30 March 2015

Assaulting the King

The Nature of Blitz

In a recent blitz game, I made an unsound sacrifice of a knight for two pawns to launch an attack on my opponent's king. Had he defended properly, my attack would have faltered. He failed. My attack, also, contained several inaccuracies. Such is the nature of blitz: blunders characterize the play of both sides.

White to move

Objectively best for White here is 15.Nc3 or 15.Qa5. In both cases, Black is somewhat better. However, I wanted to take the g5 pawn and create an open file in front of Black's king. I needed to decoy away one of the defenders of g5.


An assault on the king begins with an unsound sacrifice.

15...Qxd6 16.Nxg5 Bg7?

The first important error by my opponent.

16.h6 18.Nh3 Kh7 and Black's king remains secure. White has nothing for the sacrifice.


Positionally, this move seemed sensible, but it fails.


17...h6 18.Rg3 Kh8 19.Nf3-+ and Black should prevail.


18.h6 was best.


White to move


This second knight sacrifice is White's only chance.


19...fxe6 20.Qxh6 Ne8 21.Qg5=


Now, White's attack cannot be stopped.

20...Ne8 21.Qg5 f6 22.Qh4

22.Qg6 is better.

Black to move


22...Kf7 is objectively best, but still loses. 23.dxe6+ Qxe6 24.h6+-.

23.h6 Rf7 24.Bh5

24.hxg7 is better.

24.Re7 25.hxg7 Nxg7 26.Bg6

Black to move


Black ran out of time, but it is checkmate in three.

25 March 2015

Building upon Morphy

Ten games won by Paul Morphy are the source for nineteen middlegame positions in GM-RAM: Essential Grandmaster Knowledge (2000) by Rashid Ziyatdinov. These games have provided a core for my chess study the past ten weeks. Today, I am moving on to Anderssen -- Steinitz, London 1862, game 16 in GM-RAM. See "Game of the Week" and "To Know a Position" for detailed discussion of my training program with this book.

I am not finished with Morphy. Two more of his games crop up in Ziyatdinov's list as games 22 and 23. Morphy's eighteen games from the First American Chess Congress have become part of my teaching curriculum for youth players receiving individual instruction. Three of these are among the selections in GM-RAM. I have a yearning, but not yet a firm commitment, to work through all of the available Morphy games along with the annotations in Philip Sergeant, Morphy's Games of Chess; Macon Shibut, Paul Morphy and the Evolution of Chess Theory; and Valeri Beim, Paul Morphy: A Modern Perspective.

Studying Saint Amant -- Morphy, Paris 1858 this past week drove me to my bookcase where an old text had been accumulating dust.

White to move

From this position, Saint Amant played 17.Kh2, and the game continued 17...Rad8 18.Rad1, resulting in position 165 in GM-RAM. It is at this point that Philip Sergeant offers, "White's feeble play, which is not what one would have expected from St. Amant, gives Morphy an opportunity for a neat combination" (Morphy's Games of Chess, 150).

I spent some time exploring 17.Rad1 as an alternative for White. Three Black replies seemed worth examining.

a) 17...Bxh3 as in the game, fails. 18.gxh3 Nxh3 19.Kg2 Nf4+ 20.Kg1 Qg5 21.Qg3 and Black's attack has been parried. Materially, three pawns for the piece may give Black some endgame chances, but with all the major pieces on the board, White is better. Black could eliminate the queens with 21...Qxg3, but the rooks remain and White's pieces are more active.

b) 17...Nxh3 seems promising. 18.gxh3 Bxh3 19.Rfe1 Bg4+ 20.Kg2 Bxd1 21.Rxd1 and Black has a material advantage, while White's king remains vulnerable to attack.

c) 17...Rad8 seems best. After 18.Nb3, Black can sacrifice the knight on h3. 18...Nxh3 19.Rxd8

Black to move

19...Nf4+ is better than 19...Rxd8 because of 20.gxh3.

After mulling through these possibilities, it occurred to me that additional work would be useful on h3/h6 sacrifices. I remembered that Laszlo Polgar, Chess Training in 5334 Positions (1994) has a section in the back of the book devoted to such sacrifices. Pulling it off the shelf, I discovered that positions 4663-4762 are h3/h6 combinations. Polgar presents entire games with a diagram in each case immediately prior to a sacrifice on h3 or h6. There is plenty of study material there to occupy my time.

