Dinosaurs and Mammals

[smrf]

Current top engines seem to be highly related. How could there be really new approaches to chess programming, if competitions are always done against those top dogs? It looks like an era of dinosaurs, where it does not make sense for a new species to initially challenge just a kind of Tyrannosaurus Rex. Instead there an alternative ecological niche has to be established, where completely new (at least independently created) ideas might evolve. This implicates that those chess programs do not consist of copied and pasted parts from other code, including open source. It seems to be very hard to create such a trusted scene of independent ideas. But there is no chance to experience the coming up of a new possibly superior species like mammals by constantly deriving only from dinosaurs.

[zullil]

Current top engines seem to be highly related. How could there be really new approaches to chess programming, if competitions are always done against those top dogs? It looks like an era of dinosaurs, where it does not make sense for a new species to initially challenge just a kind of Tyrannosaurus Rex. Instead there an alternative ecological niche has to be established, where completely new (at least independently created) ideas might evolve. This implicates that those chess programs do not consist of copied and pasted parts from other code, including open source. It seems to be very hard to create such a trusted scene of independent ideas. But there is no chance to experience the coming up of a new possibly superior species like mammals by constantly deriving only from dinosaurs.

It took an asteroid collision to give mammals a chance.

[smrf]

It took an asteroid collision to give mammals a chance. :D

But they already have been there. Even as small as mice they evolved separately. Otherwise we would not exist - with or without asteroid impacts.

[hgm]

What bothers me most is the trend that everything has to be multi-core in order to compete. Making a program SMP is an advanced programming problem, that has absolutely nothing to do with Chess. It would be much more productive if people would devote their efforts to finding better search strategies and better evaluations rather than all solving the same technical problem.

[Evert]

What bothers me most is the trend that everything has to be multi-core in order to compete. Making a program SMP is an advanced programming problem, that has absolutely nothing to do with Chess. It would be much more productive if people would devote their efforts to finding better search strategies and better evaluations rather than all solving the same technical problem.

True. Of course, writing an SMP search is much easier than coming up with better search strategies and evaluation.
All SMP does is let you use more CPU time for your search. It doesn't improve the engine in any qualitative sense. I added an SMP search to Jazz because I wanted to get my hands dirty in parallel programming, and to proof to myself that I could do it. In terms of playing strength it's not really worth it, since there are tons of single core engines that are stronger than Jazz.

I think the most innovation for chess programs (in a broad sense, ie, if we don't limit ourselves to orthochess) will come from chess variants. Or things like retrograde analysis to solve particular problems.

[zullil]

It took an asteroid collision to give mammals a chance.

But they already have been there. Even as small as mice they evolved separately. Otherwise we would not exist - with or without asteroid impacts.

Yes, I know mammals were already present. And I suspect there are some chess engines already here that are differently designed than the "dinosaurs" that now dominate. But, like mice, they are weak and cannot compete. Without something akin to an asteroid strike, they will remain in the shadows, and likely head toward extinction.

[hgm]

Elephants are scared to death by mice. Why would T. Rex be different?

[cdani]

I think there are a lot of possibilities near us, just must be picked. But I understand that when there is a lot of information on anything, people tends to use it instead of thinking all from start by themselves. Andscacs is an example, where I was able to do a relative strong program with relative little time, even if I have written and thought all the code by myself, and there are some original ideas on it. But the base of its strength are ideas present everywhere on the programming chess world.

Strictly speaking all this is not bad, just the usual in every branch of knowledge in the history. There are no impairments to change anything, just our own preconceptions. Sometimes one must not take the safe path. And here, in our little world, the risk is zero; on the contrary the change it's fun!

[jdart]

I don't think the top engines are highly related.

Stockfish does much more pruning than other engines, reaches higher depths and has a significantly different eval.

Komodo and Houdini are similar in speed and depth reached but I don't think the eval is similar. They not infrequently prefer different moves.

--Jon

[zullil]

I don't think the top engines are highly related.

Stockfish does much more pruning than other engines, reaches higher depths and has a significantly different eval.

Komodo and Houdini are similar in speed and depth reached but I don't think the eval is similar. They not infrequently prefer different moves.

--Jon

Well, we're getting into taxonomy here, and the endless conflict between "lumpers" and "splitters". To me these engines are all "dinosaurs", all more or less using the same standard approaches to chess programming. Why? Because these approaches work, and engines that use them survive (and reproduce, often by cloning ). My impression was that the OP was interested in "mammals", engines that played chess using fundamentally different approaches. These must exist, hiding in the jungle, hopelessly outclassed for now, if not forever.