Last week I posted a try to find and organize checkmate patterns using a clustering algorithm combined with a focused look on the 3x3 squares around the checkmated King.
Thanks! I haven’t used mathematics since high school, but learning about programming has rekindled my interest :) I have (unsuccessfully) puzzled with tensor flow in a try to make a model that could compose mate in two problems.
Well you are braver than I! This is such a great idea imho as it makes me think about mate problems in a new way. I mean for me at least. I found tensor algebra very useful bitd in physics but have no facility with it these days sadly.
HI Martin, your work has been amazing, I am sure you can write a thesis on Chess. As per the title is concerned, "Recurring Checkmates" would be short and sweet. attracting both the beginner and advanced. we already have books on the types of checkmates but your focus is on what is most repeating with the graph that you have shared.
My guess is there are many more. This is just the clusters I found using the settings I setup. Using other settings and a larger database may find more patterns.
Thanks. This is so interesting. If you set the hamming distance to a higher number, you get fewer, larger clusters. I assume at some point there would be clusters that grouped patterns humans would normally say aren’t similar.
Last year I tried to do similar concepts about checkmate patterns, but without the use of AI or any other applications.
Hello Martin!
Your project is pretty interesting! I must say it is a nice idea to try out the new approach & see how much improvement can be made (I would select the Barsky book "A guide of the checkmate patterns" as the perfect one for the present moment).
What I noticed is the big number of checkmate patterns one can create by writing down the code (AI or any other type) or extract from the big dabatase (like lichess 5M puzzles).
However one problem I could not solve was the practical application and use of these patterns.
For example: imagine I created or extracted 100K checkmate patterns of back rank checkmate. Now the question is how many of these patterns and which one should be put into the book or workbook? And the most important criteria: what are the ways to extract only the ones that has significant and practical meaning for the chess players?
That is why I am sceptical about such projects until you have the idea how to extract and group the necessary and practical puzzles of checkmate patterns. I do not mean the project is useless, but rather point out the necessity of solving the issue of the extracting and grouping these patterns that can be really helpful. Why? Because without filtering we can easily extract 500K checkmate patterns, but I guess nobody would be interested in practice of these.
I would be very happy if you reply to my comment and let me/us know how are you going to filter out and group these patterns.
When you get to tensors my heart will explode 😉
Well done.
Thanks! I haven’t used mathematics since high school, but learning about programming has rekindled my interest :) I have (unsuccessfully) puzzled with tensor flow in a try to make a model that could compose mate in two problems.
Well you are braver than I! This is such a great idea imho as it makes me think about mate problems in a new way. I mean for me at least. I found tensor algebra very useful bitd in physics but have no facility with it these days sadly.
HI Martin, your work has been amazing, I am sure you can write a thesis on Chess. As per the title is concerned, "Recurring Checkmates" would be short and sweet. attracting both the beginner and advanced. we already have books on the types of checkmates but your focus is on what is most repeating with the graph that you have shared.
Thanks 🙏🏻 I like the idea for a name!
410 clusters? Does that imply 410 different mating patterns or would we as human players organize some of the clusters into the same pattern?
My guess is there are many more. This is just the clusters I found using the settings I setup. Using other settings and a larger database may find more patterns.
Thanks. This is so interesting. If you set the hamming distance to a higher number, you get fewer, larger clusters. I assume at some point there would be clusters that grouped patterns humans would normally say aren’t similar.
Last year I tried to do similar concepts about checkmate patterns, but without the use of AI or any other applications.
Hello Martin!
Your project is pretty interesting! I must say it is a nice idea to try out the new approach & see how much improvement can be made (I would select the Barsky book "A guide of the checkmate patterns" as the perfect one for the present moment).
What I noticed is the big number of checkmate patterns one can create by writing down the code (AI or any other type) or extract from the big dabatase (like lichess 5M puzzles).
However one problem I could not solve was the practical application and use of these patterns.
For example: imagine I created or extracted 100K checkmate patterns of back rank checkmate. Now the question is how many of these patterns and which one should be put into the book or workbook? And the most important criteria: what are the ways to extract only the ones that has significant and practical meaning for the chess players?
That is why I am sceptical about such projects until you have the idea how to extract and group the necessary and practical puzzles of checkmate patterns. I do not mean the project is useless, but rather point out the necessity of solving the issue of the extracting and grouping these patterns that can be really helpful. Why? Because without filtering we can easily extract 500K checkmate patterns, but I guess nobody would be interested in practice of these.
I would be very happy if you reply to my comment and let me/us know how are you going to filter out and group these patterns.
And the idea of the book titles:
300 most common & extended checkmate patterns
300 basic & extended checkmate patterns
300 most common checkmate patterns
The idea is exactly to find useful patterns that happens often, while also being relatively hard to solve ~ hard to notice.