Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / bad news
Parent - By Carl Bicknell (*****) Date 2012-06-23 10:21
Very sad.
Parent - By Master Om (Bronze) Date 2012-06-23 10:53
Bad. Whats the cost to organize such tours ?
Parent - By Fulcrum2000 (****) Date 2012-06-23 11:26
Sad, but not totally unexpected :cry:
Parent - - By Banned for Life (Gold) Date 2012-06-23 14:29
This is not very surprising. Many of the engine developers are psychopaths. Some are followed around by dung beetles. One can't expect great things in this environment.
Parent - By Labyrinth (*****) Date 2012-06-23 14:37
Maybe c++ rots your brain? Or does it just make you an angry person?

Carmack never seemed angry to me. We did recently see Linus Torvalds literally giving the finger to Nvidia. Who knows.
Parent - - By Nelson Hernandez (Gold) Date 2012-06-23 14:44
I've said it before, I'll say it again.  The whole situation could be remedied by having two biannual world championship cycles, one with limited and identical hardware and one with unlimited hardware.  With real prize funds and a modest amount of publicity.  There's no reason in the world why some of the top chess servers couldn't collaborate on something like this.  The hobby desperately needs it as a focal point of interest and achievement.
Parent - - By Banned for Life (Gold) Date 2012-06-23 15:31
You've proposed a solution that would be good for the hobby, but not necessarily for the hobby's leaders and their second rate engines. Bob never hid the fact that he was fine with the idea of having only a few developers, and not necessarily the better ones, at the competitions. The hobby's trajectory will remain as is as long as this leadership persists.
Parent - - By Nelson Hernandez (Gold) Date 2012-06-23 21:05
The problem with your comment is that you called them leaders rather than "so-called leaders".  I don't accept the ICGA as anything other than an overreaching private organization with delusions of grandeur and a clubby relationship with a group of programmers wedded to a superannuated status quo.  That is a unfortunate for the hobby because ten years ago and earlier they held what were generally recognized as the most consequential computer chess tournaments in the world.  Now they have generated so much ill will I don't see how they can lead the hobby anywhere except to oblivion.

A motivated group of people with experience and resources could put together a wonderful set of recurring competitions.  I know it could be done.  All that is needed is the will to make it happen.  Consider what great venues there are out there: wherever there is a high density of computer nerds with time on their hands.  Technical universities are an obvious place for such an event.
Parent - - By Banned for Life (Gold) Date 2012-06-23 23:49
I'm not so sure that engine-engine tournaments aren't going the way of the dinosaurs. There was very little interest in the events, even before the latest fiasco. If some group were actually going to try to breath life into this arena, they would do well to consider a freestyle, rather than pure engine model. The freestyle events have generated at least an order of magnitude more interest and enthusiasm, and they do a much better job of exploiting the added dimensions that you have alluded to above.
Parent - By Harvey Williamson (*****) Date 2012-06-24 00:19

> they would do well to consider a freestyle


Voila!

http://www.talkchess.com/forum/viewtopic.php?t=44168
Parent - By Nelson Hernandez (Gold) Date 2012-06-24 01:28
In terms of developing a far-flung array of international teams, sure, that plan works.  Freestyles are democratic and meritocratic at the same time.  Certainly the PAL/CSS tournaments were fun and interesting; what fueled them were the prize funds and generous publicity on Chessbase.  If something on that scale could be resuscitated that would be a step forward for the hobby, just as Linares or Wijk aan Zee are prestigious tournaments that are good for chess in general.

But what I am talking about is having an event that declares a program/programmer the "world champion".  An event that follows human forms to some extent: a championship number of rounds, classic time controls, and two programmers (selected through a candidates match) facing off head-to-head.  Set up a live broadcast, have Chessbase and/or Chess.com publicize it, have some newspapers or magazines pick it up as a novelty story, and presto--you have an event that could potentially pick up steam as it went on. 

