1
2
Poll
Which engine will be released sooner? (Closed)
| 1.hiarcs 14 | 25 | 46% | |
| 2.shredder 13 | 4 | 7% | |
| 3.rybka 5 | 5 | 9% | |
| 4.komodo 4 MP | 16 | 30% | |
| 5.critter 2 | 4 | 7% |
A question about magnus.
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
What I'd really like from the Komodo team is a realistic release date. I've been waiting for months for their product which was expected 'in a few weeks'.
Just some sort of progress update would be nice.
Just some sort of progress update would be nice.
Also, why hold back the SP 4 series update {I believe Larry said there were a few improvements made to the evaluation and search} for all those who bought K4? If SP and MP are not linked then why hold back an SP release?
> What I'd really like from the Komodo team is a realistic release date. I've been waiting for months for their product which was expected 'in a few weeks'.
Don's waiting for some more new ideas to come his way. No one has over taken Houdini and Rybka 5 hasn't come out yet -he would still have to wait on Richard V! So! Realistic Date? Hum! Soooooooooooooooooon!
> No one has over taken Houdini and Rybka 5 hasn't come out yet -he would still have to wait on Richard V!
Well, don't count on me. My interests shifted somewhat since last Critter. I am now much less focused on raw Elo in engine-engine matches, instead I am trying to make my engine more useful for analysis (while trying to stay on par with 1.4a).
>> No one has over taken Houdini and Rybka 5 hasn't come out yet -he would still have to wait on Richard V!
> Well, don't count on me. My interests shifted somewhat since last Critter. I am now much less focused on raw Elo in engine-engine matches, instead I am trying to make my engine more useful for analysis (while trying to stay on par with 1.4a).
That is awesome news!!
I think what Don is attempting is to champion both a chess engine on steroids and a highly developed chess analysis tool. I am not sure how many programmers are able to work that out within a limited scope of time. I have no doubt that you are a gifted programmer. I would not doubt for a moment that you will develop an excellent chess analysis tool. May be if the spirit moves you at some later date- you might persuade yourself develop the engine in elo strength as well.
![[mx]](/mwf/v12/flags/mx.png "Mexico")
Great decision Richard, for analyzing I still find Rybka 3 better than Rybka 4 despite the elo difference, because of its Persistent Hash, so I'm glad to see Critter going this route (opposed to, say, Stockfish where users have to fix the problems themselves.)
Have you considered going for a Critter GUI? Because at some point Critter is going to hit a block if the GUIs can't support future analysis features.
Have you considered going for a Critter GUI? Because at some point Critter is going to hit a block if the GUIs can't support future analysis features.
> Well, don't count on me. My interests shifted somewhat since last Critter. I am now much less focused on raw Elo in engine-engine matches, instead I am trying to make my engine more useful for analysis (while trying to stay on par with 1.4a).
Great decision. After all, that is what people use their engine for
> Well, don't count on me. My interests shifted somewhat since last Critter. I am now much less focused on raw Elo in engine-engine matches, instead I am trying to make my engine more useful for analysis (while trying to stay on par with 1.4a).
One thing I really wish could be improved with Critter is that although it is very good at finding winning moves fast, as soon as it finds a good move it doesn't try very hard to find a better one. The current approach may be good for getting a high elo, but if you're after the truth of a position it is frustrating.
I've seen that, it seems at some point there's too much pruning and Critter goes too deep in the main lines while leaving side lines undercovered.
For instance, an event like this:
Critter, on a position, as white, thinks move A has score 0.30 at depth 19.
User plays move B instead on the position.
Critter now, sees that this is 0.42 as soon as depth 14.
It means, that on the original position, Critter pruned move B at depth 13 or earlier. I guess some kind of option to make Critter prune less (with the price of going less deep) could be useful for analysis. In this case, the time used to go from depth 18 to depth 19 could be used to extend some moves that were cut earlier and move B may be discovered in the same time even though Critter only reaches depth 17. Losing these two plies may be fatal for games, but for analysis finding the better move is more important than having a more accurate score of main move (if Critter doesn't change its move choice next depth, and the score of the move is very close to what it was last iteration, this time was basically wasted and it may have been better to try to find a fail high in the alternative moves.)
