Is rybka ply depth display reliable?

[emerson4301972]

Some said that Rybka's node count is not real or Rybka count it differently compared to other programs.

Should we treat Rybka's ply depth the same as other programs or is it different from other program just like its node count?

[mclane]

you should not rely on rybkas output.

[Eelco de Groot]

You can add two to reported plydepths, that gives a better comparison to other programs, if you want to compare tactical accuracy or speed to plydepths between Rybka and other programs.

For instance Christophe Théron said, it is quite a while ago, that subtracting two from the plydepths of ChessTiger made searches comparable to Rybka 1.0 Beta, the programs then were finding tactical solutions at the same ply.

For Strelka clones the plydepth is I believe two plies more than in Rybka 1.0 Beta. For later versions of Rybka I think we can safely assume that

• The search has changed.
• Adding two to the reported plydepths is still more useful in comparing accuracy at least it should be possible to verify this directly.

Eelco

[kgburcham]

[d] r1q2rk1/pp1n1p1p/1n1P2p1/2b1P1B1/4NQ2/1B3P2/PP2K2P/R7 w - -

Time to solve:

Zappa Mexico II x64 0:24
Deep Hiarcs Paderborn 2007 0:41
Deep Shredder 11 x64 3:43
Glaurung 2.0.1 mp x64 15:32
Rybka 2.3.2a mp x64 18:21
Arasan 10.4 23:49

Depth to Solve

Zappa Mexico II x64 14/47
Deep Hiarcs Paderborn 15/42
Deep Shredder 11 x64 15/50
Glaurung 2.0.1 mp x64 20.03
Rybka 2.3.2a mp x64 22.34
Arasan 10.4 17.01

Kns:

Zappa Mexico II x64 1992
Deep Hiarcs Paderborn 1460
Deep Shredder 11 x64 3182
Glaurung 2.0.1 mp x64 5174
Rybka 2.3.2a mp x64 384
Arasan 10.4 1337

Total Nodes to Solve:

Zappa Mexico II x64 (49.420.211)
Deep Hiarcs Paderborn (60.567.000)
Deep Shredder 11 x64 (709.818.700)
Glaurung 2.0.1 mp x64 (4.823.808.486)
Rybka 2.3.2a mp x64 (413.799.765)
Arasan 10.4 (1.911.770.904)





Zappa Mexico II x64
14/47 0:24 +2.62 1.Rc1 Qc6 2.Rxc5 Nxc5 3.Nf6+ Kh8 4.Qh4 Qb5+ 5.Ke3 h5 6.Nxh5 Nd5+ 7.Bxd5 Qd3+ 8.Kf2 Qc2+ 9.Kg3 Rg8 10.Nf6+ Kg7 (49.420.211) 1992

Deep Hiarcs Paderborn 2007
1.Rc1 (0.87) Depth: 15/42 00:00:41 60567kN

Deep Shredder 11 x64
15/50 3:43 +4.07 1.Rc1 Qc6 2.Rxc5 Qxc5 3.Nxc5 Nxc5 4.Bh6 Nxb3 5.axb3 Nd7 6.Bxf8 Rxf8 7.Qe4 b6 8.Ke3 a5 9.f4 Nc5 10.Qd4 (709.818.700) 3182

Glaurung 2.0.1 mp x64
20.03 15:32 +4.54 1.Rc1 Qc6 2.Rxc5 Qxc5 3.Nxc5 Nxc5 4.Bh6 Nxb3 5.Bxf8 Rxf8 6.axb3 Nd7 7.Qe4 b6 8.f4 Nc5 9.Qd5 Re8 10.d7 Rd8 (4.823.808.486) 5174

Rybka 2.3.2a mp x64
22.34 18:21 +2.59 1.Rc1 Qc6 2.Rxc5 Nxc5 3.Nf6+ Kh8 4.Qh4 Qb5+ 5.Ke3 Qd3+ 6.Kf2 h5 7.Nxh5 Ne4+ (413.799.765) 384

Arasan 10.4
17.01 23:49 +1.15 1.Rc1 Qc6 (1.911.770.904) 1337

[Ovyron]

Rybka has been confirmed to subtract 3 to depth, so ply 1 is actually ply 4.

You can check this by forcing Rybka to negative depths, and she'll go to up to depth -3 (playing weaker.)

[tiger]

It was 3 plies.

[Eelco de Groot]

It was 3 plies.

Christophe! Very good! You were right from the start about that. And it is very good to see you, I hope you and your family are doing well, I am glad you have not completely forgotten us here or this silly endeavour called computerchess. But sometimes far more important things in life demand some attention too!

My Regards,

Eelco


Holding our breath for any news on underwater Tigers

[mclane]

that looks like tiger catching little fish under water...

[Matthias Gemuh]

that looks like tiger catching little fish under water...

I tought it was Tiger hiding from BigLion

Matthias.