Nezhmetdinov - Chernikov - 1962 - Queen Sac

[Lyudmil Tsvetkov]

The reason that engines seem to have a hard time with this problem is that taking the queen with exf6 is not the correct response after the queen captures the bishop.

Position after QxB:
[d]r1b2rk1/pp1ppp1p/5Qp1/q7/3nP3/1BN1B3/PPP2PPP/R4RK1 b - -

3 best Responses:

Code: Select all

 31/50	07:13	1,161,925,068	2,694,437	-6.51	Qxc3 Bxd4 exf6 Bxc3 b5 Rfd1 a5 a4 bxa4 Bxa4 Bb7 f3 Bc6 Bxf6 Rfc8 Bc3 Ra7 Kf2 h6 Rd4 Kf8 Bb3 Rb8 Bc4 Bb5 b3 Bxc4 Rxc4 Ra6 Bxa5 Ke8 Rc5 Rc6 Rxc6
 32/50	07:13	1,161,925,068	2,694,437	-0.41	exf6 Bxd4 d6 Nd5 Rd8 Bc3 Qb5 a4 Qa6 Nxf6+ Kg7 Rfd1 h5 Rd3 Be6 Nxh5+ Kf8 Bxe6 fxe6 Nf4 Kf7 Rh3 e5 Nd5 Kg8 Nc7 Qe2 Nxa8 Rxa8 Rd3 Rd8 Bxe5 Qxc2 Rad1 Qxa4 Rd4 Qc2 Bxd6 Re8 f3 Qxb2 Rb4
 32/50	07:13	1,161,925,068	2,694,437	 0.00	Ne2+ Nxe2 exf6 Nc3 d6 Nd5 Be6 Nxf6+ Kg7 Bd4 Rfc8 c3 h6 f4 Bc4 Rfe1 Kf8 e5 dxe5 Nd7+ Kg8 Nf6+ Kg7 Rxe5 Qa6 Ree1 Bxb3 Nd7+ Kh7 axb3 Qb5 Nf6+ Kg7 Nd7+ Kh7

straigth recapture ef6 instead of Ne2 loses a couple of tempos, and should be even easier win.

please see the imbalance position I referenced; I am certain a similar position could be reached from almost any line with transpositions, or otherwise black gets even worse.

but, carefully checking with SF and Komodo, those engines see draws all the time, so from now on I have decided to call Komodo Komodraw, while SF Capture-capture-check-check, as that is all the engines see well, a sequence of captures (because of quiescence search), and a sequence of checks (because of check/mate extensions), apart from that, they are totally clueless.

I really wonder how could SF see some really good quiet moves, when it has just some 100 eval terms, and lacks refined psqt? If it had 1000 or 2000 terms instead, it could have indeed seen something worthwhile, but, unfortunately, that is not the case.

deep search helps only when there are concrete threats, otherwise you need refined eval to pick the best move.

only in the critical imbalance position, apart from having skewed imbalance eval, SF has just a single one of the 5 or 6 major pawn structure eval terms, defining compact pawn structures, needed to return correct eval.

but of course, even if you have all the rigth terms, it is still very very difficult to properly tune them; so far I have not found a reasonable way of fine-tuning a larger pool of 2000-3000 terms, and, with much smaller number of terms, eval will be always skewed.

[cdani]

again, we get at this position:

[d]r4r2/pp3pkp/6p1/q2B4/4P3/4B3/PPP2PPP/R4RK1 b - - 0 7

Sure! This type of positions are a pleasure to play for the side with two bishops, a lot of chances to win and very difficult to lose. And sure the evolution of the position depends nearly 100% on quiet moves, for which the engines have not much relevant parameters.

I'm a bit repetitive of what Lyudmil told, but I think is necessary.

[yanquis1972]

again, we get at this position:

[d]r4r2/pp3pkp/6p1/q2B4/4P3/4B3/PPP2PPP/R4RK1 b - - 0 7

or with a bishop on d6, which is easily won for white.

anyone can refute that?