(This is inspired by Zappa Mexico II that seems to be excellent at covering sidelines, unfortunately, its evaluations have some major flaws that engines like Rybka or Critter punish heavily.)
For instance, an event like this:
Critter, on a position, as white, thinks move A has score 0.30 at depth 19.
User plays move B instead on the position.
Critter now, sees that this is 0.42 as soon as depth 14.
It means, that on the original position, Critter pruned move B at depth 13 or earlier. I guess some kind of option to make Critter prune less (with the price of going less deep) could be useful for analysis. In this case, the time used to go from depth 18 to depth 19 could be used to extend some moves that were cut earlier and move B may be discovered in the same time even though Critter only reaches depth 17. Losing these two plies may be fatal for games, but for analysis finding the better move is more important than having a more accurate score of main move (if Critter doesn't change its move choice next depth, and the score of the move is very close to what it was last iteration, this time was basically wasted and it may have been better to try to find a fail high in the alternative moves.)
(This is inspired by Zappa Mexico II that seems to be excellent at covering sidelines, unfortunately, its evaluations have some major flaws that engines like Rybka or Critter punish heavily.)
![[in]](/mwf/v12/flags/in.png "India")
>instead I am trying to make my engine more useful for analysis (while trying to stay on par with 1.4a).
+1
One of the best focuses in CC now. We all are with you ,who are in CC only to play correspondence chess.
May God give you time, money and the power for this hobby of yours.
Regards
I heard the SP version is buggy. (Don Dailey and Larry Kaufman are old) Maybe they are fixing it.
I'm not sure what the reference to 'Old' has to do with this but I believe you meant to say 'MP' rather than 'SP'; I understood the problem for the delay to be down to issues with the MP implementation.
> I'm not sure what the reference to 'Old' has to do with this
It makes them seem disabled or something. I'm sure age is not a problem.> I understood the problem for the delay to be down to issues with the MP implementation.
Larry said Don was trying to locate a MP bug, but that was a few weeks ago. I was hoping there would be some update.
He probably means that because they are old, they are cranky and don't like to be questioned about release schedules!
And not greedy like younger people.
BTW i still cant find any increase in ELO between H 1.5a and H2 and maybe H3 will be imaginary also,A lot of people will pay money to find out.
I am not talking about blitz games which are of no importance to me.
BTW i still cant find any increase in ELO between H 1.5a and H2 and maybe H3 will be imaginary also,A lot of people will pay money to find out.
I am not talking about blitz games which are of no importance to me.
![[us]](/mwf/v12/flags/us.png "United States")
Correct. I think talkchess recently had that overall rankings and what was the elo difference? Maybe about 10 between 1.5 to 2?
![[nl]](/mwf/v12/flags/nl.png "Netherlands")
you must be the idiot in our family.
>Which engine will be released sooner?
The least important!
Where is the option none of the above.!!
![[hr]](/mwf/v12/flags/hr.png "Croatia")
>Where is the option none of the above!!
Also, where is the option - Strelka 5.5 ?
With regard to what I've read in various forums it seems that version 5.5 is in final phase of testing...
http://imageshack.us/photo/my-images/638/strelka55.jpg/
I'm more interested in the development of Robbolito and it's next version update.
>I'm more interested in the development of Robbolito and it's next version update.
Robbolito/IvanHoe project/development looks pretty neglected (maybe even dead) for a long time (with its "original creators"), on the other hand, if Osipov really can to improve ELO rating of Strelka 5.1 for approximately ~40-50 points (look the screenshot; it is +41 currently...), then the Strelka 5.5 could become number one on all rating lists and it would be "much more interesting novelty" (in CC) especially in relation to a some half-dead (native Robbolito) project...
I'm not sure exactly when RobboLito 0.10 SMP made its appearance, but I don't think it was all that long ago...I do know it was after Fire was discontinued.
Robbolito 0.10 is not the product of the original creators (from the original team), but in this case one another programmer/compiler was made a few small improvements and was made support for working with multiple CPUs/cores, etc... But this engine is slightly weaker than the latest Ivanhoe compiles.