Throwing in some tabloid angles to the story would make it all the better, e.g. "Houdart, the brilliant but sometimes touchy author of the #1 rated program, against Rajlich, the devil-may-care Czech-American whose program won almost every tournament it entered for five years"...gallons of ink could be spilled.  Speaking of Rajlich, what the blazes is he up to these days?
Parent - - By Scott (*****) Date 2012-06-24 01:44
If it's Freestyle and you are playing as a Centaur then you shouldn't have to divulge your engine or hardware beforehand. I myself use different engines for different parts of the game.
Parent - By Harvey Williamson (*****) Date 2012-06-24 02:14
I suggest you ask peter. I don't think he, yet, realises what he is letting himself in for by making it a freestyle :)
Parent - - By Bob Durrett (***) Date 2012-06-24 18:59
Scott, I find your statement that you use different engines for different parts of the game to be fascinating!

In fact, last night as I lay there half asleep, my mind racing and so many questions kept coming ny into my head.

Here are a few of them:

(1) Do you manually switch between commercially available engines as when playing correspondence chess?
(2) Are the different engines a part of software you created yourself? Did you write all these engines yourself (Note: Using borrowed code is OK)?
(3) Are the different engines completely independent of each other? Or, do you go from one engine to the next by simply switching out certain subroutines, subprograms, or modules and then switching in others?
(4) As you switch from one engine to the next, do you lose your hash and have to start over on that?
(5) Do you have separate software running in the background which is monitoring the positions and determining when to change "engines"?
(6) Are ongoing search activity terminated and the results lost when you switch?
(7) Do you switch during the process of evaluating an end (leaf) position?
(8) How many distinct engine configurations are you currently using? For example, do you have separate middlegame engines for closed, semi-closed, open and semi-open positions?
(9) Do you accomplish switching by changing values of constants (which were declared in the source code)?

I have more than thirty questions, but the above should do for starters.

I am particularly interested in this because doing something like this is a suggestion I made at CCC several decades ago.

Bob Durrett
Parent - By Scott (*****) Date 2012-06-24 21:47
Hi Bob, I'm not an expert, but I'll answer as best I can. #1=yes, #2=no, #3=independent of each other, #4=yes, you lose hash, #5. I make the decision when to switch, #6=yes, #7=yes, #8=yes, this is ongoing trying to decide which engine is best for which situation. #9=no, that's too much for me, :)

I have been researching this for a while and I'm convinced certain engines for example play endgames better than others etc., while the "others" may play the opening and middlegame better. It's just amateur research, hope I helped! :smile:

Scott
Parent - - By Uly (Gold) Date 2012-06-24 21:54
I know you didn't ask me but I found interesting answering those questions :lol:

> (1) Do you manually switch between commercially available engines as when playing correspondence chess?


I do and a lot and my analysis relies on finding the right engine for the position and analyzing with it, but sometimes the best engine for a game changes from one position to the next! That's why I need to keep monitoring the game with several engines, and see which ones are better than others as the game advances. For an idea (not examples) of how this progresses check http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=21554 . For examples without moves check http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=417321#pid417321

> (2) Are the different engines a part of software you created yourself?


No, but some engines have implemented features that I have suggested, such as Critter Session File, Rybka Visualizer, or Stockfish Preserve Analysis,

> (3) Are the different engines completely independent of each other? Or, do you go from one engine to the next by simply switching out certain subroutines, subprograms, or modules and then switching in others?


The engines are completely independent. They are used without modification and are switched in the GUI on the chess positions that require it.

> (4) As you switch from one engine to the next, do you lose your hash and have to start over on that?


I do, probably other people have more engines loaded simultaneously without losing the hash, but I found out the decrease in analysis is insignificant, unless it's an engine that relies on hash like Stockfish Gran2 Zappa Mexico II or Rybka 4(.1 with PA On). I usually have 2 engines loaded, the main engine, or the one with the main idea I want to refute, and a challenging engine, or refuting engine, that plays the other side. The engine that "loses the fight" gets unloaded and loses the hash, while the other one remains and awaits for another engine to challenge it/try to refute it, or I've found my main move.

