The final position:

click for larger view

1. ² (0.60) 19...Ng4 20.Nd4 Bf8 21.Qf4 Ne5 22.Nxc6 bxc6

2. ± (0.66) 19...Nh5 20.Rad1 Bf6 21.Bd4 Be7 22.Rc1 Rac8

3. ± (0.78) 19...Rac8 20.Nd4 d5 21.Rae1 Nh5 22.Re3 Bf6

4. ± (0.85) 19...Nd5 20.Bd4 gxf5 21.Rae1 f6 22.Nxd5 Bxd5

For example, 18.Qd2 Ng4 19. Bb6 Bf6 20. h3 Ne5 21. Nd4 Qe7 22.Rae1! Bg5 23.Nxc6 bxc6 24.Qe2 Bh4 25.f6! check out how complex this position looks:

click for larger view

This Deep Junior 10 analysis has me puzzled as well as extremely worried.

25...Qxd3 ????Is that a typo? (25....Qh5 surely).
And then after 25...Qxd3 26.f6 is <#4 (mate in four), not 0.59>

I don't know what to think of analysis like that.

And of course the line was already faulty with 24....Qe2, where 24...Rc8 or possibly Bb3 would be what Black would play.

That is the result of 22 ply Deep Junior 10???

I can hardly believe it.
Test it again at the crucial moves, will you?
BTW, you must have a very high bus speed to get to 22 play in seven minutes.

25...Qxd3 ????Is that a typo? (25....Qh5 surely).
And then after 25...Qxd3 26.f6 is <#4 (mate in four), not 0.59>>

I've seen this behavior in Deep Junior 10 before, the first several moves of a line would be accurate, but later moves become more and more inaccurate. Rest assured though, when you actually have it calculate for the 25th move, it chooses Qh5 instead of Qxd3.

<BTW, you must have a very high bus speed to get to 22 play in seven minutes.>

Athlon X2 3800+ overclocked to 2500mhz, basically slightly faster than an Athlon X2 4400+.

Does it really really work?

I am cautiously optimistic.

I gratefully followed RV's Rybka until move 22....? where I thought it was a little too optimistic. I reckon I found a better move for Black there.

The line seems utterly amazing, <but it is ESSENTIAL that we play Bb6 at move 19 in the sequence.>

Variations on my forum, but please post your reply here or on the main page because I want to keep the pgn on top at my own forum.

Put it in your engine and see if I'm right. Almost a pawn up by the end of it.

Even 18...Nd5 seems to hold no problem for us. It will be a different game then, but not impossible.

[Event "seems good "]
[Site "?"]
[Date "2006.??.??"]
[Round "?"]
[White "our"]
[Black "game"]
[Result "*"]
[ECO "B42"]
[Annotator "monad"]
[PlyCount "53"]
[EventDate "2005.??.??"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Bd3 Bc5 6. Nb3 Be7 7. Qg4 g6 8. Qe2 d6 9. O-O Nd7 10. a4 Ne5 11. a5 Nf6 12. Be3 O-O 13. Nc3 Bd7 14. f4 Nxd3 15. cxd3 Bc6 16. f5 exf5 17. exf5 Qd7 18. Qd2 Rfe8 19. Bb6 Nd5 20. Bd4 gxf5 21. Qh6 f6 22. Qh3 Bf8 23. Rxf5 Ne3 24. Rh5 Qf7 25. Rh4 Qxb3 26. Ne4 Bxe4 27. dxe4 *

I didn't get back convinced, I think the problem with that site is that evaluations are missing at the end of the variations (I don't demand 1/100 pawn precision, usual symbols would be perfect) therefore the choice of what is the main line and what is a subvariation appears arbitrary.

Also somewhere you end a variation in a non "quiescent" position, that's the case with 19.Bg5 Ng4 20.Nd4 Bxg5 21.Qxg5 Ne3 22.f6. I spent some time to find 22...Qg4, without that move some people may think White is actually winning there!

Hiya, the reason I left it out is that i<t is all > right up to the end. There is absolutely no problem with the entire line. Actually, even better with the addition of the Thorsson bit towards the end. It is <the better part of a pawn> up all the way right to the end.

Besides, you have seen the link to the pgn at the bottom of the webpages, I hope? I have already made it so easy for you.

You can put it in any engine and see the values for yourself. (and do some checking while youre at it )

Anyways, here is another pgn file for you, with some values added.

[Event "value added"]
[Site "?"]
[Date "2006.??.??"]
[Round "?"]
[White "our"]
[Black "game"]
[Result "1-0"]
[ECO "B42"]
[Annotator "JC"]
[PlyCount "72"]
[EventDate "2005.??.??"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Bd3 Bc5 6. Nb3 Be7 7. Qg4 g6 8. Qe2 d6 9. O-O Nd7 10. a4 Ne5 11. a5 Nf6 12. Be3 O-O 13. Nc3 Bd7 14. f4 Nxd3 15. cxd3 Bc6 16. f5 exf5 17. exf5 Qd7 18. Qd2 Rfe8 (18... Nd5 eval. 1.21 --- Sightly better for ♗lack than ♘g4, but still too costly for him I would have thought. Don't forget, we are working on narrow margins here.) (18... Ng4 19. Bb6 Bf6 20. h3 Ne5 21. Nd4 Rae8 (21... g5 22. Nxc6 bxc6) 22. Rae1 Kh8 eval 1.64 So too high a price for ♗lack to deviate from ♖fe8. If he does play this, we can cope.) 19. Bb6 (19. Bg5 Ng4
This is the move that makes 19.♗g5 a mistake. (19... Nh5 ♙ie in the sky. ♘ot going to happen. And....♘g4 is dangerous.) 20. Nd4 Bxg5 21. Qxg5 Ne3 22. f6 no more than 0.09 for us.) 19... Nd5 20. Bd4 gxf5 (20... f6 21. fxg6 hxg6 22. Ne4 Qg4 23. Rae1) 21. Qh6 f6 22. Qh3 Bf8 (22... Bd8 Too optimistic from ♖V 22 ply ♖ybka 23. Rxf5 Ne3 24. Bxe3 Rxe3 25. Qxe3 Qxf5 26. Nd4 Qe5 27. Qf2 Qc5) 23. Rxf5 Ne3 24. Rh5 Qxh3 This is where Thorsson came in with an improved line taken from his Junior engine. I agree, it is better than the line I had at first, on the previous webpage stage 21. 25. Rxh3 Nc2 26. Rf1 Nxd4 27. Nxd4 d5 28. Rxf6 Bc5 29. Rf4 Rf8 30. Rg3+ Kh8 31. Rgf3 Rxf4 32. Rxf4 Rf8 33. Rxf8+ Bxf8 34. Nxc6 bxc6 35. Kf2 Bb4 36. Ke3 Kg7 0.80ish

It's stunning how much raw material there is. I am not going to sort it now, but leave that to the analysts. Here's what I thought were the most important lines discussed. Please feel free to bring other lines up if you think I've missed some. For now, if GMAN does play Ng4, I think the question of Bb6 or Bd4 is most important to work on.

<Tabanus>: Analysis (18-ply) by Fritz8 after 18.Qd2 Rfe8 19.Bb6 (bee bee six): 1. ² (0.60) 19...Ng4 20.Nd4 Bf8 21.Qf4 Ne5 22.Nxc6 bxc6

<Domdaniel> says:

After 18.Qd2 Ng4 19.Bd4 gxf5 20.Qf4 Ne5 21.Bxe5 (21.Rae1 is also fine for White) dxe5 22.Qg3+ Kh8 23.Qxe5+ f6 24.Qxf5 Qxf5 25.Rxf5 an interesting position arises, where Black has 2 bishops but White has an extra pawn.

Is this a case of Black sacking a pawn for initiative? Or White winning a pawn, confident that the knights can contain the bishops?

<kwgurge> says: If 18.Qd2 Ng4 19. Bd4 and after most any reasonable black move 20.h3 and h6 is available to our Queen. White's pressure on f6 and along th a1-h8 diagonal is very strong.

