Criteria for Training Exercises - a good read

An interesting article in Chesscafe that caught my eye.

Requirements for Training Exercises: by Mark Dvoretsky

Now let's talk about the requirements that examples used in training should meet. It is essential that the exercises should be interesting, that the solutions that are found (or even not found!) should give pleasure, and that they should be remembered for a long time, provoking a desire to keep studying.

Difficulty of solving. You should not choose problems that are either too easy or extremely difficult – here it is important to use moderation. Sometimes it makes sense to use exercises of increased complexity, but the difficulties should not be purely analytical ones. Only the pleasure received from examining a subtle and beautiful solution can compensate for annoyance at not finding the solution.

Necessity and uniqueness of the solution. Here are some examples, taken from various collections of exercises, that did not seem very successful to me.

Stefanov – Andreev
Bulgaria, 1957

[FEN "rn3rk1/4ppbp/b2p2p1/2pP2B1/q3N3/5P2/PP1Q2PP/1K1R2NR b - - 0 1"]

Black to play

In the game there was 1...Qxa2+ 2.Kxa2 Bd3+ 3.Kb3 c4+ 4.Kb4 Na6+ (4...Nc6+!?) with a quick mate. But it is obvious that both 1...Bd3+ and 1...Bc4 win easily, and after the simple 1...Nd7 White's position remains hopeless. The search for a solution in these situations does not provide any pleasure, as it is clear that you can stop on virtually any move.

Alternative routes to the goal also spoil the impression.

Khasin – Stein
Soviet Championship, Tallinn, 1967

[FEN "r4rk1/1bq2p1p/p2ppb1B/1p2n3/3QP3/2N2B2/PPP4P/R4R1K b - - 0 18"]

Black to play

18...Bg7?? 19.Rg1 Ng6 20.Qxg7# or 18...Nd7?? 19.Rg1+ Kh8 20.Bg7+ (20.Qxf6+) lose immediately. But in the case of 18...Rfd8? White develops a very strong attack by means of 19.Bh5! Qe7 (19...Bh8? 20.Bxf7+! Nxf7 21.Rg1+), and now not 20.Rxf6 Qxf6 21.Rg1+ Kh8 22.Bg5! Qf5!, but 20.Rf4! with a subsequent 21.Raf1 or 21.Rg1+.

We come to the conclusion that it is essential to sacrifice the exchange, for which Black will have more than sufficient compensation: a pawn and a strong dark-squared bishop whose opponent, the h6-bishop, will disappear from the board.

In the game there followed 18...Kh8! 19.Bxf8 (19.Rg1 Rg8; 19.Bh5 Bg7) 19...Rxf8 20.Rad1 Rd8 21.Bg2 Bg7 (threatening 22...Nc4) 22.Qf2 Rg8-/+, and Black won the subsequent battle.

Testing allowed us to find a no less promising version of the exchange sacrifice: 18...Qc5! 19.Qxc5 dc 20.Bxf8 Kxf8 21.Bg2 Ke7-/+ and 18...Bh8! (threatening 19...Ng6) 19.Qg1+ (19.Bxf8?! Nxf3-/+) 19...Ng6 20.Bxf8 Rxf8-/+. We could think for a long time trying to choose which of them is best, but the point of the exercise is something else entirely: you have to easily and quickly decide on a positional sacrifice.

Generally it is desirable that a solution be "pure," that all the moves you have to find are unconditionally the best, and if possible even the only ones. For the sake of this a coach may even slightly adjust the initial position of an exercise.

Gragger – Barcza
Olympiad, Varna, 1962

[FEN "8/1p6/6b1/p2k4/8/2P4p/1P1K4/6B1 b - - 0 74"]

1…?

The correct plan is associated with transferring the king to b3 and then preparing for it to break through on the kingside to the h3-pawn.

74...Kd5-c4! 75.Bg1-h2 Kc4-b3 76.Kd2-c1 a5-a4

White resigned because of 77.Be5 Bc2! 78.Bd6 (78.c4 b6) 78...a3 79.ba Kxc3, then Ba4, b7-b5, Kd3 and so on.

But the move 74...Ke4 did not let go of the win, as after 75.Ke2 it still was not too late to go back to the same plan: 75...Kd5! 76.Kf3 Kc4 77.Bb6 a4-+. Black could also think about 75...Bh5+ 76.Kf1 Bf3 (in the case of 76...Kd3 the bishop managed to defend its queenside: 77.Bb6! a4 78.Bc5 Kb3 79.Ba3=) 77.Kf2 (77.Bh2? Kd3 78.Bc7 Kc2-+) 77...Kd3!? 78.Kxf3 Kc2. True, as subsequent calculation shows, White can save himself here.