Since chess only has two sides, this is very effective (as it decreases the redundancy to 0 in deep lines, because one engine doesn't analyze the other side.) If chess had three sides, I'd cross-check with 3 engines simultanously.

> (5) Do you have separate software running in the background which is monitoring the positions and determining when to change "engines"?


No, I do it manually. And I don't think this aspect of the process could be automated, as most of the time it relies on intuition (seeing what kind of position is on the board, and knowing what kind of engine is best for that kind, like, with possibility of material imbalance I'd fire Rybka 3 Dynamic, if I seem to be gaining large advantage, I fire up Stockfish, etc., other main engines like Critter or Rybka 4 are loaded in all positions by default.)

> (6) Are ongoing search activity terminated and the results lost when you switch?


No, I store my analysis in Bookup. I guess most people would use Aquarium Trees for that. Also, engines with "Learning" like Critter 1.6 or Rybka 3 Persistent Hash take care of keeping results at the search level.

Unfortunately, no current free alternative exists for storing analysis, if you want to store it you have to got commercial.

> (7) Do you switch during the process of evaluating an end (leaf) position?


Yes if the line has been refuted (the refuted engine is switched out), otherwise changes usually happen near the root (once all the scores of all engines have been backtracked to the root.)

> (8) How many distinct engine configurations are you currently using? For example, do you have separate middlegame engines for closed, semi-closed, open and semi-open positions?


I don't separate them in game stage but on kind of position (closed, open, dangerous, advantageous, drawish, winning, losing, material imbalance, passed pawns, king attacks, etc.) Usually, endgames are played with the engines that were best in late middle game. I don't have preset configurations and instead build my team of engines on the fly, as usually the ones that were best in some closed position are not that good in another closed position even though the positions look very similar to the human eye. Each chess position is unique and engine "configuration" (the team of engines used in a position, and their order) is very dynamic.

> (9) Do you accomplish switching by changing values of constants (which were declared in the source code)?


I don't modify engines. But engine personalities are very useful depending on position goal. So I use Rybka Mindbreaker Settings, Zappa Mexico II Dissident Settings, Stockfish FLFH version or even old versions of engines like Critter 0.9 if I think the position demands it.

By the way, most of the engines I use are closed source, the only strong exception is Stockfish.
Parent - By Scott (*****) Date 2012-06-25 00:30
Much better answer than mine, :grin:
Parent - - By Geomusic (*****) Date 2012-06-24 04:40
We need a Deep Blue machine again to spark CC interest
Parent - By Banned for Life (Gold) Date 2012-06-24 04:49
As in many other types of competition, the differences make the competition more interesting. The requirement for everyone to use the same hardware might strike some as egalitarian, but this doesn't make for a better event. Having everyone using radically different hardware would be a lot more interesting.
Parent - - By Barnard (Bronze) Date 2012-06-23 22:17
Nelson,what is the point to run a tournament with unlimited hardware?that tournament only will be win with the stronger cluster...

imagina Hiarcs,that is near 200 elo points below rybka...and lets supose im rich,and i build a 8192 core cluster,AND Hiarcs can use it...Hiarcs will win,but being too worst than the other programs

the point is to have a tournament with ALL the engine allowed to play it,and limited and equal hardware,so we can see the strongest engine winning (in theory)

