TalkChess.com

lkaufman wrote:I only use one core for these tests as I'm looking at so many positions and each core works on separate positions. There's no interest in testing MP as it's just a speedup and roughly similar among all programs. Regarding Firebird and Robbo, if Firebird is on one core is it sufficiently different from Robbo to be worth testing separately, or is it primarily just an MP implementation of Robbo?

The MP implementation of Robbo is Igorrit....IvanHoe and FireBird are derivatives from the Robbo family....and what a family that is
Dr.D

Do those three programs generally play the same move with the same score on one core at a set depth, or do they differ substantially and frequently at the same depth?

lkaufman wrote:Do those three programs generally play the same move with the same score on one core at a set depth, or do they differ substantially and frequently at the same depth?

In general they will not play the same move with the same score at the same depth,at least these are my observations roughly as I didn't test them....
Dr.D

lkaufman wrote:Do those three programs generally play the same move with the same score on one core at a set depth, or do they differ substantially and frequently at the same depth?

Mr Kaufman, Why dont you respond on reasons for these big differences in analysis moves in the given game Anand - Kasparov between Rybka and Firebird. How come Firebird is so different here and play far more GOOD Kasparov moves?

I really would like to know your opinions on this

Bram

I haven't even looked at Firebird as I thought it was too similar to Robbolito to be worth checking out. If you can, run the same moves thru Robbo to see if it acts like Firebird or like Rybka. If it is a lot different from Firebird on this game, I'll check out Firebird and let you know what I find out.

Tomcass wrote:Firebird Test 32w

Today I see no engine stronger than FireBird 1 in the world. You are great, guys!!

Tom.

Have you tried Robbolito also?
My tests show that Robbolito 0.085g3 is still the strongest on single core.
IMHO, to decide which engine from Ippo family is the strongest, you need to involve other strong engines into the tournament also,
not only Rybka & engines from Ippo-Robbo family...

Well it is very similar to Robbolito.
In this game from move 15 to 32 for black (Kasparov), at analysis time 30 sec a move, Robbolito only finds one less Kasparov move than Firebird, namely only on move 28

But still why does these Ippolit-based programs in this analysis test, gave so much different non-Rybka moves, as you and other ppl supposing they are clones or derivatives ??

I counted a difference of 8 on 18 blackmoves in the given game example and surprisingly Firebird (at 30 sec a move) plays 15 out of 18 Kasparov moves and Rybka only 7 ??


...28...Rb8 [Rybka 3 32-bit: 28...Bf6 29.Bf2 Qa3 30.Ng1 Qa4 31.Qc2 Qd7 32.Rfe1 Be5 33.Rcd1 a5 34.c5 dxc5 35.Qc4 Bd6 36.Re4 -0.70/15 ; FireBird 1.0 w32: 28...Rb8 ! (Kasparov gamemove) 29.Rc2 Qc8 30.Ng1 Qf5 31.Rfc1 Be5 32.Re1 Rec8 33.Bg5 Rb7 34.Kg2 Bf4 35.Bxf4 Nxf4+ 36.Kh2 Nxd5 37.Re4 Nf6 38.Rd4 d5 39.Rf4 Qe5 40.Qd4 Nd7 41.c5 -0.96/22 ; RobboLito 0.085g3 w32: 28...Bh6 ! different from Firebird 29.Qxh6 Rxe2 30.Rfe1 Rce8 31.Rxe2 Rxe2 32.Rb1 Re8 33.Qd2 Qxc4 34.Bf2 Qc8 35.Kg2 Re2 36.Qd4 Rxa2 37.h4 Kh7 38.Qa7 Qf5 39.Rb7 Kg7 40.Qd4+ Nf6 41.Rb6 Qxd5 42.Rxd6 Qxd4 43.Rxd4 -0.97/20 ...

Robbo (and I presume also Firebird) is much faster than Rybka 3 but with lower quality, so it will often see a ply deeper but with less positional knowledge and missing some tactical things as well. Therefore in a given amount of time the move choice is likely to be different. Apparently in this particular game the extra ply proved far more relevant than the missing positional stuff, but over the thousands of games in my database the two perform about the same at predicting moves at one minute per move. This also means that in assessing which is stronger, Rybka 3 or a derivative, it is important to compare results against a third unrelated engine (like stockfish for example), as in a direct matchup speed trumps knowledge in general.

lkaufman wrote:This also means that in assessing which is stronger, Rybka 3 or a derivative, it is important to compare results against a third unrelated engine (like stockfish for example), as in a direct matchup speed trumps knowledge in general.

Completely agreed! When developing Stockfish we have noted that many times +50elo against former version of program, often means +20 elo against pool of engines. Remembering that Robbo possibly has been specially tuned to beat Rybka, I have doubts if it is at all stronger against pool of engines.

We (meaning both while working on Rybka and now on Komodo) have generally found that improved eval works pretty well against any opponent, and speedups work against unrelated opponents but to a lesser extent. However when speed is achieved at the expense of knowledge, it is likely that the faster version will test well against its "brother" but not well against non-relatives. I'd like to see someone run a long non-blitz match of Rybka vs. Stockfish 1.63 and one of the derivatives vs. Stockfish 1.63. I don't know which program I would bet on.