I have seen thousands such positions and know what I am talking about.

btw., if anyone tries to analyse or even implement some imbalance feature based on this position, it will most certainly fail, because:

- above imbalance is won, but, if you remove one pair of rooks, it migth be won, or draw, depending on very small factors

- if you remove another pair of rooks, Q vs BB + 2 pawns is already a draw, or even favours the queen

- so basically, not possible to implement this imbalance, unless some very refined additional features are implemented, like number of pieces for both sides with particular imbalances; the secret here is that it is not just the 2 bishops vs the queen, but also the BB vs the rooks, the RR vs the queen, the R+B vs the queen, and the pawns ve the enemy pieces; with smaller number of pawns migth still be drawn

- white has decisive edge in terms of more compact pawn structure, which weighs into the balance; imbalance without good pawn structure migth be insufficient

- it is also vital how minor pieces and rooks defend among themselves; trying to implement 3 minors vs queen feature, fro example, migth fail, simply because when this configuration is reached, the minors migth be defended or not, in one case the minors will win, in the other the queen migth have the upper hand, and search possibly has no way to distinguish between these, as positions arrived at are quiet

- etc., etc.

however, apart from that, I am 100% certain above position is a white win.

best line of play is Rd8 (to sacrifice the rook for the bishop) Rfd1, followed by a4 and c3, with massive advantage.

just waking up so i'll examine in more detail later, but i don't include your line bc i didn't think it was optimal for black. the line i gave i tried to make optimal moves for both sides, but suggestions for white are welcome. at the time i didn't see a place for f4-f5.

your (& daniels) assertion that its already proven win is not, as far as i know, true. i'm sure this sacrifice has been analyzed very deeply & i'm certain i've never seen it evaluated as winning. so between that & the machine evaluation, i'm looking for something concrete, which i've not yet found.

[yanquis1972]

specifically looking like i need a refutation of 14...d6 (after Nc3). mentioned by lyudmil but i don't see anything about it. i havent examined the sharper 14...d5 suggested, but without a line against d6 i don't see the point yet.

[Lyudmil Tsvetkov]

youre making a claim that goes against both prior human consensus & against engine evaluation. it's a bit silly to get defensive when asked for concrete evidence to support the claim...

in the line you gave, there seems to be a blunder on move 15; Kg7 should equalize.

that is one thing I can not do: analyse every possible line.

that is why I expect help/suggestions/refutations from the community.

move 15 involves precisely Kg7.

I know both positions hide much bigger advantage for white, but one has to sit for hours to find the best moves. this is a complex mg position.

my mistake; at move 13, g5 is a blunder.

i'm trying to do your work for you & not making progress, which is the point. the burden of proof is on you here. with optimal moves, i see zero reason, other than your claim, to think that Qxf6 leads to anything better than a draw.

[pgn][Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[SetUp "1"]
[FEN "r1b2rk1/pp1ppp1p/5bp1/q7/3nP2Q/1BN1B3/PPP2PPP/R4RK1 w - - 0 12"]
[PlyCount "14"]

12. Qxf6 Ne2+ 13. Nxe2 exf6 14. Nf4 d6 15. Nd5 Be6 16. Nxf6+ Kg7 17. Bd4 Bxb3 18.axb3 Qd2 *[/pgn]

I really do not know what you want from me.
I feel cold and uncomfortable, so I can not vouch for anything I say, still I can only confirm I am 97% certain white wins.

Please note, that 97% does not mean 100%, the position is extremely complex, and a mistake can always creep in.

so far, we concluded early d5 loses, Re8 seemingly is also inferior and Nezhmetdinov convincingly proved that. you (engines) suggest now d6 instead, ok. starting from your 18th move, Qd2, white plays c3, and after Qb2 (seemingly most reasonable move, but Rc8 is also possible; I can not check all moves, but I guess they more or less transpose to similar lines), white plays f4, as already suggested:

[d]r4r2/pp3pkp/3p1Np1/8/3BPP2/1PP5/1q4PP/R4RK1 b - f3 0 9

well, we are here now, and I still do not see what has changed from the other winning positions we considered. I still assess this position with some 50-70cps white edge in terms of eval at the root: same 2minorsRR vs QRR imbalance, white lacks 2 pawns, but one of them, either d6 or h7 falls soon, white has 2 strong connected storming pawns e4 and f4, very strong attacking knigth outpost on f6, plus strong attack, plus all white minor pieces are well defended.

my assessment is: almost the same as the imbalance position we arrived at earlier, so I do not see why the outcome should be different.

there are many black options to try to defend, I can not guess every move of yours or SF/Komodo, but white plan is more or less set, f4-f5 (sometimes h4 before that to prevent g6-g5), Rf1-f3 and Raf1, with massive attack. I know SF and Komodo want to go for the draw on each and every ply, but that is the wrong decision, white simply piles up all forces on the king side, giving no checks and exchanging nothing.

