future of top engines:how much more elo?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
stavros
Posts: 160
Joined: Tue Dec 02, 2014 12:29 am

future of top engines:how much more elo?

Post by stavros » Thu Feb 12, 2015 11:21 pm

after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..

carldaman
Posts: 2004
Joined: Sat Jun 02, 2012 12:13 am

Re: future of top engines:how much more elo?

Post by carldaman » Thu Feb 12, 2015 11:34 pm

stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Another 100-200 Elo should be achievable within the foreseeable future (next 3-5 years).

Let's not forget that there's plenty of room for obvious improvement - when the engines are playing without a book or with a short book, their play is highly exploitable when faced with skilled anti-computer human expertise.

The one thing that slows down progress in this area is the nature of testing, where most of it is engine vs engine, as this masks the top engines' major weaknesses.

Regards,
CL

User avatar
Laskos
Posts: 10823
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: future of top engines:how much more elo?

Post by Laskos » Thu Feb 12, 2015 11:35 pm

stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.

stavros
Posts: 160
Joined: Tue Dec 02, 2014 12:29 am

Re: future of top engines:how much more elo?

Post by stavros » Fri Feb 13, 2015 12:54 am

Laskos wrote:
stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.
in the past was "easy" to gain 150 elo in 1 year ,recently its too tough to gain 30-40 in 1 year ,so still i cant imagine such a good software to gain more than 100-150 elo ,i think we will reach the limit soon but this is just a guess,it would be interesting to get some opinions from larry ,bob etc.
i cant imagine engine 600 elo stronger than stf6 komodo 9 etc...

deefree49
Posts: 15
Joined: Sun Dec 07, 2014 2:14 am
Location: Columbus, Ohio

Re: future of top engines:how much more elo?

Post by deefree49 » Fri Feb 13, 2015 1:57 am

Interesting question. You have to keep in mind that computer hardware is still improving every year. It is a field that gets a lot of attention and there's plenty of resources available to support further research. Therefore, I believe growth in this area will inevitably continue and may even accelerate.

It is also possible that some discovery may open up a larger leap forwards. If 'quantum computing' is achieved it may send capabilities through the proverbial roof. Computers may become faster by exponential leaps.

The other front is programming. This area is slower to improve but there still is probable progress to be made. Perhaps some genius programmer will discover a new type of search that allows a leap forward in chess knowledge. I'm just speculating but maybe "fuzzy logic" might be an area to explore.

I think the future is likely to hold some real surprises. Remember, it was only 50 years since all this started. I think that in the next 20 years, we may see ratings crack the 4,000 barrier. No, I don't believe chess will ever be "solved" but we may see computers playing fantastic and even artful chess that all players will be completely awed over.

I am still very optimistic about the future of chess engines and computer chess.

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: future of top engines:how much more elo?

Post by Lyudmil Tsvetkov » Fri Feb 13, 2015 2:02 am

stavros wrote:
Laskos wrote:
stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.
in the past was "easy" to gain 150 elo in 1 year ,recently its too tough to gain 30-40 in 1 year ,so still i cant imagine such a good software to gain more than 100-150 elo ,i think we will reach the limit soon but this is just a guess,it would be interesting to get some opinions from larry ,bob etc.
i cant imagine engine 600 elo stronger than stf6 komodo 9 etc...
I fully support Kai here.

I would say 600-700 elo, maybe way above 1000, just on software.

stavros
Posts: 160
Joined: Tue Dec 02, 2014 12:29 am

Re: future of top engines:how much more elo?

Post by stavros » Fri Feb 13, 2015 2:43 am

Lyudmil Tsvetkov wrote:
stavros wrote:
Laskos wrote:
stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.
in the past was "easy" to gain 150 elo in 1 year ,recently its too tough to gain 30-40 in 1 year ,so still i cant imagine such a good software to gain more than 100-150 elo ,i think we will reach the limit soon but this is just a guess,it would be interesting to get some opinions from larry ,bob etc.
i cant imagine engine 600 elo stronger than stf6 komodo 9 etc...
I fully support Kai here.

I would say 600-700 elo, maybe way above 1000, just on software.
looks unreal.. maybe but the difficulty to write such a software would be
beyond the known programming patterns and algorithmics,sure there is big room to gain elo but with software only how u can get 500 or 1000 elo more? u need 100% knowledge of chess!! also u need the fastest possible search at the same time ! in theory yes but how can a programmer put that knowledge ? i remember the first rybka 1.0 beta had 5+ MB size but "slow" search and we assumed was due to more knowledge.now sf6 uses <1MB and plays 350 elo more with less knowledhe and way faster search
the only road i see is the monte carlo algorithm but meeds a super fast machine

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: future of top engines:how much more elo?

Post by Lyudmil Tsvetkov » Fri Feb 13, 2015 3:44 am

stavros wrote:
Lyudmil Tsvetkov wrote:
stavros wrote:
Laskos wrote:
stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.
in the past was "easy" to gain 150 elo in 1 year ,recently its too tough to gain 30-40 in 1 year ,so still i cant imagine such a good software to gain more than 100-150 elo ,i think we will reach the limit soon but this is just a guess,it would be interesting to get some opinions from larry ,bob etc.
i cant imagine engine 600 elo stronger than stf6 komodo 9 etc...
I fully support Kai here.

