lkaufman wrote:"i consider rybka's search a piece of crap. Just mainline checking and never trying to find a better move." I'm not saying you are right or wrong, but Vas would never have used this type of search in Rybka if it did not win against a similar evaluation using a conventional search. Do you have any data to support your apparent belief that Rybka would be even stronger with a conventional search? Has anyone else tried this sort of search with an eval used in an conventional program to see if it helps or hurts?
I believe that the main reason for rybka's strength is the search.
I remember that I tried to change strelka2.0's evaluation(based on rybka1 beta) to piece square table evaluation and I was amazed how strong it is at fast time control inspite of the bad piece square values(it did not had the material tables so it clearly overvalued the pawns) and the only knowledge that it had was to do average between opening and endgame
piece square table based on the stage of the game.
It could beat Joker1.14 at 1+1 time control 24-11 and 5 draws
and Joker has CCRL rating of more than 2400 at 40/40(I used all the possible positions after 1 ply for the test that means (positon after 1.a3
position after 1.a4,...).
It may get worse results at long time control but
I will not be surprised if piece square table version of rybka3 can get CCRL rating above 2500
Uri
this is uninteresting experiment as normal rating is 3100, so you lost 600 and also you played another engine with shitzero knowledge which is a big flaw in the experiment. Further you use great low level code from many many different topprogrammers against an engine from a professor in Amsterdam who did not do much effort to get much speed out of joker. At superfast time controls speed matters and i bet you ran single core. So the experiment is total flawed from every direction.
equip it with crafty eval and play it versus crafty 23.1 i'd say at some more mature time control as it'll be single core. but i bet you won't do that as that is an afternoon of hacking for you to get that done, you prefer to do an afternoon of posts instead of an afternoon of effort for a more serious attempt i guess.
I am not going to work on copying Crafty's evaluation but
as far as I know joker is a fast searcher based on nodes per second and I also used only the 32 bit version of strelka.
It may be interesting to know the result of Doch with the same piece square table that is from board.c of strelka2.0(I see that I did not give all the relevant code because I gave only piece square table and not piece basic values but it is easy to get also the piece values 10,5,3,3,1 that is multiplied by 3399)
lkaufman wrote:"i consider rybka's search a piece of crap. Just mainline checking and never trying to find a better move." I'm not saying you are right or wrong, but Vas would never have used this type of search in Rybka if it did not win against a similar evaluation using a conventional search. Do you have any data to support your apparent belief that Rybka would be even stronger with a conventional search? Has anyone else tried this sort of search with an eval used in an conventional program to see if it helps or hurts?
In November 2008, I did an experiment - replacing the evaluation function in Strelka 2.0 on the evaluation of Rybka 3. In tests on short control (4/40) the results remained the same as that of the Strelka 2.0. At long (40/40) results have grown by about 50-70 Elo.
I went through the Rybka code forwards and backwards and took many things.
lkaufman wrote:"i consider rybka's search a piece of crap. Just mainline checking and never trying to find a better move." I'm not saying you are right or wrong, but Vas would never have used this type of search in Rybka if it did not win against a similar evaluation using a conventional search. Do you have any data to support your apparent belief that Rybka would be even stronger with a conventional search? Has anyone else tried this sort of search with an eval used in an conventional program to see if it helps or hurts?
In November 2008, I did an experiment - replacing the evaluation function in Strelka 2.0 on the evaluation of Rybka 3. In tests on short control (4/40) the results remained the same as that of the Strelka 2.0. At long (40/40) results have grown by about 50-70 Elo.
The gap between rybka3 and rybka1 beta is 238 elo for 64 bits and 207 elo for 32 bits
Based on my memory Strelka2.0 is only slightly stronger than rybka1 beta so you practically got only 50-70 elo out of 200 elo that rybka3 got.
It strongly suggests that without a superior search rybka3 could not even get half of the improvement that she got.
lkaufman wrote:"i consider rybka's search a piece of crap.