Unnecessary additional variations can be avoided by moving the white bishop to h2 in the position in the diagram. In that case you then get strictly one method of winning.

It goes without saying that the above-mentioned requirements for an exercise are not dogma, and sometimes it makes sense to ignore them. So, a second solution is allowable if it is clear and instructive.

Sumpter – Stream
Australia, 1964

[FEN "r2n1r1k/5p1p/1n1p3P/q3p3/pb2PN2/1P2QP2/4BPR1/BK4R1 w - - 0 1"]

1.?

Both kings are in danger. It is important to be the first to create mating threats. 1.Nh5? ab is bad, so we have to look for something more decisive.

1.Qe3xb6!! Qa5xb6 2.Nf4-d5 Bb4-c5

If 2...Qxf2 3.Rxf2 Bc5, then 4.Rg7! (threatening 5.Nf6) 4...Ne6 5.Rfg2. In the case of 2...Qa5, the outcome is the same as on 2...Bc5.

3.Nd5-f6 Qb6xb3+ 4.Ba1-b2 Nd8-e6 (or 4...Nc6) 5.Rg2-g7

Mate is unavoidable.

An appealing combination! But another path to the goal is no less beautiful and convincing: 1.Nd5! Nxd5 (1...ab 2.Rg8+! Rxg8 3.Rxg8+ Kxg8 4.Qg5+ Kf8 5.Qe7+) 2.ed! (but just not 2.Qg5? Ne6) 2...f6 (both 3.Bxe5+ and 3.Qg5 were threatened) 3.Qf4!!.

The limits of calculation. Which moves and variations that illustrate them should the coach examine in the solution to the exercise and require to be found, and which ones can the student rightly ignore? Let's look at the next example.

Perez – Najdorf
Torremolinos, 1961

[FEN "7r/6Q1/pB1pkn2/7q/1p6/1P6/P1r5/1N3RK1 w - - 0 34"]

1.?

Both kings are in danger, and the outcome of the battle depends on whether White can chase the enemy king into a mating net with endless checks.

34.Qxf6+? Kd7 35.Qg7+ Kc6 36.Qc7+ Kd5 37.Qb7+ Rc6 (the checks ended) 38.Rc1 was played in the game, and White resigned.

Having convinced ourselves that with a capture by the queen on f6 the attack quickly fizzles out, and concentrated on the search for a sensible alternative, we can find a beautiful combination.

34.Rf1xf6+! Ke6-d5 35.Nb1-c3+!! Rc2xc3

35...bc 36.Qb7+ Ke5 37.Qe7+ Kd5 38.Rxd6#.

36.Qg7-g2+ Kd5-e5 37.Bb6-d4+!Ke5xd4

[FEN "7r/8/p2p1R2/7q/1p1k4/1Pr5/P5Q1/6K1 w - - 0 38"]

1.?

In a practical game a player who reaches this position in his calculations will decide it is worth going for, as the attack looks formidable, and it can be carried out by various means (the most tempting is 38.Qd2+ or 38.Rxd6+, but there are also other checks). And there is nothing better, anyway, which is easy to prove by process of elimination. The player will think exactly the same thing in training, so the task of "calculating the variations to mate" from the initial position will not seem interesting to him. And in general, solving these kinds of artificial problems often leads to the development of a harmful skill: calculating variations even when you do not need to.

On the other hand, if we pose the question differently: "find the best path for conducting the attack," or, even better, just ask them to make a decision for White, we get a decent exercise that develops inventiveness. And the choice and calculation of the most efficient method of pursuing the enemy king in the position that arises by force after Black's thirty-seventh move could be a topic for a new exercise aimed at training the ability to calculate.

We can immediately reject 38.Rf4+? Kc5, and Black wins.

In the case of 38.Qf2+?! Kd5 39.Qd2+ Kc6! 40.Qxd6+ Kb7 41.Qxa6+ Kc7, nothing more than perpetual check is evident.

In the book from which this example was taken, a suggestion is 38.Qd2+!? Rd3 39.Qxb4+ Ke5 40.Qf4+ Kd5 41.Qxd6+ Ke4, and now, as is easy to prove, both 42.Qf4+ and 42.Re6+ win. Instead of 38...Rd3, more stubborn is 38...Ke4. The next variation seems to be forced: 39.Qf4+ Kd3 40.Rxd6+ Kc2 41.Rd2+ Kc1 (41...Kb1 42.Qe4+) 42.Rh2+ Kb1 43.Qe4+ Kc1 44.Rxh5 Rxh5 45.Qe1+ Kb2 46.Qe2+ Rc2 47.Qxh5 Kxa2 48.Qd5 (but not 48.Qa5+? Kxb3 49.Qxa6 with a theoretical draw). The final position is apparently won, but still this path to the goal does not look like the most convincing one.

