Engine-assisted game vs Stockfish, move 5 discussion

[MooMoo2]

The most popular move on the move 4 discussion thread was ...Nf6. Stockfish played 5. Nc3, and the FEN is:
rnbqkb1r/pp1p1ppp/4pn2/8/3NP3/2N5/PPP2PPP/R1BQKB1R b KQkq - 2 5

Game so far: 1.e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3

Depth: 46 ply. Nodes searched: 804,034,774k (~804 G)

[WMHalsdorf]

5.Nc6

[LaurV]

The "acceptable" continuations from all the DBs I could get my handsmouse on, are Nc6, Bb4, d6, a6, Qb6. I don't see much benefit in taking the queen out so early, so I think we have to decide between the others. I like both Nc6 and Bb4, and less the pawn moves. I am still thinking to it. I installed an old version of Rybka (v4.0) which will help me in avoiding obvious mistakes, and give it a run for few hours (still running) although I assume that the game is too early (and the board too full/complex) to have a definitive answer from "pure engine", without the opening books.

[gd_barnes]

a6 and Qb6 are not played often in high level play.

5...Bb4 leads to positions where white wins much more frequently according to the opening databases, i.e. 5...Bb4 6. e5 Nd5 7. Bd2 and white is winning a majority of all games in the DB even when draws are considered.

5...Nc6 6. Ndb5 d6 7. Bf4 e5 8. Bg5 is OK but I think we can do better.

5...Nc6 6. Ndb5 Bb4 7. a3 Bxc3 8. Nc3 is weak for black according to the DB. Also if we want a win we want to avoid swapping pieces so early.

The most played move is d6 and I think it is the best one. The DB seems to confirm that it gives us the best winning chances.

Most played is:
5...d6 6. Be2 Be7 or a6 and black is winning more frequently than he is losing in most variations.

Although played less frequently, the DB seems to show that white gets decent winning chances with:
5...d6 6. g4!

But...the rules state that we can install an older version of SF and use its engine so that is what I did. I feel that doing so gives us reasonable winning chances because we can try out different moves and see how the newest version of SF is likely to respond. Therefore I plugged in 5...d6 and let it think for about 5 mins. Its response was 6. Be3, which is the 3rd most played move in the DB but...according to the DB is not the best response. I also tried putting SF in multi-PV mode, which shows its analysis of its top several moves. It never considered 6. g4 even in its top 5 moves so I think we can rule out that response from it.

So...if we play d6, continuing the idea of using the DB for our moves and using the older version of SF to determine white's moves, the most likely continuation is:
5...d6
6. Be3 a6
7. Be2 Be7
8. O-O O-O

Here black is winning more frequently than it is losing in almost every line according to the DB. Draws were also < 30%. I like our winning chances here. Also SF rates the position as about +.15 for white, which is less of a white advantage than the opening position (about +.25).

Maybe others want to try this motif of letting an engine (preferably an older SF engine) think for a much longer period of time on white's moves white using the DB for our moves. Perhaps it will come up with better moves for white but I seriously doubt it.

I think that the only way that we have a chance to beat SF is to use it against itself to try to find moves where it responds in just a slightly less than optimal way. I do not feel that the slightly newer version of SF that we are playing against will come up with any better moves in the opening than older versions.

[gd_barnes]

Update to the above post. After letting SF think for 10 minutes it came up with:

5...d6 6. f4!

Although 6. f4 appears to only be the 5th most played move in the DB it is still played quite often and it is a good one. One in which white is winning quite a bit more frequently than black in most variations.

It's interesting that it never did consider the 2nd most frequently played move of 6. g4, which also appears good.

So...now I'm perplexed. Maybe LaurV's 5...Nc6 is best.

Edit: After 15 minutes, SF changed back to 6. Bc3. So it's not clear how it would respond.

[axn]

Originally Posted by gd_barnes

Therefore I plugged in 5...d6 and let it think for about 5 mins. Its response was 6. Be3, which is the 3rd most played move in the DB but...according to the DB is not the best response. I also tried putting SF in multi-PV mode, which shows its analysis of its top several moves. It never considered 6. g4 even in its top 5 moves so I think we can rule out that response from it.

This is a dangerous line of reasoning. For the amount of time the moomoo is running it (>10hrs?), it is very possible that SF will find g4 (assuming that it is theoretically the optimal move). Especially since there is a _huge_ ELO increase between the old versions and the latest one.

[gd_barnes]

Originally Posted by axn

This is a dangerous line of reasoning. For the amount of time the moomoo is running it (>10hrs?), it is very possible that SF will find g4 (assuming that it is theoretically the optimal move). Especially since there is a _huge_ ELO increase between the old versions and the latest one.