By the way, regarding my previous post (regarding new version of Strelka), should see this rating list (for example):
http://computerchess.org.uk/ccrl/4040.live/cgi/compare_engines.cgi?class=32-bit+Single-CPU+engines&print=Rating+list&print=Results+table&print=LOS+table&table_size=12&cross_tables_for_best_versions_only=1
It is not difficult to imagine what might happen (on that rating list) if Strelka 5.5 to be stronger for ~40 Elo points compared to version 5.1...
By the way, regarding my previous post (regarding new version of Strelka), should see this rating list (for example):
http://computerchess.org.uk/ccrl/4040.live/cgi/compare_engines.cgi?class=32-bit+Single-CPU+engines&print=Rating+list&print=Results+table&print=LOS+table&table_size=12&cross_tables_for_best_versions_only=1
It is not difficult to imagine what might happen (on that rating list) if Strelka 5.5 to be stronger for ~40 Elo points compared to version 5.1...
> Robbolito 0.10 is not the product of the original creators
Yes, and it causes confusion. Old named engines should be kept for the original versions of the authors. Even Stockfish changed its name from Glaurung to avoid this! It's just common sense.
It is not difficult to imagine what might happen (on that rating list) if Strelka 5.5 to be stronger for ~40 Elo points compared to version 5.1...
This is a big if, if most of S5.5's strength is due to use of Houdini... My apologies to the author if this is not the case of course.
This is a big if, if most of S5.5's strength is due to use of Houdini... My apologies to the author if this is not the case of course.
>This is a big if
Not very big
if you finally throw you look at this screenshot :http://imageshack.us/photo/my-images/638/strelka55.jpg/
Strelka 5.5 vs Strelka 5.1 -> +2647 =3678 -1675
Moreover, the difference (in Elo) between Houdini 2.0c and Strelka 5.1 is very small, therefore, even if that improvement of Strelka 5.5 will be very small, it might be enough to Strelka that this engine took the first place on the rating list.
When Osipov turns out a 64 bit engine - I'll start to take note.
>When Osipov turns out a 64 bit engine
Actually, I think (AFAIK) that any very good source-code compiler can (without any problems) to compile equally strong chess engine (from 32-bit source code) also for 64-bit platforms...
Something more serious problem could be the MP implementation and its overall effectiveness, etc., but we shall see what will happen about it...
Not interest what or who put it out! Edit-no one really knows who put the damned thing out anyhow.
>Not interest what or who put it out!
In this case it is a very important difference, because various compilers can not more significantly to improve the performance/strength (or speed) of that engine (from the source-code), in a situation when a team of ("native") programmers no longer have the technical capabilities (or ideas, or interest...) or predispositions (eg.; work with some of the strong GMs, etc.) that they would significantly to improve that engine.
You are talking much to much in general terms. Relax your concerns. I think Norm and company have something of a handle on what they are doing. (Although, they never seem to be able to pass go! ).
).
>You are talking much to much in general terms
With the obvious reason
, because I can only write about what I was previously read on a various Russian forums, and I was read about this:By kosmodrom: (08-04-2012 17:03) [ Source: gladiators-chess.ru forum ]
Звонил только что Юрию Осипову. Ситуация такая:
1. Новая Strelka5.5 выйдет в конце апреля.
2. Strelka5.5 будет на одном ядре на 30 пунктов Эло сильнее одноядерного Houdini.
3... etc...
+ about this:
http://imageshack.us/photo/my-images/638/strelka55.jpg/
>I think Norm and company have something of a handle on what they are doing
Maybe they have, maybe not (the potential No.1 engine), but (also) keep in mind that the Strelka is free engine while the Komodo is commercial (as well as Houdini
) and the big question is (about the latest and "long awaited" Komodo MP engine) when indeed he will be finished (before or after the Strelka 5.5 will be available) and will he indeed to be stronger than Houdini 2 and Strelka 5.5/5.1 ?