<Rookfile> says: Right, move 22 seems to be the place to look for improvements. I was wondering about

18. Qd2 Ng4 19. Bb6 Rfe8 20. h3 Ne5 21. f6 Bf8 22. Be3 Rec8 23. Bh6 Qd8 24. Nd4 Rab8.

<Karpova> in reference to a Rybka line:

2. = (0.23): 18...Ng4 19.Bb6 Rfe8 20.Nd4 Bf8 21.Qf4 Ne5 22.Nxc6 bxc6 23.d4 Nd3 24.Qd2 Nb4 25.Rf3>

<Thorsson> came up with a better move for white. Here's what it could look like: 20.h3 (instead of Nd4) Ne5 21.f6 Bf8 22.Be3 Qe6 23.Bh6 (Nd4 is met by 23...Nc4) Nd7 24.Bf8: Rf8: 25.Nd4 Qe5 26.Nc6: bc 27.d4 Qh5 28.Qf4

<Thorsson> responds to <Eaglewing> : <eaglewing: <FHBradley> I asked: <What if 18. Qd2 Ng4 19. Bb6 h6?> You mentioned a quiet 20. Qe2. <20 Qe2 h5 21 h3 Nf6 Nd4>. I add, as Black I would consider 20. ... Nh6. And in any case with Qe2 those high expectations of <Qd2> could be faded away.>

I did post about this line twice already. But in the line given above, 18.Qd2 Ng4 19.Bb6 h6 20.Qe2 h5 21.h3 Nf6 then 22.Nd4 looks like it's winning.

20...Nf6 straight away is better.

<Naruddin Hodja> says:

18. ... Ng4 19. Bd4 (best, I think) Bf6 20. Ne4 Bxd4 21. Nxd4 Bxe4 (or else 22. f6 follows) 22. dxe4 Nf6 23. Qf4 and white retains a space edge and will soon start pressuring black's weak d6 pawn.

<Thorsson> responds to <Rookfile> : <RookFile: Well, I suppose that black might try 18. Qd2 Ng4 19. Bb6 Rfe8 20. h3 Ne5 21. f6 Bf8 22. Be3 Rec8 23. Bh6 Qd8 24. Nd4 Rab8>

21.Rae1 is another possibility. They key is kicking the Ng4 straight away.

<Thorsson> responds to <Rookfile> : <Rookfile: E) 18. Qd2 Ng4 19. Bb6 Rfe8. I did not see this logical try addressed in Karpova's post, that I was given to look at tonight. Let's say you continue 20. Nd4. All of a sudden, 20....Bf8 sets up a threat of ....Ne3. But, white attacks too. 21. Qf4! (This prevents ...Ne3 because of 22. Nxc6! the b6 bishop hits the e3 knight) Ne5 Hmm. Does white have anything here? Not really - he can kick the knight, but after 22. Nxc6 bxc6 23. d4 Nc4, I think black is close to equal. Fooling around with various tries for ten minutes with Fritz seems to confirm this. So, an improvment evidently needs to be found to 20. Nd4 in this line.>

20.h3 Ne5 21.f6 Bf8 22.Be3 and Black looks in a wee bit of trouble.

<Dionyseus> says: I think I found an improvement, instead of 22.Nxc6, white should play Rae1!

For example, 18.Qd2 Ng4 19. Bb6 Bf6 20. h3 Ne5 21. Nd4 Qe7 22.Rae1! Bg5 23.Nxc6 bxc6 24.Qe2 Bh4 25.f6!

Both Rybka and Deep Junior 10 love this position, Rybka evals this as 1 pawn up, DJ10 says almost 2 pawns up.

<Rookfile> discusses some various Ng4 lines:

B) 18.... Ng4 19. Bb6 Bf6 20. h3 Ne5 21. Nd4 Qe7 22. Nxc6 bxc6 23. Ne4 c5 24. fxg6 fxg6 25. b4 Nd7 +0.39.

What to make of this? 20. h3 does not impress me, white spends a tempo to make black play a move he's going to play anyway. That's probably why this comes out as only 0.39.