It found 6. f4 fairly quickly, which I think is nearly as good as 6. g4. See my follow up post. 5...Nc6 may be best. I definitely don't care for 5...Bb4.

[LaurV]

Nc6 seems the best in my opinion. Followed by Bb4 (in order of preference, not in game moves order). I don't know why I don't like the pawn move.

Bb4 seems a lot more aggressive, and avoids exchanging the knights on c6 (which exchange I see it as a path to a draw. Or at least, a small step toward a draw). But we can't do much with the aggressivity there, and exchanging the horse in c3 for our bishop seems really in our disadvantage. In case of Bb4, his e5 forces us to a sequence of moves leading to exchange also.

Also, with Nc6, he still can force us to move the bishop to b4, after his Bf4. This last move is not much frequent in the databases and I don't know why (yet!) because it seems a very good move for white, and it certainly gets the maximum "score" from Rybka 4, after few hours of thinking. (i.e. I am talking about: Nc6, Bf4, Bb4, we have no other good move after his Bf4).

I will look more into d6 in the evening, but for some reason I don't like it.

[gd_barnes]

After looking more at d6 and SF's likely responses to it, I did not like some of the positions that we got into if SF responds with f4 or g4. Therefore I agree that Nc6 is likely our best try.

[LaurV]

Yes, I already voted for Nc6. The problem I see with d6 is that in the same time we block the black-field bishop and we open the king. If after 5...d6 white goes with 6. Bb5+, I don't see any good move for us. Of course Bb5 may not be the best white move either (not really in the books), but it is enough good for me to cut all my interest for 5..d6. What the hack do we play then? 6...Nc6 is out of discussion, we lose big (pawn and position) if white exchanges. Moving the king to e7 lock everything, and queen in front is totally silly. Retreating the horse 6...Nfd7 just sets us two (or three?) moves back, we are crowded there, with no development, and white has all the space in the world, the center of the table, and the sente. The only possible left moves for us are 6...Nbd7, or 6..Bd7. Last is bad, after 7.Qf3, white can push e6 and let us forked (f6 horse, or b7 pawn with the rook together). White is well defended and again it had all the space on the board, the center, and the sente. 7..Qd3 is alo good for white, now we really can not beat that bishop in b5.

This leaves us to 6..Nbd7. Here we are safer and the position seems good for us, we control row 5, but strategically we are too crowded, no space to develop, both bishops are blocked, one horse linked.... We are back 2-3 moves. Not better chances than a draw.

Therefore 5..d6 may be a good move for black, for competition and face to face game, but here we want a win against one of the strongest engines in the world, so we have to take some risks... Otherwise we are wasting the time.

[gd_barnes]

In response to 5...d6, despite initial appearances of a cramped position, black is winning a large majority of the time in most variations in the DB with:
5...d6 6. Bb5+ Bd7 7. Bxd7+ Nxd7 (or Qxd7) [or 7. Qe2 a6]
-or-
5...d6 6. Bb5+ Nbd7 7. O-O (or Bg5) Be7

You mentioned that 5...d6 6. Bb5+ Bd7 7. Qf3 is good for white due to the eventual threat of pushing the pawn with e5. But 7. Qf3 is not in the DB at all because white never gets the chance to push the pawn. After 7...Bxb5 8. Ncxb5 a6 9. Nc3 Nbd7 (or Be7) black is in good shape. Older SF confirmed that this position is nearly dead even or a slight edge for black.

You also mentioned that 5...d6 6. Bb5+ Bd7 7. Qd3 is good for white. But once again 7. Qd3 is not in the DB for higher level play. Perhaps because simply ignoring the queen side and developing normally with 7...Be7 8. O-O O-O results in mostly an even position.

Also with 7. Qd3 or Qf3 black cannot as easily mount a king side because either f4 or g4 are no longer possible. He can play one but not the other and he needs to be able to play both to mount an effective king side attack. Whereas playing the queen to e2 threatens both e5 and an eventual f4 and g4 advance. Therefore 7. Qe2 must be better than 7. Qd3 or Qf3. Black needs to take more immediate action to remove or back off the Bishop on b5, which is why 7...a6 is most commonly played. But even with 7. Qe2 a6 black has nearly equalized.

Therefore I would not be afraid of 6. Bb5+. I would be more afraid of 6. f4 or g4 where white's open position and the ultimate king side attack proves problematic for black. For this reason I also voted 5...Nc6.