the unlimited hardware is only a course for fools...one of the pasts icga events,Jonny running on 800 cores...that is simply,absurd;wasting that very big amount of money to feed the ego of some people...that money can be very good invested if were donated to an ONG,and not just to feed up some egos
Parent - - By Pia (****) Date 2012-06-23 23:04
But least after the match there will be 8192 core Hiarcs playing stronger chess than ever.
And everyone would be trying to beat it with a more powerful hardware or smarter software.
Then we'll get even more games on the way to perfection :smile:
Parent - By Barnard (Bronze) Date 2012-06-24 19:09
that is a complet waste fo money to feed up some fools ego;too much better use the money to build a school in Africa,for example...or send food to the people that dies because they havent food or even water
Parent - - By Nelson Hernandez (Gold) Date 2012-06-23 23:33
As was demonstrated by the event you cite, more cores do not equal invincibility.  What matters is hardware combined with software, with software encompassing chess engines, opening books, and the panoply of technicalities falling under the broad heading "implementation".  The point being that we cannot predict the future or how either hardware or the various sub-branches of software will evolve.  All you can do is create a competitive structure where the best in the world can compete and are properly incentivized to do so.
Parent - - By Barnard (Bronze) Date 2012-06-24 18:35
you are right Nelson,but that wasnt my point

my point is that if 'x' team can afford to create a cluster of 16384 cores AND take benefit of the cluster with a very good engine that cna use all the capabilities of the cluster,that 'x' team will win all the tournaments,and i think that is just to feed up some persons ego's

what i think is too much better is,instead of feeding up some ego with that monters,use the money to help a ONG because a lot of people dies

and i think you will agree with me,that has more sense (lets asume the cluster costs 200.000 euros) give that money to buld schools in Africa,or help people who dies for the AIDS,than feeding up the ego of a person that thinks:''i build that cluster and i win all the tournaments''
Parent - By Harvey Williamson (*****) Date 2012-06-24 18:46

> give that money to buld schools in Africa


It would buy a lot of Fish. I can supply some recipes
Parent - - By Nelson Hernandez (Gold) Date 2012-06-24 19:43

> give that money to buld schools in Africa


Following this line of thinking every potential venture, investment or initiative in the world would need some central authority to determine its nobility, virtue, justice, moral worth, etc. on some relative scale.  Such a program of systematic value judgment doesn't sound the least bit appealing to me, as it would in a short time enslave the entire world.  A system of ordered liberty is the optimal modus vivendi mankind has thus far managed to devise.

What is a better use of your time, building schools in Africa or participating in computer chess?

I would argue that I do not have the right to judge that in your case, much less the right or authority to enforce your doing one thing or the other.  It is for you to decide and for the rest of us to accept man's nature as it is, rather than as some utopian wishes it ought to be.
Parent - By Barnard (Bronze) Date 2012-06-24 23:23

>What is a better use of your time, building schools in Africa or participating in computer chess?


now that you made me that question,i will answer you:go to Africa,and help people...that is why i have plans to go to live to Zambia,one of the poorest countries on the world...

you can ask Paul,since i told that to him a lot of time ago
Parent - - By Banned for Life (Gold) Date 2012-06-23 23:40
imagina Hiarcs,that is near 200 elo points below rybka...and lets supose im rich,and i build a 8192 core cluster,AND Hiarcs can use it...Hiarcs will win,but being too worst than the other programs

Nope. It doesn't work that way. Hiarcs on an 8192 core cluster won't be any more competitive than Hiarcs on a 16 or 32 core machine, where its performance tops out. It takes a lot of time and effort to build a good cluster. Vas has actually done it. Team Hiarcs has not. And how well did Jonny do on 800 cores?

Do you believe that all competitions should be single core so that Komodo can compete on an equal playing field? Or in disabling TBs because Stockfish doesn't support them? Setting the bar really low to allow all to compete makes for a pretty boring competition.
Parent - By keoki010 (Silver) Date 2012-06-24 01:03
+1 totally agree on this. It's not necessarily the cores but the software search that determines the outcome. :grin:
Parent - - By Barnard (Bronze) Date 2012-06-24 18:30
not Alan,competitions not should be in 1 core,or without tb

what i mean is if we allow unlimited cores in the competitions,the guy with better hardware surely win 80 % of the events

of course is very very very difficult do a cluster like rybka cluster,and make the software works on it