22 March 2015

The Opera Game

H.R.H. the Duke of Brunswick is a thorough devotee to Caissa; we never saw him but he was playing chess with some one or other. We were frequent visitors to his box at the Italian Opera; he had got a chess-board even there, and played throughout the performance. On our first visit "Norma" was performed. The Duke's box is right on the stage; so close, indeed, that you might kiss the prima donna without any trouble. Morphy sat with his back to the stage, and the Duke and Count Isouard facing him. Now it must not be supposed that he was comfortable.
Frederick Milnes Edge, Paul Morphy: the Chess Champion (1859), 172.
A casual reader of Frederick Edge's Paul Morphy, could easily surmise from this passage that Vincenzo Bellini's Norma was performed on stage when Paul Morphy played his "most famous game".* However, David Lawson, who researched matters thoroughly, asserts that the game was played during a performance of Barber of Seville by Gioachino Rossini.
On the second of November they heard the Barber of Seville, during which Morphy played his most famous game, the Duke again consulting with Count Isouard. In his games with Morphy, the Duke always had a partner, sometimes two, Counts Isouard and Casabianca consulting against him.
Lawson, Paul Morphy: the Pride and Sorrow of Chess (2010), loc. 3271 Kindle ed.
Lawson's documentation is not on par with his research. Verifying his claims, and the counter-claims of many writers before him, has proven vexing. Edward Winter's outline of several unknown details and the contours of disagreement is well-worth reading (citation below).

It seems clear from these two passages in the books by Edge and Lawson that Morphy played several games against the Duke of Brunswick in consultation with one or more Counts. Several of these games were played at the opera as a guest of the Duke. Yet, it seems, only one record of these games remain.

This one recorded game was a stellar performance by Morphy. It is included by Rashid Ziyatdinov in GM-RAM: Essential Grandmaster Knowledge as game 14. It was my study game a week ago, and this past week comprised my "lesson of the week" for youth chess players. It is an instructive illustration of effective and ineffective pins. My annotations were written without sight of a chess board, giving me confidence that I am beginning to understand this game.

The Game

Morphy,Paul  -- Duke of Brunswick, Count Isouard [C41]
Paris, 1858

1.e4 e5 2.Nf3 d6 3.d4 Bg4

The inadvisability of this pin is addressed in "Lasker's Rules".


Black to move

The rapid conclusion of this game may give the impression that the Duke and Count were weak chess players. But, here, they demonstrate understanding of zwischenzug--something many of my young chess students need to learn.

The immediate 4...dxe5 loses a pawn.

4...Bxf3 5.Qxf3 dxe5 6.Bc4

This simple checkmate threat is far too effective against scholastic players. Even at the Washington State Elementary Chess Championship, games will end quickly with Scholar's Checkmate.


The Duke and Count block access of White's queen to f7.


Morphy renews the threat against f7, adding an attack on b7. From b7, the queen may snatch the rook.

Black to move

Thinking that many of my students could quickly find the answer, I asked: How can Black protect both the pawn on f7 and the rook on a8? I was disappointed at the answers.

7...Qe7 8.Nc3

8.Qxb7 Qb4+ 9.Qxb4 Bxb4+ 10.c3

8...c6 9.Bg5

Morphy's pin of the king's knight is vastly more effective that when his knight was pinned, although he, too, violates the letter of Lasker's Rule.


9...Nbd7 seems better.

White to move

10.Nxb5! cxb5 11.Bxb5+ Nbd7

White to move

Morphy has managed to pin both of Black's knights.


Both White rooks will come to d1 for the battle of d7.

12...Rd8 13.Rxd7

It was important to capture with the rook, temporarily sacrificing an exchange when already down material, because this capture keeps a Black piece pinned on d7.

Rxd7 14.Rd1 Qe6

White to move

White is down a rook for two pawns, and yet has more force in the battle. If Black is able to mobilize their bishop and other rook, this game will become one of Morphy's rare losses.

15.Bxd7+ Nxd7 16.Qb8+

Forcing Black's response and leading to checkmate. This queen sacrifice is the essence of the Opera Checkmate, a combination that leads to a more elementary bishop and rook checkmate.