![[jo]](/mwf/v12/flags/jo.png "Jordan")
Donno about Houdini 2 but Komodo will definitely not be stronger than Strelka 5.5 in the statement of it's current status is valid
Dr.D
Dr.D
What forum are you referencing? So far all I have seen come down the pike -even from the previous touting of Vitruvius was gas when it came down to playing toe to toe against Houdini. It is just a matter of fact- no one has to be a fanboy to make these observations and pronouncements. They all start out with a bang and some hope of gaining head wind then fail the current.
>What forum are you referencing?
The first information (in Russian) is taken from gladiators-chess.ru forum, the screenshot (+testing) is taken from another Russian forum but that link (forum) is not allowed here, but you can "Google it" (if you wish) with the words "Тесты Стрелки/Strelka" [Post #366 in this thread].
I actually may be a member of that forum. I gave up going there because of the language difference and the pain of translating through google chrome.
>I gave up going there because of the language difference and the pain of translating through google chrome.
One useful tip: if you use Google translate, it is sometimes better to type one by one word at a time instead you use copy-pasted method with entering the whole text. Unfortunately, this is a very slow method...
Also, from my point of view, English grammar is much more complicated than my Cro grammar
, the problem is that sometimes Google Translate does not give me some good advices , and then it occurs what I call - "Tarzan english".
No! It is the built in "web page" translator" - sometimes it works and sometimes it becomes inaccessible.
>No! It is the built in "web page" translator" - sometimes it works and sometimes it becomes inaccessible.
BTW, of course, I know for Google Chrome and his implemented translator
, but I'm talking (in my case; Opera@USB 11.62) about the "classical" Google Translate service which has nothing to do with Google Chrome.For example, with the help of Google Translate service, our friend Nelson (almost
) "became an expert" for the Croatian language.
I don't know about Norm-he may or may not have it in him to go the extra mile. He has come close with Fire and it did show some promise. Robbo .10 may be on par with Critter 1.4.a. -I've done only minimal testing so that remains inconclusive. What strikes me there is the developers did way with the RobboTables and there are scant parameters available. But its strength is apparent.
>I don't know about Norm-he may or may not have it in him to go the extra mile. He has come close with Fire and it did show some promise. Robbo .10 may be on par with Critter 1.4.a. -I've done only minimal testing so that remains inconclusive.
BTW, of course, I am not a programmer or compiler, so it is difficult for me to talk about something I do not have much experience or knowledge
, but I have the impression that people (compilers) like Norman S. or @PetarPan (etc.), they have achieved the maximum from the last available IH source-code, the problem is (for fans/compilers of that engine) that after was the IH 999946 source was published, there was some long period without that something significant (or anything?!) was happened from the "native" Robbo/IH programmers (or "Decembrists"), therefore it is possible that they maybe do not want to work on the improvement of that engine anymore, or maybe they lost interest after Houdini became a commercial engine (it is possible that they feels a bit "cheated and exploited" in this regard, therefore, viewed from their perspective, why would they continues to helping [non-indirectly] to author of the Houdini engine (for example), now when the Houdini also became a commercial product and that "in their eyes" means something like [another one] "a greedy capitalist", or something similar... )Therefore, it is very possible, that they do not have interest to working on this IH project anymore, on the other hand, AFAIK, Jury Osipov is something like a "one man show"
, in his case everything is possible , I would not even be surprised, if he would indeed managed to do it with Strelka 5.5 engine.
The other thing is: Deep Fritz 13 is due out mid-May.
Magnus changed his account,he has a new one
Oh, how did you find out? What is his new user name?
And are you Barnard?
And are you Barnard?
same style writing,same poor english,same stupid things,etc...
i will tell you in private
yes,i am,you can see in my profile,i just changed my name,not opened a new account
i will tell you in private
yes,i am,you can see in my profile,i just changed my name,not opened a new account
>Magnus changed his account, he has a new one
Who is that Magnus ?
Which Magnus ? BTW, which was his previous account/nickname?
Did he was banned here?
Magnus is an annoying guy asking always the same type of speculative and annoying questions,and of course,questions that no one has the answer
His name is Magnus Friedman
And not,he never was banned,not at least that i know
His name is Magnus Friedman
And not,he never was banned,not at least that i know
1
2
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill