How to Use Engines Effectively

How to Use Engines Effectively

Illingworth
GM Illingworth
|
5

A common discussion point in chess circles is: What can we learn from chess engines?


I have already shared some ideas on this in my recent post on How to Find the Best Opening Moves.


However, because this was quite a specific theme, I want to share how else engines can help us improve, when we use them the right way.



I'm aware that there are some who advise against using chess engines at all. This argument generally goes along these lines:


1) Engines don't understand anything about chess.


2) Because they don't use words, only moves and evaluations, humans don't have a way to derive meaning from these moves and evaluations. 


Therefore, engines are useless in helping humans understand a position.



For what it's worth, I do believe that many players become too reliant on the engine, using it as a 'crutch' or as gospel, just taking the moves and assessments for granted, without trying to understand them. If you watch a chess broadcast with engine analysis live (as I liked to do for over a decade), you will see this for yourself. Here's one example:



When I was watching this game, and this position was reached, most of the spectators assumed that this was a trivial win for White. Not just because of White's two extra pawns (which are both doubled on the h-file), but because the engine was giving extremely high evaluations - higher than +6 - for White.

However, those who have analysed endgames with the engine will know that the assessments should be taken with a grain of salt, because '+1.27', for instance, does not tell us whether the endgame is a win or a draw. 


It turns out the engine's highly positive assessment for White is justified here - but what the engine won't tell you directly is that White has to find exactly the right idea here to win. Do you see it?


Well done if you solved the puzzle! However, do you understand why the solution wins? And, why Carlsen's move, played in the game, only draws? 

We'll test your understanding of the second question in the next puzzle:


The engine still claims White is winning after 46.h6, which confused a lot of the spectators watching this game live. However, those who know their rook endgame theory will have recognised the Vancura Position in the above game, and possibly even found the way to reach it, as Aronian did. 

Now let's see why 46.Kc2! wins. It turns out the key to White's victory lies in zugzwang:


From this example, you can see that we had to analyse the position ourselves in some depth, to make sense of the play, and understand White's winning plan:

a) First, prevent the Vancura Position by tying up Black's rook.

b) Force Black into Zugzwang by pushing the h-pawn and a-pawn...

c) So that ...Kg8 can be met with Rg7+ and a7, decisively freeing White's rook from being trapped in front of his a-pawn. 

d) Don't fall for Black's stalemate trap!


When using engines, we should continue our analysis until the position has clarified (no major tactics). To understand the tactical point of a move, we only need to ask the natural question 'What happens after this (obvious) move?' to eventually see what the tactical idea is. Let's take such a dynamic example.


It is Black to move in this GM game. The engine gives 0.00, but as I mentioned in the last article, such positions will often be easier for one side to play. In this case, the onus is on Black to defend against White's attack. How would you do so? 

There are a few ways, but the clearest is 30...Rd8!, forcing White's hand, as 31.Qxd8 Qc1 32.Kh2 Qf4 33.Kh3 Qg4 34.Kh2 Qf4 35.g3 Qxf2 is a clear draw by repetition. If White tries to play on with 31.Qe2, Black will free his king with 31...h6, for at least equality. 

In the game, Black was in serious time trouble, and blundered with the natural but incorrect 30...Re7?. Can you see how White exploits this slip?


You may have found the winning move, but to understand why it works, you need to have seen White's key trick after Black's natural move 31...h6, trying to free Black's king and nullify the Rg8+/Qd8 tactic. What is the move for White?


You probably understood the main point of this example already, but I'll explain it anyway. If we had gone through this game with the computer, not asking questions, but just taking the engine's word for everything, we would have seen that 31.Ra8! wins, but not understood why, or turned this into a lesson that makes us a stronger player.





Through a deeper investigation, we have learned a very clever tactical motif, and reminded ourselves not to make assumptions, but to consider the forcing moves, and whether we can ignore the opponent's threat after all. 

To test if you fully grasped the above ideas for White, let's see how you go with defending as Black, after the move White played in the game, 31.g3?:


The above examples were quite complex, as proven by the fact even strong Grandmasters did not play perfectly in the game. So, for the benefit of less experienced players, let's demonstrate these methods in a much simpler form:


I doubt it will surprise many people to learn that White is clearly better in this position. At depth 35, Stockfish gives an evaluation of +1.10 for White. But why is that the case? 

We can often make sense of an evaluation, where there are no major tactics or forcing moves, by considering the strategic factors in the position. More specifically - what advantages do both sides have, and how significant are they? I have found that Stockfish's assessments are mostly quite reliable, so here I will explain why Stockfish gives such an assessment. 



I don't see any advantages for Black in this position, so let's consider what White's advantages are here:

- Bishop pair advantage (White has two bishops, Black has only one remaining)

- White has a clear lead in development (currently 2 to 0)

- White has the initiative (mate threat on f7)

The engine's evaluation did seem a bit optimistic to me at first - what if Black just defends the pawn with 5...Nf6That leads to the puzzle position below:


What are the main things you learned from this post? 


The key to using engines effectively is to: (example questions from the first game, Carlsen-Aronian)

a) Ask questions (what is the difference between 46.Kc2 and 46.h6?)

b) Consider alternative moves (what if Black tries to check from behind?)

c) Understand why a move works (why is the Vancura Position a fortress?), and

d) Understand why a position is winning/better/equal (why is the position after 51.h5 Zugzwang?)


This process does take some thinking and effort, but I hope that these examples have convinced you that it is well worth the extra discipline  

What's a lesson you learned from a game or position you analysed with the engine? Share it in the comments below!

You can find more of my puzzles and content at:


If you found this post helpful in directing your chess improvement, and would like more specific and personal direction toward achieving your chess dream, applications are open to become a private student of mine:


I'll see you in the next post!