Imagine you are a chess teacher and it's Monday morning 9 AM. Your day starts and your first private student comes in. As a teacher, you prepared a lesson about double attack. An interesting question is: why double attack? Why not a lesson about another tactical motif or the endgame? How do you decide as a teacher what to teach your students and why?
Russian grandmaster Igor Bondarevky, coach of former world champion Boris Spassky, made the technique of flashcards popular in the '60s to create personal training programs in chess. He effectively countered the dilemma with this performance analysis tool about what to teach his students. Even 60 years later, this technique is still popular to decide what to focus on next in chess training. Well known chess coaches like Mark Dvoretsky and Adrian Mikhalchishin still use it. How does it work?
Coaches look for critical moments in the games of their pupils - so-called turning points. The coaches take these critical positions in the game where mistakes were made or brilliancies were played and classify them with their didactical skills (e.g. flashcard one is a double attack, flashcard two is a pin and so on). After collecting a few dozen flashcards of different key moments in several games, they look for recurring patterns of themes (e.g. figure 2). If a pupil missed a double attack multiple times during their games, you can imagine it makes sense to create a personal training program where the pupil practices on double attacks during training. At the same time it's possible to look at what's going great and which themes pupils do spot a lot. You could cater to their strengths as a coach.
Fast forward to today and Artififical Intelligence (AI) being able to automatically spot didactical motifs in chess on large scale. You are probably already connecting the dots. What if teachers could have access to a flashcard database with thousands of their pupils brilliancies and mistakes. Just in a matter of minutes, a database where everything is categorized is in the palm of their hands. Learning profiles can be created based on patterns found in thousands of flashcards. And if learning profiles can be created, an AI is also able to recommend training material to improve a pupils game. Just like with recommendation with movies on Netflix or books on Amazon. But the possibilities don't end here.
First, historical analysis on a technique level is possible if you collect games over decades. For example, which age was there a huge increase in the amount of double attacks, pins and discovered attacks you spotted? Imagine having a tool where you can filter on quality of decisions, time spent per move, time control and much more variables.
Second, it's possible to cluster data of players. Imagine being a coach on a school or a club and having access to all learning profiles of all these students. You could look for overlapping themes and group students together more efficiently.
Third, comparing players is possible. How does your development look like in comparison to your friends, competitors or world class players? How does the development of a player in Asia looks like in comparison to a player in Europe? How do two clubs compare to each other on performance in certain themes?
This is exactly what we have created and are working on. We believe with enough support of the chess community, we are able to be a great guide to boost training efforts in chess. We will be able to function as a compass for individuals who want to spent their limited time wisely improving their game.
Cooperation between machines and teacher
Even though ‘robots’ recognize didactical patterns in chess, chess teachers don’t need to start packing their bags and look for another job. Just like with the story of the turtle and the hare in the first article, both have valuable skills and enhance each other when they work together. A robot can offer scale and create thousands of flashcards in just a few seconds. Humans on the other hand can offer detail. Some positions are not clear cut and easy to categorize. Humans are better at the fine details of didactical concepts.
The next time it's Monday morning 9AM, the chess teacher can use the data gathered from their students and the feedback from the program to focus their energy in the right direction. AI can support the chess teacher with analytics possibilities. The human teacher looks at the details and keeps in control what to work on next with their pupils.