signs whatsoever, that the cluster rental program is a success or if it has a reasonable number of customers, at all? I wouldn't be sure about that. There is so much speculation and imagination involved.
I think the "normal" release engine is much, much more interesting and also I do not think that any compromise in terms of achievable strength is affordable. The competition is too strong.
The cluster has it's huge hardware advantage anyway.
you are right...
but one problem of the 'released' rybka (or other engines),is that in just a few months,is RE,and all its secrets are showed and implemented in other programs
and that is why actually there are that lot of top programs,and when one one programmer discover one 'secret',and make the engine,lets say,100 points elo stronger,be for sure that the engine will be RE and the 'secret' will be implemented in the other engines (or at least in the engine of the guy who RE)
but one problem of the 'released' rybka (or other engines),is that in just a few months,is RE,and all its secrets are showed and implemented in other programs
and that is why actually there are that lot of top programs,and when one one programmer discover one 'secret',and make the engine,lets say,100 points elo stronger,be for sure that the engine will be RE and the 'secret' will be implemented in the other engines (or at least in the engine of the guy who RE)
If all one is looking for is an ELO result based on a number of games played then assuming you are judging No 1 in this way I guess you mean the IPON list. Currently the IPON list has R4.1 some 64 ELO points below H2 and with the H2 author claiming the new Houdini will be launched around Sep 2012 with a gain of around another 40ish ELO points then it seems Rybka will need to be at least 110 ELO over R4.1 to stand a chance of taking the IPON No 1 spot.
As well as the above Rybka will have to compete with the new Critter and Komodo engines so I don't believe it is going to be that easy, unless Rybka is released now perhaps. Releasing now might mean that Rybka will catch the competition cold for a few weeks at least and hold the IPON No 1 spot for a while at least; assuming Rybka has around 70ish ELO more than R4.1. If not I guess we will all have to wait a while; perhaps too long!
BTW, don't expect great things from chess engines for a while, I think chess engine programmers need to find another significant breakthrough before we notice anything major; for now it just feels like small adjustments.
As well as the above Rybka will have to compete with the new Critter and Komodo engines so I don't believe it is going to be that easy, unless Rybka is released now perhaps. Releasing now might mean that Rybka will catch the competition cold for a few weeks at least and hold the IPON No 1 spot for a while at least; assuming Rybka has around 70ish ELO more than R4.1. If not I guess we will all have to wait a while; perhaps too long!
BTW, don't expect great things from chess engines for a while, I think chess engine programmers need to find another significant breakthrough before we notice anything major; for now it just feels like small adjustments.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill