How do you generalize tactics?

With the ai review it is easy to see the mistakes you made and what variation would have worked, but the exact same situation is unlikely to show up again (unless it’s mistakes in joseki, but let’s assume you are practiced enough to not make mistakes there). How do you distill the essence of the tactics so you can do something similar in other games? Or do you think that is not possible?




Generally we get around it with proverbs and pattern recognition, involving common shapes and liberty counts, as well as practicing a lot of tsumego (not necessarily limited to L&D tho)

Personally I would recommend (as Kaworu is doing) to first think of “What would I like to get out of this situation.” Of course, this should be incredibly obvious upon being restated as “What is your strategic goal?” as it is plainly obvious that your tactical goal should be influenced by your strategic goal (and your strategic goal by your possible tactical goals).

Take it down to bare essentials: If your goal is to cut, then you start reading ways to cut, probably drawing on various tesujis to ensure the cut; if it’s to connect, then start reading those ways, also drawing on various tesujis to ensure it. If your goal is impossible, you should reevaluate your strategic situation based on that, and then look for a new tactical goal. Same with living, running, killing, etc.

The Clossi Method that Kaworu posted is considered a very solid strategic framework, although I’d argue it misses the most fundamental goal of “What options do I have for winning the game?”, but that’s likely intentional so as to prevent weaker players (of which I believe I fall into the category of) from making big leaps in judgement that could cause them to lose the game while trying to win it.

But if what you want is tactics, that’s the primary aim of tsumego: to give you familiarity with common shapes to make reading easier (as well as help you hone your reading), and thus helping compress the tactical situation into more manageable forms.