Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / What software enable phone komodo to beat quad shredder 10
- - By rocket (****) Date 2019-03-16 09:34 Edited 2019-03-16 12:25
It's not depth going by their respective numbers, and not number of positions analyzed (obviously).

So what are the improvements in the algorithm that enable such a handicapped modern engine to beat quad core old engine?

It's not Shredder 10 specifically, any engine of that era it beats running on a mere smartphone.

Even assuming the eval is vastly superior, how could this compensate all that processing power?

Shredder had a fairly balanced material value table, and eval in general.
Parent - By Vegan (****) Date 2019-03-16 15:18
Recently I moved from a quad core CPU to 8 threads, CPU now is a R5 2400G

Not much improvement in gameplay for any particular engine

My iPhone 7 plus is not nearly as powerful so I do not use it for any games
Parent - By Labyrinth (*****) Date 2019-03-16 22:24
Vastly better pruning. The nodes it does analyze are more relevant and this makes all the difference in the world.
Parent - - By Carl Bicknell (*****) Date 2019-03-16 22:41
Pruning. SF is able to work out better which moves DON’T need analysis, so it can spend more effort on moves which do. Given how large a search tree can become in chess, good pruning can compensate for many times the processing speed of a lesser engine.

Another way of looking at it is that every speed doubling is worth 50 Elo. So work out the elo difference between SF and Shredder 10 on equal hardware, now divide that number by 50. That’s how many speed doublings your quad has to be faster than the phone. And it just can’t do it.
Parent - - By rocket (****) Date 2019-03-17 11:57
What if Shredder 10 had 1 year per move vs Stockfish 10's 30 seconds, would Shredder win then?
Parent - By Vegan (****) Date 2019-03-17 21:49
shredder is an old engine and nowhere nearly as polished as what stockfish has accomplished

i only play tournaments with open source as too much plagiarism out for my tastes
Parent - - By Labyrinth (*****) Date 2019-03-18 01:09
Probably yes, if it could be configured with an enormous permanent hash that would withstand so much analysis, although there's still a possibility that it wouldn't win because shredder would also be pruning good moves early on in some cases, and this would happen less with SF. It would be great to see the games though, but this is obviously infeasible.
Parent - By rocket (****) Date 2019-03-18 05:23
Will a move that is pruned out of the search ever be digged up again with enough time or is it forever discounted by the engine? I mean literal pruning is that the move is cut out of the search tree so how could it ever resurface?
Parent - - By rocket (****) Date 2019-04-05 18:37
You might not know this but in human chess, superior pruning also accounts for why Kasparov, Carlsen etc beats the other top guys. It's a curious organizing feature of their brains that allow them to discard lines that their opponents waste time on, and get to the core of the position.

In computer chess however, at least back in the Fritz days, evaluation accounted for most of the descisive games, not search, not depth, not tactics.

Vas talked about how master level human games are mainly decided by tactics, while elite computer games are decided by evaluations.

The super gm games, however are decided by "superior pruning", when accounting for why Carlsen and Kasparov so readily beats very strong gms.
Parent - - By Labyrinth (*****) Date 2019-04-05 23:38

>Will a move that is pruned out of the search ever be digged up again with enough time or is it forever discounted by the engine?


I suspect the latter, but I could be wrong. Perhaps it will only revisit an early discarded move if the other moves lead to fails at some depth, but this could require untenably deep calculations so wouldn't necessarily happen in practice.

>You might not know this but in human chess, superior pruning also accounts for why Kasparov, Carlsen etc beats the other top guys.


I don't think that you can really 'graft' computer calculation mechanisms onto humans so easily. Intuition and understanding based on pattern recognition is quite a different animal. Humans don't need to think 'linearly', they can imagine dream positions or piece placements that are not possible in the current position, then work towards those placements.

It would be pruning if the top players mostly calculated the same amount linearly and evaluated similarly, or that the same conclusions could only be reached via the same mental mechanism, but this simply isn't the case.
Parent - By rocket (****) Date 2019-04-06 12:18
<Intuition and understanding based on pattern recognition is quite a different animal>

Everyone has that at super gm level. That's not what separates the very best.
Parent - By rocket (****) Date 2019-04-06 12:31

>Humans don't need to think 'linearly', they can imagine dream positions or piece placements that are not possible in the current position, then work towards those placements.<


The method you are referring to is not always applicable. It's mostly in positional advantageous circumstances. There are times for brute force measures in which Carlsen, Kasparov, Kramnik definately sorts through positions superior to the middle of the road gms, combined with their positional understandings. The positional understanding allows them to sometimes enter muddy waters, when it's too difficult to even "prune it", and expect to end up in a favourable position.

and sometimes of course they simply miscalculate one of the lines. The engines equivalence is discarding a winning line.
- By rocket (****) Date 2019-03-17 12:52
Alright, so what enabled superior pruning compared to in 2006?
Up Topic The Rybka Lounge / Computer Chess / What software enable phone komodo to beat quad shredder 10

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill