Learning Techniques for Go

This article focuses on learning techniques based around Bloom’s taxonomy and how they apply to Go. A lot of the material is taken from Justin Sung’s YouTube channels: “Justin Sung” and “iCanStudy”.

Bloom’s taxonomy and encoding

Here is a short explanation of the taxonomy layers, taken from Colorado College’s page on “Bloom’s Revised Taxonomy”.

  • Remember: You can retrieve relevant knowledge from long-term memory.
  • Understand: You can explain something.
  • Apply: You can use the knowledge in a new situation.
  • Analyze: You can break the material into its constituent parts and determine how the parts relate to one another and/or to an overall structure or purpose.
  • Evaluate: You can make judgments based on criteria and standards.
  • Create: You can put elements together to form a new coherent or functional whole.

Spaced-repetition software (SRS), which is commonly used when learning kanji or facts or solving tsumego, only covers the lower layers.

SRS focusses more on the information retrieval aspect, not on the information encoding aspect. SRS is better than doing nothing at all but it’s not the goal, it’s a stepping stone. Just because you understand something doesn’t mean you are able to do something with it.

There is a famous quote: “What I hear, I forget. What I see, I remember. What I do, I understand.” —Xunzi (340 – 245 BC), Confucian scholar.

Information in isolation is forgotten. Information with lots of relationships is held onto. When you try to understand information, you are still seeing it in isolation. You still only retain it while you study it.

“Skill” is the ability to apply the knowledge. You need game experience for that.

Deliberate practice, mindful training—analysing and dissecting concepts and sequences—helps to “encode” knowledge into the long-term memory. Once you have internalized something you won’t forget it—that is, you don’t need to repeatedly recall it to short-term memory.

To train deliberately, don’t look at positions in isolation; relate them to similar positions and other concepts, to a big picture. This gives your brain the chance to figure out how to organize the materials. The brain is good at structuring and organizing if you give it sufficient material to work with (“cognitive load”).

The “apply” layer is the beginning of higher-order learning, where you are focussing on the relationships. To solve a problem you need to relate the information to other pieces of information.

When you “analyze” you are combining and comparing different concepts with each other. You find similarities and differences; you group them together; you break them apart and see how the parts relate to each other.

When you “evaluate”, you don’t just compare concepts; you rank and prioritize them. You decide what is the most important in the current situation or in different contexts.

To “create” is the highest step; there you create new knowledge; you use your deeply connected understanding of the concepts in order to theorize about possibilities.

If you operate on a higher layer of the taxonomy you will automatically get the lower layers for free. When you “apply”, you also “remember” and “understand” faster and more deeply, moreso than if you just tried to “remember” or “understand” in isolation.

Your brain is focussing on whatever you are framing it to focus on. If you start learning something in order to solve a problem—”applying”—then everything you “remember” and “understand” is framed in a way that is related to the problem, so it already creates relationships. That is more efficient than starting by “remembering” first, next “understanding” in isolation and only then putting it together by “applying”. Doing the first two steps separately is redundant and takes a lot of time. You can still fill in the gaps in your knowledge later by memorizing and understanding the missing pieces.

Phrasing it in terms of the taxonomy layers. you could use SRS to “remember” and “understand” some joseki but then fail to “apply” it in an actual game or don’t know what to do when the opponent deviates from your known sequences. Instead you should actively “analyze” the sequences, see how they relate to each other, and also “evaluate” them, that is, contrast and compare them and know when to choose which sequence.

David Bull says something similar in “David Bull on the ten-thousand hours rule”.

Another way to improve the process is to increase the “cognitive load capacity”—that is, how much information you can give the brain to work with. If you immediately write down your conclusions you ease the burden; you empty your mind again. But this is not necessarily a good thing. You have a record of your thoughts but you are not encoding them.

Instead, hold onto that information in your mind, consume more and ask yourself how this is related to that; how it fits into the big picture. By gradually increasing the amount of information you hold onto before getting to conclusions, you can increase the cognitive load tolerance.

If you have trouble with the higher layers like evaluation—comparing, constrasting, prioritizing—and analyzing, start with a lower layer first, like applying, then work your way up. Applying is still better than just memorizing. But also realize that the overwhelming confusion is fuel for growth. Confusion is a sign of higher intrinsic cognitive load, which promotes higher-quality encoding.


When you are faced with a large amount of information, first create a mental schema, like a relational database schema. Don’t put everything in one big pile but store the information as relations. For Go, this is already easier because we can classifiy sequences and concepts using terms like “opening”, “middle game”, “endgame”, “attack and defense”, “invasion”, “reduction”, “joseki”, “tsumego” etc. If the individual categories are still too big, subdivide them. For example, you can subdivide “tsumego” into “corner tsumego” and “side tsumego” or “stones on the second line”.

This process of finding meaningful relationships and similarities is called “chunking”. Chunking refers to the process of taking individual pieces of information and grouping them into larger units. By grouping each data point into a larger whole, you can improve the amount of information you can remember. You don’t look at individual stones but at patterns. Also, you don’t isolate, for example, individual shapes on the second line but consider all such positions. You will then be able to generalize things like “on the second line, six stones are dead, seven are unsettled and eight are alive”.

The process of continuous chunking is engaging higher-order learning, which facilitates better encoding.

Creating a good structure of big-picture chunks and relationships is the hardest part of learning but you should do this first. Once your structure is stable you can add new information and consolidate ideas a lot faster.

Most of the chunks mentioned above are universal for Go players but be aware that your own structures and notes will be easier to understand for you and not others because you did the active process of thinking about the material and encoding it.


I still believe in the traditional pro advice of repeatedly solving relatively easy problems. That is, SRS can be useful in drilling. But not in understanding, relating and analyzing new concepts.

Also I believe that most of the above is common sense; of course you will internalize knowledge better if you don’t just repeat it mindlessly.

Use syntopical reading, that is, to get information from multiple sources talking about the same topic. This allows you to build a more robust foundation because you don’t just learn linearly. Imagine different Go teachers explaining the same or similar concepts using different words and examples.