What does your Rybka play here?

[Ovyron]

On my computer, my Rybka 1CPU 32bit 64MB always wants to play 2. Nf3!? allowing 2...d4! I realized that the bad performance of Rybka in my book-less tourneys is mainly due to this.

However, another member on Rybka forum can't reproduce Nf3! He tried on different computers and his engine only plays d4. I tried reinstalling, different GUIs, and other things but Rybka is stuck on Nf3. The only way to make it play another move is to change the outlook, but I think it could be a problem with my hardware (Athlon 2.1Ghz.)

[d]rnbqkbnr/ppp1pppp/8/3p4/8/2N5/PPPPPPPP/R1BQKBNR w KQkq - 0 1

I'd like to know if someone else has Rybka loving Nf3 so much, or if it's only me.

[swami]

Well, I'd guess it depends on the version that is used....you haven't specified the version number.

Also, It may well out put different move other than Nf3 as the depth increases, ask the guy at what depth Rybka shows d4 or did it ever show Nf3 etc

[Mike S.]

I can reproduce 2.Nf3 with single Rybka 2.3.2a 32-bit, default engine settings, and with 64 MB hash on a Pentium D945 3.4 GHz:

Analysis by Rybka 2.3.2a 32-bit :

2.Nf3 Nf6 3.d4 Nc6 4.g3 g6 5.Bg2 Bg7 6.0-0 0-0 7.Ne5 Bf5 8.Bf4 Nb4
= (0.19) Depth: 5 00:00:00
= (0.09) Depth: 17 00:02:24 7809kN

I replaced Rybka with another engine, and loaded it again but with 256 MB hash (resulting only in a small speedup):

Analysis by Rybka 2.3.2a 32-bit :

2.Nf3 Nf6 3.d4 Nc6 4.g3 g6 5.Bg2 Bg7 6.0-0 0-0 7.Ne5 Bf5 8.Bf4 a5
= (0.19) Depth: 5 00:00:00
= (0.09) Depth: 17 00:01:54 6184kN

Now, I switched to multi-pv mode(3) and waited until depth 16 was finished again.

Analysis by Rybka 2.3.2a 32-bit :

1. = (0.09): 2.Nf3 Nf6
2. = (0.09): 2.d4 Nf6 3.Bf4 g6 4.g3 Bg7 5.Bg2 0-0 6.Nf3 Bf5 7.0-0 Nc6 8.Ne5 a5
3. = (-0.02): 2.e3 Nc6 3.Nf3 Nf6 4.d4 g6 5.Bd3 Bg7 6.0-0 0-0 7.a3 Be6 8.Bd2

I think that it is possible that mp Rybkas cannot always reproduce Nf3, because d4 gets the same eval and they will sometimes get d4. Although, I am not sure if the evals are exactly identical, or have more digits internally.

I think Rybka, but also any other engine, should have opening book or opening variation support, at least for some moves. Also, if you provide some solid variations which are really beyond doubt, it will help your testing because it will create variety and allow more games which make sense. More than two games between the same opponents and under the same conditions (x-y and y-x) without any book are nonsense.

Here is a PGN with four short variations, which can be used as opening database, for eight games per pairing, with switching sides (also called switch colors). Two end with a white move and two end with a black move. You can also be sure that no engine is favoured, by these very short move sequences.

Copy the PGN, paste it into a text editor and save with .pgn extension. Then it is ready to use, in Fritz or Arena (and probably also in other interfaces).

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]
[ECO "E15"]
[PlyCount "7"]

1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. g3 *

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]
[ECO "B54"]
[PlyCount "7"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 *

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]
[ECO "E61"]
[PlyCount "8"]

1. Nf3 Nf6 2. d4 g6 3. c4 Bg7 4. Nc3 O-O *

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "*"]
[ECO "C77"]
[PlyCount "8"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 *

[Ovyron]

I can reproduce 2.Nf3 with single Rybka 2.3.2a 32-bit, default engine settings, and with 64 MB hash on a Pentium D945 3.4 GHz:

Analysis by Rybka 2.3.2a 32-bit :

2.Nf3 Nf6 3.d4 Nc6 4.g3 g6 5.Bg2 Bg7 6.0-0 0-0 7.Ne5 Bf5 8.Bf4 Nb4
= (0.19) Depth: 5 00:00:00
= (0.09) Depth: 17 00:02:24 7809kN

I replaced Rybka with another engine, and loaded it again but with 256 MB hash (resulting only in a small speedup):

Analysis by Rybka 2.3.2a 32-bit :

2.Nf3 Nf6 3.d4 Nc6 4.g3 g6 5.Bg2 Bg7 6.0-0 0-0 7.Ne5 Bf5 8.Bf4 a5
= (0.19) Depth: 5 00:00:00
= (0.09) Depth: 17 00:01:54 6184kN

Thank you! So this means there's nothing wrong with my setup, or with my Rybka, or with my GUI, or with my hardware.

As for the book-less games, I only play 1 with no book, then on later games I force one move depending on the results. For example, if Rybka played Nf3 and won, on the second game it repeats it and I force a different reply for black (so it tries to improve.) If Rybka lost, I force a different first move (for example e4) and so on.

In case of a draw, I may keep the first moves on the game and force a different move later (for white as it should try to make use of moving first) because I don't think the engines should play very differently if they don't want to play too risky.

Really, if some move is good against the opponent I don't see why should we force the engine to play another one, so I call this "Manual book learning" and it's specific against an opponent (Rybka has a different memory of games against opponent A than against opponent B so I force different moves.)

Thanks for the PGN, it'll be useful for choosing moves to force.