Rook & Elite
Hey, I’ve been looking into how classic puzzles can be turned into tactical decision trees—thought it might be a neat angle for us to compare.
Nice idea. Let’s start by listing the puzzle’s steps, then spot where each choice splits into branches. Once we have the nodes, we can attach probabilities and payoff values. Ready to roll it out?
Sounds good. Let’s map the sequence first and then break each fork into its own node. I’ll keep the list simple and we can layer in the numbers once the structure is clear. Let's start.
Give me the sequence, step by step, no fluff. We'll map each fork to a node. Let's start.
Which puzzle do you want to break down?
Let's pick Minesweeper. It’s classic, has clear decision points, and the payoff is immediate—no emotional fluff, just data. Ready?
1. Start: choose first square to click – Node 1
2. Result: square opens and shows a number N – Node 2
3. If N = 0, open all adjacent squares automatically – Node 3a (no decision)
4. If N > 0, decide to either
4a. click another adjacent square – Node 4a
4b. place a flag on a suspected mine – Node 4b
5. For each click in 4a, the result is either
5a. open a safe square (shows number) – Node 5a
5b. hit a mine – Node 5b (game over)
6. For each flag in 4b, update counts on adjacent numbers – Node 6
7. Repeat step 3 for any new zeros, step 4 for any new numbers, until all mines are flagged or the board is cleared – Nodes 7, 8, …
8. End: either all non‑mine squares are revealed (win) or a mine was clicked (lose).
Great, we’ve got the skeleton. Next step: assign probabilities to each branch and add payoff values. Let’s start with node 4 decisions—those are where we actually trade risk for reward. Ready?
For node 4 the trade‑off is straightforward.
- 4a (click a neighbour)
– Probability it’s safe = (remaining safe squares)/(remaining unknown squares).
– Payoff if safe = +1 (you gain a revealed number).
– Payoff if mine = –10 (game over).
- 4b (flag a suspected mine)
– Probability it’s a mine = (remaining mines)/(remaining unknown squares).
– Payoff for a correct flag = 0 (no extra gain, but you remove that risk).
– Payoff for a wrong flag = –1 (wastes a flag, doesn’t help).
You can compute the probabilities from the current board state: count how many unknown squares remain, how many mines are still unflagged, and what the adjacent numbers say. Once you have those figures, plug them into the payoff formulas above. That gives you a weighted expected value for each choice, so you can pick the branch with the higher expected reward.