16...Nxb8 17.Rd8# 1–0

*Frank Marshall called Morphy's opera game, "the most famous game of all time" (Comparative Chess [1932]). See Edward Winter, "Morphy v the Duke and Count," Chess Notes (updated 20 March 2014).

19 March 2015

Wasting Time

We are past the Ides of March,* and I am still pursuing my project for the year that began in early December. Each week, beginning Wednesdays, I am going through one game from the 59 in GM-RAM: Essential Grandmaster Knowledge (2000) by Rashid Ziyatdinov. These games comprise the source for 120 middlegame positions that represent significant tactical and positional understanding to aspiring players if each is understood thoroughly.

This week's game is Saint Amant -- Morphy, Paris 1858. It is the fifteenth game in GM-RAM, and the tenth Morphy game. Today's post concerns only the opening moves.

De Saint Amant,Pierre Charles Fourier -- Morphy,Paul  [C54]
Paris, 1858

1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.c3 Nf6 5.d4 exd4 6.cxd4 Bb4+

The moves so far are fairly standard in the Italian Opening. Two moves are equally popular in this position.


Yesterday, I posted a position from a reference game that continued: 7.Nc3 Nxe4 8.0–0 Bxc3 (8...Nxc3 is familiar from the games of Greco) 9.d5 Ne5 10.Qe2 0–0 11.bxc3 Nxc4 12.Qxc4 Nd6 13.Qd3 Qf6 14.Re1 b6 15.Bg5 Qf5 16.Qxf5 Nxf5 17.g4 f6 18.Bf4 Nd6 19.Bxd6 cxd6 20.Nd4 Bb7 21.Nf5 g6 22.Ne7+ Kf7 23.Re3 Ba6 24.Rae1 Rae8 25.h4 Bc4 26.a3 Rh8 27.Kh2 b5 28.Kg3 a5 29.h5 Bb3 30.f4 Bc2 31.f5 g5 32.h6 Bb3 33.Kf2 Rhf8 34.Re4 Bc2 35.R4e3 Bb3 36.Re4 Bc2 37.R4e3 ½–½ Gashimov,V (2740) -- Dominguez Perez,L (2713) Nice 2010.

7...Bxd2+ 8.Nbxd2 d5 9.exd5 Nxd5 10.0–0

10.Qb3 is more popular, and may be stronger.


White to move


Looking through this game, my initial thoughts at seeing this move was recollection of Philip W. Sergeant's comment on the third match game between Paul Morphy and Alexander Meek at the First American Chess Congress, 1857. Sergeant wrote, "In nothing was Morphy so fortunate as in the frequency with which his opponents played P-R3" (Morphy's Games of Chess [1957], 44).

However, Sergeant's comment was instantly tempered in my mind by a comment I read last week in a more recent book. In this case, the reference game is Morphy -- de Riviere, Paris 1863, where Morphy played 9.h3.
Morphy himself was more apt to play this move than would be a modern master. So we have here a concrete example of evolution in strategic understanding; today, beginners are warned against such "wastes of tempo." But as we shall see, such "inaccuracies" don't render the ensuing play transparent.
Macon Shibut, Paul Morphy and the Evolution of Chess Theory (1993), 9.
Moreover, in some situations, even modern masters will play h3 or h6. When is such a move a useful or even necessary prophylactic move? When is it a waste of tempo? General statements such as Sergeant's substitute facile principles for concrete evaluation.

John Watson offers a balanced view in Secrets of Modern Chess Strategy: Advances since Nimzowitsch (1998).
Whether in closed, semi-open, or open positions, flank pawn moves are regularly employed for a variety of reasons, ... Pieces are moved any number of times, if necessary, to achieve strategic goals. And the number of pawn moves in the opening can range from one to eight or more, depending on the requirements of the position. (15)
Several moves are more popular from the diagram than 11.h3, but the best move is not well established yet.

Perhaps, we can say that it would be better for White if he develops one of the rooks. 11...Bg4 does not appear to represent a credible threat worthy of expending a tempo to prevent. As the subsequent course of the game demonstrated, the advanced h-pawn could become a target.

*See William Shakespeare, Julius Caesar (1599).

18 March 2015

The Sentinel

A lonely sentinel stands on the forward point of a fortress. He cannot be removed by the enemy, but nor can he retreat. As long as he stands there, the forces back home are safe from attack.

