Chess patron Isador Samuel Turover, throughout his life, offered cash prizes for brilliancies in chess games. In 1930 Turover gave a 500 lire brilliancy prize at the tournament in San Remo. In 1974 he established the annual World Brilliancy Prize, the first winner was Michael Stean who received $1,000 for his win against Walter Browne at the 21st Chess Olympiad in Nice.

First,

Where did black go wrong? Although not all moves are precise, Stockfish believes the move that starts to get black into trouble was 10 b6.

Stockfish prefers 10 .. Be7

After the game move (b6)
1) +0.32 (18 ply) 11.Qd3 Qc7 12.Qg3 h5 13.h3 h4 14.Qe3 Be7 15.f4 O-O 16.Rae1 Bb7 17.Bxe6 fxe6 18.Nxe6 Qc6 19.f5 Rae8 20.Nxf8 Rxf8

After Be7:
1) -0.11 (20 ply) 11.Qe2 O-O 12.Rad1 Qc7 13.a4 Bd7 14.g3 Qc5 15.Kg2 g6 16.Rfe1 e5 17.Nf3 Kg7 18.h3 Rac8 19.Qd2 b5 20.axb5 axb5

The next move that really changes the game is 14 b5:

1) +1.99 (23 ply) 15.e5 Qb6+ 16.Kh1 dxe5 17.Qg6+ Kd8 18.Qf7 Qc5 19.fxe5 Rf8 20.Qxg7 Nh5 21.Rxf8+ Kc7 22.Qxe7+ Qxe7 23.Rf7 Qxf7 24.Bxf7 Nf4 25.Kg1 b4 26.Na4 Bxg2 27.Nc5 Bc6 28.Rd1 Nh3+ 29.Kf1 Ng5 30.Ne6+ Nxe6 31.Bxe6 Rf8+ 32.Kg1 Re8 33.Rd6

Better is Qc7:

1) +1.24 (17 ply) 15.Kh1 Qc6 16.Rad1 h5 17.Nd5 Rh6 18.c4 Ng4 19.Bf5 Kf8 20.e5 Qe8 21.Nxe7 Kxe7 22.exd6+ Kf8 23.d7 Qd8

The problem with computer analysis is that we can wse ethe 'anser' but not the 'why'. My guess is that 14 Qc7 stops e5. After 14 Qc7 15 e5:

1) +0.67 (21 ply) 15...Qc5+ 16.Kh1 dxe5 17.fxe5 Qxe5 18.Rae1 Qg5 19.Bh3 Rd8 20.Qe2 O-O 21.Qxe7 Ba8 22.Qe6+ Kh8 23.Qxb6 Ng4 24.Qg1 Qh4 25.Bxg4 Qxg4 26.b3 Qg5 27.Re2 Bb7 28.Ne4 Rxf1 29.Qxf1 Bxe4 30.Rxe4 Qd2 31.Re2 Qd4

Stockfish doesn't do a good job of explaining how it got from 0.3 type position it a 1.2ish type position, I'd probably need to run on a higher ply. It shows good scores but changes them after the recommended moves are made.

The bishop sacrifice 19 ... Bg2 ch
doesn't work against a computer (and apparently not against Stean either). I suspect Browne wanted to shake things up.

After 19 ... Bg2ch:

1) +3.85 (22 ply) 20.Kxg2 Rf8 21.Rad1+ Kc7 22.Qxg7 Rg8 23.exf6 Rxg7+ 24.fxg7 Qg5+ 25.Kh1 Qxg7 26.Rf7 Qxf7 27.Bxf7 Rd8 28.Nd5+ Kb8 29.c3 Bh4 30.Bg6 Rf8 31.Rd4 Bg5 32.Re4 Kb7 33.Kg2 Bc1 34.b4 Bg5

Instead, Rf8 maintains the bad position:

1) +1.85 (25 ply) 19...Rf8 20.Qxg7 Nh5 21.Rxf8+ Kc7 22.Qxe7+ Qxe7 23.Rf7 Qxf7 24.Bxf7 Nf4 25.e6 Bxg2+ 26.Kg1 Bc6 27.e7 Kd6 28.Re1 Be8 29.Bxe8 Rxe8 30.Ne4+ Ke6 31.Nc5+ Kf6 32.Nxa6 Rxe7 33.Rxe7 Kxe7 34.Nc7 b4 35.Kf2 Kd6 36.Kf3 Nd5 37.Nxd5 Kxd5 38.Kf4

It was specualted that 14 ... Bc8 may be best for black, but it doesn't:

1) +1.99 (22 ply) 15.Bb3 Kf8 16.e5 dxe5 17.Qxd8+ Bxd8 18.fxe5 b5 19.Kh1 h5 20.Rad1 Bg4 21.Rd2 Bc7 22.exf6 Rd8 23.Nd5 Rh6 24.h3 Bb6 25.Kh2 Ba5 26.Rdf2 Be6 27.fxg7+ Kxg7 28.Ne7 Kh8 29.Nf5 Bc7+ 30.Kg1

Finally, it ws asked what about 30 Ka6:

One of the posters got it right according to the Silicon monster:

1) mate-in-5 (23 ply) 31.a4 Bb8 32.axb5+ Qxb5 33.Nxb5 a4 34.Rxb8 a3 35.Rd6#

<<Breunor> I went to check what the Silicon monster thought of the key moves in this fascinating game.>

If you're going to use Stockfish to analyze positions you need to let it search much deeper than low 20-plies in order to have any confidence in his evaluations. Somewhere in the high 30-plies is a minimum, and somewhere in the 40-ply range is preferable. Of all the major engines, Stockfish is the one that prunes its search tree the most aggressively and therefore it can reach deeper search depths than all the other major engines in a similar amount of time.

The tradeoff is that it is the most susceptible to missing good moves, particularly at low plies. So you need to let it search deeper so that lines with good moves are less likely to be discarded from its search tree. Even better, IMO, is to check the analysis of every position with multiple engines, since different engines have different evaluation functions and different heuristics for pruning their search tree. Most of the time they will come up with different evaluations and different lines, and even different move rankings. And every engine analysis needs to be verified manually, checking for adverse influence of the horizon effect, and, if time permits, do forward sliding to ensure that promising alternate better moves are not missed.

As an example, following up on <whiteshark>'s 10-year old notion that Black was already lost after 12...Be7 (M Stean vs Browne, 1974 (kibitz #8)) I had Stockfish 9 analyze the position. Here is its principal variation following 12...Be7:

click for larger view

[+0.57], d=44: 13.Nxe6 fxe6 14.Bxe6 Kf8(2) 15.Kh1 b5(3) 16.e5 dxe5 17.fxe5 Qxd3 18.cxd3 Rd8 19.Rad1 b4 20.Na4 Rd4 21.Rc1 Rxd3 22.Nc5(4) Bxc5 23.Rxc5 Rd2(5) 24.Bh3 Kg8 25.Rc7 Ng4 26.Rxb7 Nf2+ 27.Kg1 Nxh3+ 28.gxh3 Kh7(6) 29.Rff7 Kg6(7) 30.Rxg7+(8) Kf5 31.Rgf7+ Kxe5(9) 32.Rbe7+(10) Kd5 33.Rd7+ Ke6 34.Rfe7+ Kf6 35.Rxd2(11) Kxe7 36.Kf2 Rc8 37.Kf3 Rc4 38.Kg3 Rc5 39.b3 Ke6 40.h4 Kf7 41.Kf3 a5 42.Kg4 h5+ 43.Kf4(1)

(1) The first thing to do, I think, is to verify whether Stockfish's initial evaluation was reasonable or whether the horizon effect distorted it. So I do what I call "forward leaping" (a pun on the widely used forward sliding) which aims to do just that. Restarting the analysis from this position Stockfish evaluates the resulting positions at [+0.90], d= 49 after either 43...Ke6 or 43...a4 so Stockfish's original evaluation was not unreasonable; the horizon effect did not significantly affect the original evaluation. White has a slight advantage but nothing more; certainly not a winning one. Still, I did some forward sliding to see if Stockfish could uncover better moves than the ones it listed in its principal variation.

(2) Stockfish immediately deviates from Browne's 14.b5 and removes its king from the e-file, even if it means that Browne must give back the piece, block the Rh1, and emerge a pawn down. But at least he reduces White's attacking chances by exchanging queens. However, that would seem to be a difficult decision to make OTB.

(4) 22.Rc7 seems more aggressive but it amounts to pretty much the same thing after 22...Be4 23.Rc4 Bb7. And at d=35 Stockfish evaluates 22.Nc5, 22.Rc7, and 22.Rd2 approximately equal.

(5) Now Black seems to have adequate counterplay.

(6) I thought that 28...Rxb2 held promise but after 29.Rff7 Black is lost.

(7) 29...Rg8 would be very passive, although at d=45 Stockfish evaluates the resulting positions as even, with evaluations of [+0.08] after either 30.Rxb4 and 30.Rf2 and [0.00] after 30.Ra7.

