Data-Driven Thinking -- Applying The Chicago Approach to Chess

Data-Driven Thinking -- Applying The Chicago Approach to Chess


Imagine you have reached the following position in a game. It's black to move.

How should you proceed? How can you formulate a plan? Which moves should you consider?

In this article, I am going to describe a new method, inspired by The Chicago Approach (Details), of thinking about such positions.

The Chicago Approach is a structured process to generate, refine, and apply ideas in a given situation. It consists of two steps. In the first step, we evaluate the situation by asking different kinds of questions. The answer to each question forms the basis of a potential idea. In the second step, we assess the value of the generated ideas. We evaluate ideas using data, rather than intuition. We then select the most useful idea and apply it to the given situation. The following figure illustrates these steps.

The Chicago Approach Step 1 - Generate ideas through relentless inquiry

Consider the position introduced above.

To generate ideas (or possible plans), we need to ask different kinds of questions. Each question is based on a lens (or perspective) that provides a different way of looking at the position. The number and variety of questions that you ask will determine the quality of the generated ideas. In the position above, we can ask the following:

  • Material Lens
    • What is the material balance?
    • Are there any imbalances? such as bishop vs. knight? 
    • ... 
  • Tactical Lens
    • Is either king vulnerable? 
    • Which pieces are unprotected? 
    • Which pieces are protected by only one other piece? 
    • ... 
  • Positional Lens
    • Which squares are weak? 
    • Which pieces can I improve? 
    • Does my opponent have an isolated pawn? 
    • What can I attack in the opponent's half of the board? Can I attack his king or a weak pawn? 
    • ...
  • Prophylactic Lens
    • What is my opponent's plan? 

The answer to each question forms the basis for an idea. Consider the following answers (and potential ideas) to the questions listed above:

  • Material Lens
    • Both sides have the same material. 
  • Tactical Lens
    • White's king is vulnerable. The g-file is open. Black's king seems safe. 
    • The white knight on b3 and the white bishop on f3 are both protected only by the queen on d1. There might be a tactic to exploit this fact.  
    • The black bishop on a6 is unprotected. White might be able to plan a tactic around this. 
  • Positional Lens
    • The square on e4 is somewhat weak. The Black knight can occupy this square. 
    • The white squares around the black king are weak. If Black can get rid of the light-squared bishops, then Black's pieces can occupy those weak squares. 
    • The Black knight on e7 is not well-positioned. We should improve its position. 
    • The Black rook on f8 is not involved in the game. We should improve its position. 
  • Prophylactic Lens
    • White would like to get rid of the doubled pawns on the f-file. One way for white to achieve this is to play f5. 

At the end of the first step, we have a list of ideas. Not each one will be useful, and some could even be trivial. The more ideas we generate, the more likely it is that one will work in the given position. Therefore, it is essential to ask as many questions as possible.

The Chicago Approach Step 2 - Evaluate the generated ideas using data

After completing step one, we have a list of ideas (or plans). How should we rank these ideas to select the most optimal one to apply to the given position?

We measure the usefulness of an idea using data, rather than intuition. In chess, we have two sources of data available to evaluate an idea. The first is to calculate the concrete variation associated with the idea. The second is to recall past games where this idea was successfully used in similar positions. 

When evaluating an idea, we must first try and calculate the corresponding variation to sufficient depth. I will not go into strategies for calculating in this blog since there are several excellent books on the subject. We can reject most ideas by calculating the corresponding variation for a couple of moves.

However, for several ideas, it is not always possible to calculate the corresponding variation. For instance, for the positional idea of trading light-squared bishops and occupying the weakened squares, it is difficult to calculate all variations. To evaluate such ideas, we should recall past games where similar plans were successfully used. If we cannot recall a single game where an idea has worked in the past, then we should consider alternative plans. 

More generally, our ability to assess an idea depends on the quality of available data. The quality, in turn, is a function of the type of idea we are evaluating. To accurately assess a tactical idea, we should improve our ability to calculate variations. On the other hand, to better evaluate a positional idea, we need to increase our knowledge of typical games, positions, and patterns. 

In our example, black uses the idea of trading light-squared bishops and occupying the weakened light squares with their pieces. Look at the comments in the solution to the puzzle for more details.  


For a given position, we follow two steps.

  1. We generate ideas by asking different kinds of questions. In this step, we want to generate as many ideas as possible, regardless of quality.
  2. We evaluate ideas using data.
    • For some ideas (usually tactical), data-driven evaluation implies calculating the corresponding variation to sufficient depth.
    • For other ideas (mostly positional), data-driven evaluation involves recalling past games (having similar positions) where related ideas were successfully used.

Therefore, to become a better chess player, we should improve both our ability to generate ideas (by looking at a position in different ways) and our ability to evaluate ideas (calculation skills + knowledge of typical patterns).