I suppose the black position should crumble at some point, but that migth take 20-30-50 plies or more: too long for engines to see in search, and too complicated for humans to pass an easy judgement.

again, I guess this is a win, but this is too complicated and long, with too many sidelines, so the probability for a very small number of drawing lines still exists.

at any case, if white has the attack all the time and over 90% winning probability for most lines after the queen sac, while without the sac the game is a straightforward (and dull) draw, I do not see why you simply do not accept Qf6 is best and deserves 2 exclamation marks?

[Lyudmil Tsvetkov]

indeed, for most the lines, I get to a white plus or significant white plus from SF, but that is only some 25 moves on or so. that makes 50 plies - impossible to see for engines.

white alternates doubling the rooks on the f file with f4-f5, h4-h5 and Re1-e3-h3, depending on what black plays. the black king has a hard time, either in the king corner, or in the center.

if someone tries to follow with engines, the fact that top engines would like to draw straigth away, and, if white avoid the draw, the engine briefly sees black advantage, but then wants to draw again 5 moves later, and then 10 moves later, that should certainly strike a bell: if white can draw 10 moves later, why should white agree to a draw much earlier, especially when having the attack?

engines have terribly wrong time here, this is a position for which engine terms are lacking altogether or improperly tuned, whether the search or eval.

[yanquis1972]

i'm not sure exactly what's so painful about this, but i apologize. i'm not attempting to torture you; you claimed Qf6 is absolutely winning & i do not see it. if your claim was that its a !! & gives best winning chances, thats entirely different & i think youve made a very good case.

as is youve given me concrete lines & plan that i can springboard off of (i think with d6 & the removal of the light sq bishop the attack is weaked) which i greatly appreciate.

[yanquis1972]

[d]r4r2/pp3pkp/3p1Np1/8/3BPP2/1PP5/1q4PP/R4RK1 b - - 0 9
i dont know that its objectively best, but i think Rfc8-c5 (after either f5 or h4) allows black to hold.

[Lyudmil Tsvetkov]

i'm not sure exactly what's so painful about this, but i apologize. i'm not attempting to torture you; you claimed Qf6 is absolutely winning & i do not see it. if your claim was that its a !! & gives best winning chances, thats entirely different & i think youve made a very good case.

as is youve given me concrete lines & plan that i can springboard off of (i think with d6 & the removal of the light sq bishop the attack is weaked) which i greatly appreciate.

nothing to apologise for, John, I appreciate very much your interest, it is just that I am not in my best form currently.

thanks you very much for the kind words!

as said, if my eval assessment for some 60cps or so white advantage is correct, than this is just above the drawing lines in the mg (possibly some 45cps accurate eval), so expecting quick win, or mate, is not the way to go. there is no forced mate above, neither a very quick win, a long struggle is still ahead by toughest black defence, and seemingly the lines you suggest are good tries to hold.

still, barring miracle, white should be winning.

[Lyudmil Tsvetkov]

[d]r4r2/pp3pkp/3p1Np1/8/3BPP2/1PP5/1q4PP/R4RK1 b - - 0 9
i dont know that its objectively best, but i think Rfc8-c5 (after either f5 or h4) allows black to hold.

me neither, there so many lines, your suggested seems indeed a good try to try to defend.

I can not, especially now, check all reasonable lines, but, just as an illustration, and fully following my intuition, white should continue with Rfc8 f5 Rc5 Ne8 Kf8 Nd6 (this is an important pawn, white further improves its imbalance, nowhere to hurry) Rc7 b4 (again, no hurry, simply keeping the pawn and improving pawn structure):

[d]r4k2/ppr2p1p/3N2p1/5P2/1P1BP3/2P5/1q4PP/R4RK1 b - - 0 13

and, much to the chagrin of SF and co, this position very much favours white, quite probably winning. white continues with e5, f6, Rae1-e3-h3, slowing building up pressure, and black should crumble at some point. at any case, here SF score changes from negative for white to a white plus, I do not have the time to check how big that plus becomes, but in terms of eval the position is very favourable for white.