White to move

This position is from Gashimov -- Dominguez Perez, Nice 2010. It arose via a Greco Gambit in the Italian Opening. I was looking at the game this morning as a reference game to this week's new "game of the week": Saint Amant -- Morphy, Paris 1858.

Gashimov put his rooks on e3 and e1, then advanced his pawns on the kingside. Everything became closed. Black's rooks could not enter the game via the e- or c-file. White's rooks could not leave the  knight unsupported.

14 March 2015

One Game

The past few weeks I have been trying to avoid long strings of online blitz games. It's not always easy. Instead, I try to play a single 15 minute game with a ten second increment. Yesterday's sole game was short, but easily could have lasted longer.

Stripes,J (1999) -- Internet Opponent (2036) [B41]
A Website, 13.03.2015

1.e4 c5 2.d4 cxd4 3.Nf3 

I've been playing around with this move order because I've been studying the games of Paul Morphy.

3.c3 The Smith-Morra Gambit is the usual reason White plays 2.d4.


3...e5 4.c3 (4.Nxe5?? Qa5+–+)

4.Nxd4 a6 5.c4

I play a move that has given me trouble when I've been on the Black side in this position.

5.Nc3 is the most popular choice.

5.Bd3 does well statistically, and was the recommendation of something I read when I started playing the Kan several years ago.

5...Nf6 6.Nc3 Bb4

White to play


7.Qd3 is probably best. See Carlsen -- Anand, Game 6.

7.Qf3 is good.

7...Qc7 8.Nc2

8.Qb3 also scores well.

8...Be7 9.Be3


9...0–0 10.Be2 Nc6 11.0–0 Ne5 

11...Rd8 offers White more difficulties.

12.b3 d5!?

White to move

At first, this move looked dangerous, then I saw that my vulnerable knight would be the second White piece to capture on d5.


13.cxd5?? Qxc3-+.

13.f4 was also good.

13...exd5 14.Nxd5 Nxd5 15.Qxd5

Perhaps Black has compensation for the pawn.

15...Be6 16.Qe4 f5 17.Qd4 Bf6 

17...Rfd8 18.Qb6

18.Qb6 Qxb6 19.Bxb6

Black to move

19...Rac8 20.Rad1

I had been thinking about the vulnerability of this rook on a1

20...Nd7 21.Bd4 1-0

White is a pawn ahead, but there's plenty of play left in the game. Black's resignation seems premature. I was not really tested in this game, but it's always nice to win.

12 March 2015

Lesson of the Week

My chess lessons this week have been tailored for different groups and individuals.


Watching one of my young pupils chase a lone king with a queen and a knight until I, as the tournament director, stepped in and called the game drawn by the fifty move rule influenced my choice of lessons for the groups comprised mainly of beginners. Two after school clubs worked on elementary checkmate with queen and king vs. lone king (see "Teaching Elementary Checkmates").


Last Thursday, my copy of the Encyclopedia of Chess Combinations, 5th ed. (2014) arrived. I immediately put it to work. My advanced club that meets on Thursday afternoon was offered some incentive to work through the first six problems in the volume. Two students took home chess books after solving all six correctly in one hour's labor. At the March Madness youth tournament on Saturday, I created a problem solving contest featuring eight problems from this text. The two participants who scored best (four correct answers) each won a copy of John Nunn's Chess Puzzle Book.

One of these two winners meets with me on Wednesday afternoons for individual lessons. We spent a fair portion of our time looking at some of the tactical alternatives at two critical positions in Bird -- Morphy, London 1858 (see "Hitting the Books"). An abbreviated version of this exercise will be used today and tomorrow with other groups and individuals.

Bird,Henry Edward -- Morphy,Paul  [C41]
London, 1858

1.e4 e5 2.Nf3 d6 3.d4 f5 4.Nc3 fxe4 5.Nxe4 d5

White to move


White should have sacrificed the knight on e4 for an attack. 6.Nxe5!

6...e4 7.Ne5 Nf6 8.Bg5 Bd6 9.Nh5 0–0 10.Qd2 Qe8 11.g4 Nxg4 12.Nxg4 Qxh5 13.Ne5 Nc6 14.Be2 Qh3 15.Nxc6 bxc6 16.Be3 Rb8 17.0–0–0 Rxf2!

17...Bg4 should win easily. The merits of Morphy's brilliant attack are measured by White's defensive resources that Bird missed in the next diagram.