Interesting is the move 20. Bd4! No, this is not an offer of a draw by 20.... Be7 21. Bb6 etc. because you would deviate from 21.Bb6 with 21. f6!! winning on the spot, because black's knight has nowhere to go. So, black sticks his knight in the way with 20.....Ne5 instead, and now comes 21. Ne4! Bh8. So this is a hard position to evaluate. I really don't like the bishop on d4, because it gets in the way of the knight - but white does seem to have some advantage here because of the wayward bishop on h8. Something like 22. Bc3 seems to address this concern and promise a slight edge for white.

C) 18. Qd2 Ng4 19.Bg5 - this transposed into a known position from my 18. Bg5 research - it's pretty good for white.

D) 18. Qd2 Ng4 19. Bb6 Rae8 20. h3 +0.54

Evaluation: 19...Rae8 is logical, because it attempts to demonstrate that the field of operations is on the kingside now, and black has numerical superiority there. How does it work out in terms of variations, though? 20. h3 Nf6 21. Be3 looks real good for white. So, I guess, 19....Rae8 needs to be ruled out, it boxes in the f8 rook and sets up a scenario by which black can lose the exchange.

E) 18. Qd2 Ng4 19. Bb6 Rfe8. I did not see this logical try addressed in Karpova's post, that I was given to look at tonight. Let's say you continue 20. Nd4. All of a sudden, 20....Bf8 sets up a threat of ....Ne3. But, white attacks too. 21. Qf4! (This prevents ...Ne3 because of 22. Nxc6! the b6 bishop hits the e3 knight) Ne5 Hmm. Does white have anything here? Not really - he can kick the knight, but after 22. Nxc6 bxc6 23. d4 Nc4, I think black is close to equal. Fooling around with various tries for ten minutes with Fritz seems to confirm this. So, an improvment evidently needs to be found to 20. Nd4 in this line.

<pilot> says: Has someone already analyzed this line further?

18.Qd2 Ng4 19.Bd4 Bf6

It seems to do a good job killing our kingside attack and the bishop theatens check. All I could see was:

20.Bxf6 Nxf6 and we lose our attacking bishop to black's bad bishop

<Ron> says: I have now switched my vote to Qd2. Here is a sample line from this position: 18. Qd2 Ng4 19. Bg5(!?) Re8 20 h3 Ne5 21 f6 Bf8 22 Bh6 Kh8 23. Rae1 b5 24. Bxf8 Rxf8 25. d4 Qa7

<Willem Wallekers> says: Everybody is talking about 18. Qd2 Rfe8 and 18. Qd2 Ng4 19. Bd4 Rae8 or 18. Qd2 Ng4 19. Bb6 Nd5, but what about 18. Qd2 Ng4 19. Bd4 gxf5 and 18. Qd2 Ng4 19. Bb6 Bf6?? Or 18. Qd2 Ng4 19. Bg5 f6? I need evidence on these three variations with 18. Qd2 Ng4 before I switch from Nd4 to Qd2.

<Wassily> discusses some Ng4 lines:

18. Qd2 Ng4 19. Bd4 Rae8 20. Ne4 Bd5 21. Qf4

18. Qd2 Ng4 19. Bd4 Rae8 20. fxg6 fxg6 21. Rae1

18. Qd2 Ng4 19. Bd4 Rae8 20. h3!? Nf6 21. Qh6

<Thorsson> says: OK if I look atth elines posted on 18.Qd2 I see that Rybka and Fritz want to answer with the natural 18...Rfe8, while Shredder has 18...Ng4. As Fritz also likes Ng4 the following move we should examine Ng4 before getting to the main line.

I too had looked at this line quickly and liked 19.Bb6 in response. I now have two Black responses:

A. 19...h6 20.Qe2 Nf6 21.Qe3

B. 19...Bf6 20.h3 Ne5 21.Nd4 Rae8 22.Rae1

I don't honestly think this can be good for Black.....

<Azaris> says: <If 18...Ng4 then as <Dionysius> has pointed out 19. Bb6 preserves whites advantage and permits a later return to the a1-h8 diagonal>

This is actually a good point. Say 18.Qd2 Ng4 19.Bb6 Bf6 20.h3 Ne5 21.Nd4 and now something like 21...d5?? fails miserably to 22.Bc5 .