(8) Since this apparently leads to the loss of the Pe5 I thought that perhaps 30.e6 held promise but after 30...Rxb2 Stockfish at d=42 indicates that Black can achieve a draw by repetition after 31.Rxg7+ Kf6 32.Rg2 Rb1+ 33.Kf2 Rc8 34.e7 Rb2+ 35.Kf3 Rc3+ 36.Kf4 Rxg2 37.e8Q Rf2+ 38.Kg4 Rg2+ 39.Kf4 since 39.Kh4 leads to mate after 39...Rc4+ 40.Kh5 Rg5+ 41.Kxh6 Rh4+ 42.Qh5 Rhxh5#. 30.e6 loses a valuable tempo and that's all that Black needs to secure the draw!

(9) Since the Pe5 does not seem to be going anywhere and it's not easy to protect, I thought that perhaps 31...Rxb2 protecting the Pb4 and threatening the Pa2 seemed better. And indeed it seems that the horizon effect is starting to demonstrate its curse since after 32.e6 Stockfish at d=42 evaluates the resulting position as exactly even, [0.00], and all other moves by White results in evaluations in Black's favor.

(10) And Stockfish disdains capturing Black's Pb4 and protecting White's Pb2, preferring to keep its rooks connected on the 7th rank and apparently trying for a draw by repetition as its best chance.

(11) White can force a draw by repetition with 35.Rf7+ after either 35...Ke6 36.Rfe7+ or 36...Kg6 37.Rg7+ since 37...Kh5 leads to 38.Rxd2 Rc8 39.Rd5+ Kh4 40.Rg6 Rc1+ 41.Kg2 Rc2+ 42.Kf3 Rc3+ (good old horizon effect!) 43.bxc3 h5 44.Rgg5 Kxh3 45.Rxh5#. Apparently Stockfish as White wants more. But as the "leaping forward" evaluation showed, there isn't much more to be had.

click for larger view

Analysis by Houdini 4: d 25 dpa done

1. + / = (0.50): 14...Qc7 15.Kh1 Qc6 16.Qg3 h5 17.Qxg7 Rh7 18.Qg5 Kd8 19.Nd5 Nxd5 20.Qg8+ Qe8 21.Qxh7 Nf6 22.Qg7 Bxe4 23.Rae1 Qg6 24.Qxg6 Bxg6 25.Re2 Bf8 26.Rfe1 Ra7 27.c4 Rg7 28.f5 Be8 29.Rd1 Bc6 30.h3 Kc7 31.Kh2 Re7 32.Rd4 Bh6 33.Re1 Rg7 34.g3 Bf3 35.Rd3 Be4

2. + / = (0.52): 14...Kf8 15.Kh1 b5 16.e5 dxe5 17.fxe5 Qxd3 18.cxd3 Rd8 19.Rad1 b4 20.Na4 h5 21.exf6 gxf6 22.d4 h4 23.Nc5 Bxc5 24.dxc5 Rxd1 25.Rxd1 Rh5 26.Rc1 Ke7 27.Bh3 Re5 28.Kg1 Bc6 29.Bg4 f5 30.Bh5 Kd7 31.Kf2 Bd5 32.Bf3 Kc6 33.Rc4 a5 34.Rxh4 Kxc5 35.Bxd5 Rxd5 36.Ke3 Re5+ 37.Kf3 Re6 38.Rh7

click for larger view

Analysis by Houdini 4: d 22 dpa done

1. + / = (0.33): 12...Qc7 13.Rad1 Be7 14.Bxe6 fxe6 15.e5 0-0 16.exf6 Rxf6 17.f5 exf5 18.Nxf5 Raf8 19.Nd5 Bxd5 20.Qxd5+ Kh8 21.c3 Qd7 22.Ne3 Rxf1+ 23.Rxf1 Rxf1+ 24.Kxf1 Bg5 25.Kf2 Bxe3+ 26.Kxe3 b5 27.Kd3 Kh7 28.h4 g6 29.a3 Kg7 30.Qd4+ Kg8 31.g4 Qe6 32.Qe4 Kf7

click for larger view

BLACK TO MOVE
In this position Browne had an opportunity to return the píece and run away with his king to b8 and a7. At any rate, Black has a clear advantage, but 18...Kc7!? would at least have put Black's King at a safer place, for example: 19.Qxe7+ Kb8 20.fxe5 Re8 21.Qxg7 Nh5 22.Qxh6 (or 22.Qg4 Qxe6) Rg8, and Black has counterplay while his King will be safe. (Diagram)

click for larger view