Intuition and the Sacrifice

How do you know if a sacrifice is correct or incorrect? In the absence of a calculable forced line leading to mate or winning back the material – how can you assess whether the compensation is enough?

Let’s look at the following position:

White has two bishops and a (very) temporary lead in development. However, in a stonewall-type of position, the bishops might not be especially relevant. How could White open up the position?

The pawn advance c3-c4 would be hard to arrange because of the pressure on d4. Playing f4-f5 seems to achieve more, but right now that square is under Black’s control. Preparing it with g4 does not look too appetizing. White will regret the weaknesses in front of his king. A strong answer to 13.g4 would be 13…h5. Then in the event of f4-f5, not one but two files will be opened in front of the white king, and the black king will be spirited away to safety on the queenside.

Do you see the idea of the immediate 13.f5? After 13…gxf5, White sacrifices the bishop with 14.Bxf5. Lines will be opened and the black king will be caught in the center. But White will be down a piece for only one pawn. Will he have enough compensation?

If Black doesn’t take the piece (i.e. after 14.Bxf5 he plays 14…0-0-0), I am sure you can appreciate the value for White in opening the f-file and liberating the dark-squared bishop. But for such a modest positional gain is it worth risking the sacrifice of a whole piece? If White is wrong, he might be lost – but if he is right (and Black doesn’t take the bishop) he will gain only a pleasant advantage.

I am sure there are many who would not bother to risk calculating the sacrifice. It is quite possible the gains of sacrificing the piece, if Black doesn’t take it, could be less than the risks. Nevertheless, chess is not a game of probabilities like this, like poker. You must try to play the best move, even if it is risky. If you do not, you may avoid some risk but you will ultimately be punished.

The 1999 champion of Yugoslavia, GM Miroslav Tosic, didn’t avoid the sacrifice. He played 13.f5! in the game, followed by 14.Bxf5. I would guess he played it pretty quickly too. It is not hard for a GM to assess the position after 15.Qxf5. And the declining of the sacrifice was hardly worth thinking about. Clearly the change in structure would be greatly to White’s advantage.

I met Tosic shortly after I first came to Serbia in March of this year. Later he invited me to visit his town in the south of Serbia called Svrlig. There I played two simuls against local players and an exhibition match of rapid chess against him. Although it is a very small town (somewhere around 5,000 people, I think) the audience for that match was quite large.  It is hard to convey how fantastic was this trip to a very strange rural place in the mountains of southern Serbia, and I will always remember it.

Now, let’s see Miroslav’s game against GM Vladimir Raicevic.

The key position, of course, was before 13.f5. How exactly can one assess the position after the sacrifice? How can one decide if White has enough compensation, and what exactly does "compensation" mean? After all, logically if the game continues on and on, and White neither wins back his material nor checkmates his opponent (or gives perpetual check), then he should lose. So does "compensation" mean that ultimately - even if no human or even computer is capable of calculating it - White will win back his material or give checkmate? Does "enough compensation" mean that, in the event of perfect play by both sides - White will regain the material to reach an equal position, or force perpetual check? 

I suppose that theoretically the above reasoning is correct. However, in the typical position after such a sacrifice, it is impossible for humans to calculate to the point where the 'compensation' becomes actualized. In most such positions, computers also cannot calculate far enough to determine how real the compensation is - and let's hope that it remains that way!

This means we need to look with our eyes and judge for ourselves whether the compensation will be enough; whether some time - maybe ten, twenty or more moves from now we will be able to convert the advantages that our material sacrifice gave us into something more concrete. It is not an exact science. One thing to note is that the postion in the above game after the sacrifice cannot be accurately evaluated without calculating a number of short variations emanating from it. We can all see that White has better development and more active pieces aimed at the black king, which will crush him if Black cannot escape. The question is - can Black escape with the king or somehow develop the pieces to protect it? To answer the question requires calculation of short variations - 15...Be7, 15...Qd7, 15...Nd8, and 15...Bd6.

Let's compare this to the following position, from an article I wrote a few weeks ago about my game against IM Zelko Bogut:

The sacrifice he made prior to this, it appears, was not quite correct - i.e. White probably did not have sufficient compensation. This means that in the event of perfect play by Black, I would end up regaining the initiative while keeping some of my material advantage (or convert the material advantage into some other advantage). However, to reach that point would be very difficult, so in practice he had almost enough compensation. This can be compared to the Tosic game, where White clearly had more than enough compensation. It requires very fine judgement to distinguish between these. The key point, though, is that an evaluation cannot be made without looking at the future possibilities for the players.