38.Rxd6+! Ke5 (38...Ke3 39.Qf2+ Ke4 40.Rd4+ Ke5 41.Qf4+ and so on) 39.Qd5+ Kf4 40.Qd4+ (40.Rf6+ Ke3 41.Re6+ Kf4 42.Qd4+ is also strong) 40...Kg3 41.Qf2+ Kg4 42.Rd4+ Kg5 43.Qf4+ Kg6 44.Rd6+ Kh7 45.Rd7+ Kg6 46.Qf7+ Kg5 47.Qe7+ Kg4 48.Rd4+ Kf5 49.Rd5+ Kg4 50.Qxb4+.

We do not have to calculate this variation precisely to the end, as in the course of things White had some good alternative options. It is enough just to convince ourselves that a given method of conducting the attack is the strongest.

The problem. As a rule, when you offer an exercise it is enough to indicate whose move it is, and the problem will be making a decision for the player whose on move, finding the best option for them.

But sometimes, especially when examining complex positions, you can also use artificial problems.

Lipnitsky – Smyslov
Moscow, 1950 (variation from the game)

[FEN "r4rk1/pp1Nbppp/2n1p3/3p4/3PnB1P/P1NQP3/1q3PP1/RB2K2R w KQ - 0 15"]

1.?

The strongest continuation is 15.Nxf8!, but the variations that prove this are not very engaging. Another possibility – taking with the knight on e4 – leads to interesting complications that are difficult to calculate. So here it makes sense to set the problem: "Evaluate the consequences of 15.Nxe4."

15.Nc3xe4!? d5xe4

A forced reply. We can establish that after 16.Qxe4 g6! White is in no condition to hold onto his big material advantage. You have to postpone taking the pawn and choose an unexpected zwischenzug.

16.Ra1-a2!!

The following variations arise:

16...Qxa2 (16...Qc1+? 17.Qd1) 17.Qxe4 f5 18.Bxa2 fe 19.Bxe6+ (or first 19.Nxf8), with White a pawn up.

16....Be7-b4+! 17.Ke1-d1!

Of course, not 17.ab? Qxb4+ 18.Qd2 Qxb1+.

17...Qb2xa2 18.Bb1xa2

Now 18.Qxe4?? is impossible here because of 18...Qd2#.

18...e4xd3 19.a3xb4 Rf8-d8 20.Nd7-c5 Nc6xb4 21.Ba2-b1

(or 21.Bb3), and with two pieces for a rook White retains real winning chances (analysis by Isaac Lipnitsky).

By changing the problem we can make the solution to the exercise easier or more difficult, shifting the accent from certain problems to other ones.

Kunneman – NN
Berlin, 1934

[FEN "6k1/pp3p1p/2p3pB/3rb3/8/2P2QP1/Pq3P1P/4R1K1 w - - 0 1"]

The "zinger" in this position is that on 1.Qf6 (with the idea after 1...Qxc3 of giving mate by means of 2.Qg7+! Bxg7 3.Re8+) Black finds a beautiful way to save himself: 1...Qc1!! 2.Qxe5 Qxh6, with equal chances.

If in the problem the request is to evaluate the consequences of 1.Qf6, we make it considerably easier to find the answer, as the solver thereby concentrates their attention on the search for double-edged combinational blows. On the other hand, if we suggest just making a decision in the initial position, we thus set a very difficult problem: you have to overcome the strong temptation to "win the game" immediately by means of 1.Qf6?!, convincing yourself that neither 1.Rxe5? Rxe5 2.Qf6 Qb1+ not 2.c4? Rd8 give you anything, and only then choose the best of the quiet moves that strengthen White's attack.

In my old article two candidate moves were mentioned: 1.Kg2 and 1.Qe3, and preference was given to the latter. Later, having convinced myself that Black could hold the defense here, I analyzed the position again, and, in the end, reached a completely non-obvious solution: 1.g4!? You can see the details in the chapter "Can the problem be solved?" in my book School of Chess Excellence 2: Tactical Play.

After he read the book, the German analyst Claus Dieter Meyer subjected the position to detailed computer analysis. It confirmed almost all my conclusions, except for the most important one: it became clear that after 1.g4 Black could also count on saving himself with precise defending. So the problem turned out to be insoluble.

In conclusion, I would like to point out that a good support for independent training can be a book of problems in which the exercises are sorted not by tactical method, as is usually done, but by the skills for conducting the battle that are being trained.

My book Remember Your Opponent, published in Russian and German, was designed according to this principle – and I hope it will come out in an English edition. I can also mention Jacob Aagaard's book Grandmaster Preparation: Calculation, and for players with a relatively low rating, Artur Yusupov's series of books.