I never saw Rybka's sources, but I saw Ippo ones and I don't know if they are similar, but what I can say is that in Ippo evaluation is nothing special, almost dumb, even less sophisticated then in Stockfish, instead search is a master piece.
lkaufman wrote:"i consider rybka's search a piece of crap. Just mainline checking and never trying to find a better move." I'm not saying you are right or wrong, but Vas would never have used this type of search in Rybka if it did not win against a similar evaluation using a conventional search. Do you have any data to support your apparent belief that Rybka would be even stronger with a conventional search? Has anyone else tried this sort of search with an eval used in an conventional program to see if it helps or hurts?
In November 2008, I did an experiment - replacing the evaluation function in Strelka 2.0 on the evaluation of Rybka 3. In tests on short control (4/40) the results remained the same as that of the Strelka 2.0. At long (40/40) results have grown by about 50-70 Elo.
I'm curious. There have been reports, unconfirmed, that the 60-70 Elo advantage seen by Robbo in fast time-controls vs Rybka 3, drops severely at slower time-controls. Do you think the case you described is true of Robbo and others?
I'm running my own tests using the SilverSuite, but the slow TC (40/40+40/40) is still too early to tell, even if it seems to agree with your data so far.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
I also compared the number of nodes in the Rybka 3 and the Strelka 2.0 when searching at the same depth. Naturally, the real number of nodes and the real depth. In Rybka the number of nodes was 5 times less.
If you use the usual formula: 100 * ln(speed),
then the reduction in the number of nodes in 5 times is roughly equivalent to +160 Elo.
I went through the Rybka code forwards and backwards and took many things.
lkaufman wrote:"i consider rybka's search a piece of crap.
I never saw Rybka's sources, but I saw Ippo ones and I don't know if they are similar, but what I can say is that in Ippo evaluation is nothing special, almost dumb, even less sophisticated then in Stockfish, instead search is a master piece.
I think you have to add that this was a _quotation_ of Vincent Diepeeven, but not Larry Kaufman's opinion.
lkaufman wrote:"i consider rybka's search a piece of crap.
I never saw Rybka's sources, but I saw Ippo ones and I don't know if they are similar, but what I can say is that in Ippo evaluation is nothing special, almost dumb, even less sophisticated then in Stockfish, instead search is a master piece.
I think you have to add that this was a _quotation_ of Vincent Diepeeven, but not Larry Kaufman's opinion.
You made my day, Luca! Plus the thought of Vas' face if he ever had read this thread too fast and came upon this incorrect quotation from Vincent/Larry.
lkaufman wrote:"i consider rybka's search a piece of crap.
I never saw Rybka's sources, but I saw Ippo ones and I don't know if they are similar, but what I can say is that in Ippo evaluation is nothing special, almost dumb, even less sophisticated then in Stockfish, instead search is a master piece.
I think you have to add that this was a _quotation_ of Vincent Diepeeven, but not Larry Kaufman's opinion.
You made my day, Luca! Plus the thought of Vas' face if he ever had read this thread too fast and came upon this incorrect quotation from Vincent/Larry.
I mistaken also myself !!!!
Actually it was very strange that LK could talk like that of something that gave him the bread for some time. That's the reason I was "shocked"
lkaufman wrote:"i consider rybka's search a piece of crap.
I never saw Rybka's sources, but I saw Ippo ones and I don't know if they are similar, but what I can say is that in Ippo evaluation is nothing special, almost dumb, even less sophisticated then in Stockfish, instead search is a master piece.
I think you have to add that this was a _quotation_ of Vincent Diepeeven, but not Larry Kaufman's opinion.
You made my day, Luca! Plus the thought of Vas' face if he ever had read this thread too fast and came upon this incorrect quotation from Vincent/Larry.
I mistaken also myself !!!!
Actually it was very strange that LK could talk like that of something that gave him the bread for some time. That's the reason I was "shocked"
LK does not talk like that regardless of the facts.
It is clearly not his style to say about a search of some program that it is a piece of crap.