Isn't the solver algorithm too complicated? about kaboom

[chdoc]

Additionally, there is one interesting implication of the current setup: in the case of guessing, the player is able to influence the game somewhat (by choosing a field and forcing it to be empty). This is what makes it truly different from a normal "pre-determined" minesweeper. There is no hidden state anymore, and you could argue the game becomes a complete information game (with human and computer as players), like chess or go.

Indeed, this is what currently motivates me to come back to this game again and again: I'm trying to understand what's a good strategy for forcing empty fields. Due to the complete information, this is actually more a psychology question than an algorithms question. I find the SAT solver to be much more reliable than me when doing "parity" arguments along long boundary regions. So I try to force fields that allow me to subsequently apply simple reasoning steps.

[pwmarcz]

With your proposal, wouldn't I need to know what are the "fields the player can/cannot know are empty"? And to determine that, I need some kind of reasoning engine anyway, even if it's brute-forcing all the possibilities. So it wouldn't be so different.

Additionally, there is one interesting implication of the current setup: in the case of guessing, the player is able to influence the game somewhat (by choosing a field and forcing it to be empty). This is what makes it truly different from a normal "pre-determined" minesweeper. There is no hidden state anymore, and you could argue the game becomes a complete information game (with human and computer as players), like chess or go.

I think the main source of complication is 1. treating boundary and non-boundary differently, 2. caching past results. These are performance optimizations, and perhaps with a different algorithm they would not be necessary.