18.Bxf2 Qa3 19.c3 Qxa2 20.b4 Qa1+ 

White to move


After 21.Kc1, Black has a draw by repetition. Does Black have anything better? This position is worth playing out with a teacher or fellow student.

I like 21...a5, which Garry Kasparov claims is unclear. Kasparov suggests 21...Bf5 with a slight advantage for Black.

21...Qa4+ 22.Kb2 Bxb4 23.cxb4 Rxb4+ 24.Qxb4 Qxb4+ 25.Kc2 e3 26.Bxe3 Bf5+ 27.Rd3 Qc4+ 28.Kd2 Qa2+ 29.Kd1 Qb1+ 30.Kd2 Qxh1 0–1

11 March 2015

Hitting the Books


A distinctive element of GM-RAM: Essential Grandmaster Knowledge (2000) draws in some readers and pushes many others away. Rashid Ziyatdinov's book contains diagrams without analysis. The author explains that the book is more of an exam than an instructional book. As such, it is an open book exam that can be taken and retaken until the desired score is achieved. Co-author Peter Dyson suggests that GM-RAM, "can be thought of as both a study outline and as an evaluation tool" (9).

Fifty-nine "classic games" are the source for 120 middlegame positions. Ziyatdinov addresses the definition of "classic". Games are not classic merely because they were played a long time ago. The games in GM-RAM:
...have been analyzed in great detail by many strong players from different periods, different schools of chess, and different ages and generations. It is only after a game has withstood these many different perspectives--these "tests of time"--that it can be considered a classic. (77)
Ziyatdinov directs his readers to analysis of these games by other writers. Alternately, he writes, "a chess trainer can help teach the necessary knowledge" (13). He provides a list of references. This list offers a secondary curriculum. Most, if not all, of the the endgame positions in GM-RAM can be found in Yuri Averbakh's Comprehensive Chess Endings, which comprises the bulk of the texts listed for endgame study.

The middlegame books listed are another matter. Most of the games are from the nineteenth century, but the recommended middlegame books include the two volumes of My Best Games of Chess by Alexander Alekhine; and Bobby Fischer, My 60 Memorable Games. Also listed are Averbakh, Chess Middlegames: Essential Knowledge; Paul Keres, and Alexander Kotov, The Art of the Middle Game; and Hans Kmoch, Pawn Power in Chess. There is minimal analysis of the games of Adolf Anderssen and Paul Morphy in these books.


For the past few months, I have been systematically working through Ziyatdinov's fifty-nine games at the rate of one each week. This past week, my game has been Bird -- Morphy, London 1858 (chessgames.com link). I have not done well on my study of this game. The week has been filled with activities that interfere with personal study, and an exciting new book arrived as well, Encyclopedia of Chess Combinations, 5th ed. (2014), putting my chess study time on another course. Hence, my study of this game will carry over another week. I will press on, though, adding this week's game: Morphy's Opera Game.

I have print editions of three good books on Paul Morphy: Philip W. Sergeant, Morphy's Games of Chess (1957); Macon Shibut, Paul Morphy and the Evolution of Chess Theory (2004); and Valeri Beim, Paul Morphy: A Modern Perspective (2005). In addition, I have the Kindle edition of David Lawson, Paul Morphy: The Pride and Sorrow of Chess, new. ed. (2010) and access to older books, such as Frederick M. Edge's 1859 Paul Morphy: The Chess Champion, via GoogleBooks. Other books that contain analysis of Bird -- Morphy include Garry Kasparov, My Great Predecessors, part I (2003); and Max Euwe, The Development of Chess Style (1968).

Kasparov's book offers a good entry point to the most important historical analysis and sometimes modern computer evaluation of this analysis. It behooves me to invest the time to work through the analysis of this week's and last week's games in all of these books.


Ziyatdinov's GM-RAM contains two essential middlegame positions from Bird -- Morphy. These are separated by a single move. The positions are before 17...Rxf2 and after 17...Rxf2 18.Bxf2. Studying Kasparov's analysis last night focused my attention much earlier in the game.

White to move
After 5...d5
Kasparov credits Johannes Zukertort with having pointed out the improvement from this position that refutes Morphy's dubious opening choice. Bird could have gained an advantage had he properly applied knowledge from the ancient work of Pedro Damiano.