<Karpova> says: What about: 18.Qd2 Ng4 19.Bg5 f6 (19...Bg5: Qg5: allows f6 and loses) 20.Bh6 Rfe8 21.fg hg 22.h3 Qe6 24.Nd4 Qe5 25.Qf4 Nh6: 26.Nc6: Qf4: 27.Ne7:+ Re7: 28.Rf4: and my calculator gives 0.35, depth 13.

<Rookfile> says:

Does black get to play moves too?
Something like 18. Qd2 Ng4 19. Bd4 gxf5 20. Qf4 Ne5 21. Rae1 Bf6

<Dionyseus> says: <18. Qd2 Ng4 19.Bd4 gxf5 20. Rf4 Ne5 21.Raf1 Kg8 22.Rxf5 Rg8 23.Ne4 and black is lost.>

No no no, that's bad. 20.Rf4?? allows Qe6!

20.Rf4?? Qe6! 21.Re1 Ne5! 22.Bxe5 dxe5 23.Nd4 Bc5 and black is winning:

Analysis by Rybka 2.1o mp 32-bit:

24.Kf1 Bxd4 25.Rxd4 f6 26.Qf2 Rad8 27.Rf4 Rxd3 28.Rxf5 Qd6 29.Qe2 Rd2 30.Qc4+ Kg7 µ (-0.85) Depth: 16 00:00:19 4477kN

Forget about 19.Bd4, Bb6 must be played instead!

<Ughaibu> says: What's wrong with 18.Qd2 Ng4 19.Bd4 Bf6 20.Bf6 Nf6 21.Qg5?

<kwgurge> says: Ng4 after Qd2 has been shown to be very good for white. After Bd4 and h3, the N is either lost, retreats to f6 or goes to e5 at which point the f6 push you are so fond of happens and black cannot prevent it. The White Queen is then in position(from d2) to go immediately to h6 with mating threats.

<twinlark> gives computer analysis: 1. (0.58): 18...Ng4 19.Bb6 Bf6 20.Nd4 Bg7 21.Qf4 Nf6 22.Nxc6 bxc6 23.Qh4 Rb8

<kwgurge> says:

A)18.Qd2 Nd5 19.Bd4 Bf6 20.Qh6 Bxd4 21.Nxd4 Nf6 22.Rf4 Kh8 23.Rh4 Rg8 24.Rf1 Qe7 25.Nc4 Bxe4 26.dxe4 Qf8 27.Qg5

<jepflast> says: OK, I was complaining about Ng4, but after some investigation I think this will be satifactory for White. One line is given by noctiferus above, except I have 21...Kh8 22. Nd4 Rg8 23.Rae1 gxf5 24.Nxc6 bxc6. Or black can try 19...Rfe8 20. h3 Ne5 21.f6 Bf8 22.Be3.

<Noctiferus> line referenced by <Jepflast> above:

Shredder10 main lines of deep analysis 16 ply on 18.Qd2 18...Ng4 19.Bb6 Rae8 20.h3 Nf6 21 Be3 Rc8 22.Bg5 Bd8 23. Nd4 Nh5 24 Rae1 Re8 +0.88

One word of caution, however: with these (likely) improvements, white is not ahead by 0.88 anymore as in noctiferus's line. It's only 0.3 to 0.5.

1. = (0.19): 19.Bb6 Rfe8 20.Nd4 (20.h3 Ne5 21.f6 Bf8 22.Be3 Qd8 (better than Qe6) 23.Bh6 Nd7 24.Bxf8 Rxf8 25.Qf4 Re8 26.Nd4 0.10)Bf8 21.Qf4 Ne5 22.Nxc6 bxc6 23.d4 Nd3 24.Qd2 Nb4 25.Rf3 Bg7

2. = (0.11): 19.Bg5 f6 20.Bf4 g5 21.Qe2 Ne5 22.Bg3 Bb5 23.Rfd1 d5 24.Kh1 Qxf5