Thank you. I believe we came much to the same conclusions, but your way of putting it helped clear things up for me. See if you agree with these, please:
1. There aren't pure tactical moves in a unsolved game: we don't have the true winning probability for any move and can only estimate it with the help of strategy.
2. Games are always more tactical than the real world, because the restrictions that define them always reduce the search space (reduce entropy, as you put it).
3. You can approach chess (and life!) thinking that "tactics are the servants of strategy", as Mikhail Botvinnik said, or the opposite. In the end, you're going to need a good command of both, and a developed intuition of when to use each, to reach your goals in situations with imperfect information (any interesting situation). While a perfectionist may argue that striving for the impossible exactitude of tactics is where the beauty really is, and a more creative player may say that the impossible complexity of strategy is magical, they just might be thinking of the exact same thing.
About the third paragraph, do you mean the part about Deep Learning game engines? You got me there that I was rambling, because I used to work with Machine Learning but haven't in a while, and I didn't study these engines' algorithms' in-depth before making my comments. Still, I think that analyzing what they are doing can be very useful for this kind of discussion. A kind of empirical test to what we're maybe trying to understand philosophically. Although technical, this paper about AlphaGo explains these experiments better than I could: https://deepmind.com/research/publications/2019/mastering-game-go-without-human-knowledge
Hello, everyone. This is my first post here. I apologize if this idea is obvious, but I've been thinking about it for a while and would appreciate some outside input (which, I guess, is me doing strategy instead of tactics right now).
First, what I understand is the definition of the two central terms in this discussion, from the perspective of a player during a game of chess and considering only valid, in-game moves (i.e. we're not considering real world aspects, such as tiring out an opponent with less stamina, etc.):
- Tactics: picking the next move with highest probability of winning in a given position;
- Strategy: picking the move or sequence of moves that will guide the board to a position where you are more familiar with the probabilities, and can from there more easily find the correct tactics (e.g. strategically picking an opening, because there too many moves with very similar winning probabilities at this point).
From the way I defined those, you can already see what I'm trying to say: strategy in only a way to fill the holes until you get to a level of complexity where you're able to think tactically. Which is in agreement with the people who say that chess is 90% tactics and 10% strategy (given limited computation capabilities, because with infinite computation it would, as most things, be solved).
This, I believe, is also the way modern Deep Learning engines approach chess. Their approach, in my understanding, is mostly tactical, using probabilities given by a neural network, but sees a big (about 10%?) improvement when adding tree-based search as a strategy, according to the definition above, to simplify the search space when it gets too complex.
An idea that follows from that is that chess, when considering only valid, in-game moves, and given the average human's capacity for pattern recognition and probability computation, is mostly a specialized training for tactical thinking, because its complexity is low enough for us (and our current computers) to approach it in this 90%-10% fashion. While the real world, where complexity is just infinitely higher and you can strategically influence starting conditions, desired outcomes and sometimes the rules, is much more strategic. Thus, a game with higher search-space complexity, that therefore forces us to use more strategy (e.g go), is more applicable to the real world? Or is real-world strategy so complex that a lifetime of tactical training with chess is more effective given a person's limited lifespan?
Also, is real-world chess, with all the aspects that I decided to ignore (stamina, how you decide to train, bad-faith actions, mindset, etc.) already much more strategically balanced than 90%-10%?
In addition, I'm aware that my definition of tactics above may be overextended so as to be encompassing aspects that some would consider as part of strategy. Therefore, feel free to discuss these starting definitions, if you find that is the best way to make the conversation more fruitful.
Thank you to anyone who took the time to read so far.
Best,
Marcelo