Euwe does not offer a source, but notes, "Nowadays it is known that the answer to Black's chosen variation is 6.Nxe5! dxe4 7.Qh5+, White getting an irresistable attack in return for the sacrificed piece" (29).

06 March 2015

Sacrifice Everything

Black to move

This position is Problem 12 in Encyclopedia of Chess Combinations, 5th ed. (2014), published by Šahovski Informator (Chess Informant). For the past several years, I have been enjoying the electronic edition of Anthology of Chess Combinations, 3rd ed (2005) and will continue to do so. The Anthology was part of my selection with credit that I received for placing fifth in a Chess Informant Reader's Contest (see "Best of the Best: Chess Informant Reader's Contest").

There are many benefits to electronic books. They are light weight and require minimal storage space. The electronic edition of Anthology of Chess Combinations includes a "solver's kit"--a software interface with search capabilities and other features. Solving positions onscreen makes using the Anthology simple. See my review at "Anthology of Chess Combinations".

Print books still have their merits in this electronic age. With a pencil, I can track how I did on each problem and when I solved it (or failed). If I put the book away for a couple of months, my bookmark will remain in place and I can resume where I left off. In my review of the Anthology, I noted that tracking completed problems was the most important feature absent from the software.

The Encyclopedia, 5th edition adds approximately 300 problems over the total in the Anthology, 3rd edition. The organization of problems has changed somewhat. Morphy's famous queen sacrifice (see "Morphy's Immortal") was the first problem in the electronic edition of the Anthology and appears as number 191 (the first of the advanced in the annihilation of defense section) in the Encyclopedia.

I ordered the Encyclopedia last week and qualified for free shipping because I added the Informants missing from my collection. It arrived yesterday. I am happy!

I correctly solved the problem in this post in two or three minutes.

04 March 2015

Decoy Sacrifice

Lesson of the Week

Young players in my school clubs, as well as those receiving individual instruction, are presented with a position from an online game played last week.

Black threatens to bring the rook to g2, forking queen and pawn and demanding further concessions from White. White has the opportunity to make some threats, too.

Qa2 threats the pawn on f7. Black must either defend f7 or find something better.

White to move

Some groups are also working through nine problems on a worksheet, many of them decoys (see "Lesson of the Week [19 December 2012]).

03 March 2015

Manual of Chess Combinations

In January, a young woman showed up at a youth chess tournament that I was running with an interesting book of chess combinations. She told me that this book had been recommended by her coach, who is a Russian master.* She also said that she was working her way through the book and finding many challenging problems that demanded creative and flexible thinking.

I looked up the book on Amazon and added it to my wish list. It is Sergey Ivashchenko, The Manual of Chess Combinations, vol. II (Russian Chess House, 2002).

I almost bought the book that day, but I waffled because I was also planning to buy the Encyclopedia of Chess Combinations, 5th edition (Chess Informant, 2014). At the 23rd Dave Collyer Memorial Chess Tournament, a used copy of Ivashchenko's was among the books that IM John Donaldson brought to sell. It was $10. I grabbed it instantly while he was still unboxing the other books.

Ivashchenko claims that this book is designed to lift players from 1600-1700 up to 2200 or higher. Hence, in theory, I should be able to work through the first two or three sections rather quickly, and then find more challenging problems. Indeed, I have solved a few of the early problems in two minutes or less. Nonetheless, a couple have proven more difficult.

After five minutes with Problem 15, I worked out the idea and verified that it would succeed, but I missed a nuance in Black's defense and so did not have the whole sequence correct.

White to move

*The young woman has been playing in my youth tournaments since fourth or fifth grade. She is now a high school senior. When she was in middle school, I coached her for a month or so to prepare her for the Idaho Girls State Championship, where she tied for first. She now plays as well as a solid A Class player, albeit with a much lower rating. I was happy that I was not paired against her in the Collyer. If she keeps playing chess through college, she will become a master.

01 March 2015

The Critical Position

This position comes from my round two game against Bob Rajala in the 23rd annual Collyer Memorial. He played a Caro-Kann, and I struggled to find an advantage. We reached a position where I thought that I could be in trouble with a single inaccuracy.

Black to move

Bob played 28...Qg6, much to my relief. I exchanged my rook for his bishop and knight, and ten moves later he resigned.

I expected, and he had planned, 28...Qg5!

Should I play 29.g3 or 29.g4?