I would say 600-700 elo, maybe way above 1000, just on software.
looks unreal.. maybe but the difficulty to write such a software would be
beyond the known programming patterns and algorithmics,sure there is big room to gain elo but with software only how u can get 500 or 1000 elo more? u need 100% knowledge of chess!! also u need the fastest possible search at the same time ! in theory yes but how can a programmer put that knowledge ? i remember the first rybka 1.0 beta had 5+ MB size but "slow" search and we assumed was due to more knowledge.now sf6 uses <1MB and plays 350 elo more with less knowledhe and way faster search
the only road i see is the monte carlo algorithm but meeds a super fast machine
Current Stockfish has much more knowledge than Rybka 4, and also much better search.

Yes, it is going to take more refined, square-specific chess knowledge, and much better understanding what moves to look at preferably and which main sequences to follow.

But it is doable.

bob
Posts: 20918
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: future of top engines:how much more elo?

Post by bob » Fri Feb 13, 2015 4:42 am

stavros wrote:
Laskos wrote:
stavros wrote:after the latest stf6 komodo 8 etc, i wonder how much elo more can be achieve via programming,on the same hardware of course.
100,200,300 elo more? it could be a poll but anyway just a "food of thought"
my personal feeling not more than 100 elo. more? it would be a miracle
dont forget pls on the same hardware! lets say a medium pc 2 core etc..
Assuming, after many extrapolations, that perfect engine has ~4,500 Elo points, till chess is solved, 600 points will be gained by software, 600 by hardware, so my bet would be 600.
in the past was "easy" to gain 150 elo in 1 year ,recently its too tough to gain 30-40 in 1 year ,so still i cant imagine such a good software to gain more than 100-150 elo ,i think we will reach the limit soon but this is just a guess,it would be interesting to get some opinions from larry ,bob etc.
i cant imagine engine 600 elo stronger than stf6 komodo 9 etc...
Elo is a relative measurement. All you need to see an elo increase of 600 is to introduce a new program that is 1200 stronger than existing programs. Everyone will migrate up the Elo list a bit except for those (most likely human) that simple can't compete whether it be 2800 or 3400 they are playing against.

I don't see any rational way to extrapolate "the ultimate Elo". One has to make too many assumptions. It might well be that chess is drawish from the starting position, or it might be almost completely winning for one side. Once an engine or two penetrates that draw barrier and begins to see deep enough to push the draw rate down, assumptions fail. Ergo I don't see any reason for worrying about something that is speculation at the very best. There will always be "that next new idea" that produces a significant advancement. They tend to come in spurts rather than steady progress. Spurts, then many refinements of that idea, then another spurt.

Extrapolation might sometimes be fun, but it is ultimately pointless. They will get to wherever they get, factoring in hardware and software advances. One day, quantum or optical computing might provide the next huge hardware jump. Or they might provide nothing interesting at all. My guess, based on watching this happen since the late 60's would be that there is no upper bound one can establish. Given the Elo system, you can easily calculate the upper bound in rating for a program that wins 100% of the games against a program with a legit rating of 3200. But that bound is artificial since it is relative to the best there is at that moment. But the best there is can advance also.

You might try solving the old "how many angels can dance on the head of a pin?" question. About as useful. :)

Sorry to have to real insight. This is one of those theoretical questions that drive some to drink, others to new career choices; while the "smart ones" don't give it any serious thought since there are simply way too many unknowns.

bob
Posts: 20918
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: future of top engines:how much more elo?

Post by bob » Fri Feb 13, 2015 4:47 am

deefree49 wrote:Interesting question. You have to keep in mind that computer hardware is still improving every year. It is a field that gets a lot of attention and there's plenty of resources available to support further research. Therefore, I believe growth in this area will inevitably continue and may even accelerate.

It is also possible that some discovery may open up a larger leap forwards. If 'quantum computing' is achieved it may send capabilities through the proverbial roof. Computers may become faster by exponential leaps.

The other front is programming. This area is slower to improve but there still is probable progress to be made. Perhaps some genius programmer will discover a new type of search that allows a leap forward in chess knowledge. I'm just speculating but maybe "fuzzy logic" might be an area to explore.

I think the future is likely to hold some real surprises. Remember, it was only 50 years since all this started. I think that in the next 20 years, we may see ratings crack the 4,000 barrier. No, I don't believe chess will ever be "solved" but we may see computers playing fantastic and even artful chess that all players will be completely awed over.

I am still very optimistic about the future of chess engines and computer chess.
Clearly silicon is not going to continue to advance as fast as in the past. And with no real viable alternatives, that form of processor design is really reaching a wall. Obviously you can't make a wire from a single atom of copper, and just as obviously you can't make a wire from something less than an atom, if you want it to conduct electrons. We will see additional cores for a few years, but that is a pittance compared to what we were seeing 20 years ago going forward. I don't think there is any ultimate barrier. Concepts like "100% of chess knowledge" are meaningless. How can one quantify 100% of something that is probably infinite in size? Same for search strategies. There's room for improvement everywhere, and we will continue to see it so long as a few keep carrying the ball forward.

Post Reply