what i mean is,for example,hiarcs team is able to do the same than vas did with the cluster,and also lets supose that 1 billionaire buys 16384 core cluster AND hiarcs team is able to run with eficency the Hiarcs engine on the cluster,surely will win at least 90 % of the tournaments

but that wasnt my point;my point was that,instead of investing that huge amount of money with that clusters,that money can be better use to help a ONG to help people who dies because they havent food to eat...that is my point,that instead of wasting that amoun of money to feed up some ego's,that money can be better invested helping a ONG
Parent - - By Banned for Life (Gold) Date 2012-06-24 22:21
First, it took me a while to figure out that what you call an ONG, I call an NGO...

I am not convinced that most NGO effort isn't wasted or even counterproductive. There are many places that require better schools, or maybe micro-loan programs to provide start-up capital for small businesses. Any aid that doesn't provide a pathway to self sufficiency is doomed to failure. As an aside, my sister used to hang out with a fair number of people that run some large NGOs here in the US. As far as I was concerned, these people were the scum of the earth...

But getting back to the much more interesting discussion about money in computer chess, I'm not convinced that the Hiarcs team has less access to resources than Team Rybka. They just don't have the high end cluster software to take advantage of the resources. It's not Vas' fault that only a few people have gone through the effort to develop this type of software. It's really the same issue as playing MP engines against Komodo and then requiring the other engines to play on a single core because Don hasn't implemented MP on Komodo.
Parent - By Barnard (Bronze) Date 2012-06-24 23:43
i never atacked vas for the rybka cluster that he can has,and others not

im just telling that i honestly think that that huge amount of money can go to a NGO (ONG in Spanish,im sorry),and will be more helpful to help people that for example,hasnt water

rybka 300 cores,jonny 800 cores,and maybe in a few time,we will see 1200 cores to run a program...

the only i tell is that very huge amount of money can help a lot of people that really need it,im not telling more Alan
Parent - - By Bob Durrett (***) Date 2012-06-24 01:37
It seems to me that there is some merit to having tournaments with unrestricted hardware but only if the motivation for entry is enticing enough. A large incentive would stimulate the development of stronger and newer kinds of chess playing machines. There is also the potential to develop public interest, attracting rich sponsors.

If all work is limited to making small improvements to Hyatt's open source software, using inexpensive computers, then where will be the stimulus for creativity, innovation, and real progress in the world of thinking machines? A long time ago, Dan Corbett once told me (in a bulletin board post responding to something I had posted) that the then existing algorithms were the only way to go. But the World changed and we now have completely different kinds of hardware with seemingly unlimited possibilities for completely new kinds of hardware. What will motivate chess programmers to write programs to run on the new kinds of machines?

For me, of course, I care most about software I can afford. Fortunately, however, there are people out there with more money than they know what to do with.

Just my opinion. :-)

Bob Durrett
Parent - - By Nelson Hernandez (Gold) Date 2012-06-24 02:15
Here are some traditional motivations and incentives:

1.  Power
2.  Money
3.  Egoism
4.  Glory
5.  Sex

#1 has no application to computer chess, though Bob Hyatt and gang have made a valiant effort.  Computer chess could potentially and legitimately offer opportunities in #2, #3 and #4.  As for #5, typically devotion to computer chess renders purely imaginative all potential activity in that area, unless you are willing to trade #2 for #5.
Parent - By Homayoun_Sohrabi_M.D. (***) Date 2012-06-24 02:27

> As for #5, typically devotion to computer chess renders purely imaginative all potential activity in that area, unless you are willing to trade #2 for #5. <IMG class=qButton title="Quote selected text" alt="Quote selected text" src="/mwf/mi_quote.gif" width=20 height=10>


:lol::lol::lol:
Parent - By InspectorGadget (*****) Date 2012-06-24 16:33
I blame it on the IPPOLIT :sad:
Up Topic The Rybka Lounge / Computer Chess / bad news

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill