Recovery
Undo and retry costs: when resetting beats another undo
Use a simple credit model to compare one more undo, a reversible probe, and a complete reset.
A constructed recovery model with explicit credits and costs. It does not describe the undo policy, pricing, or save behavior of any specific game.
On this page
Name the cost before you spend it
An undo is not free even when the button looks free. It may spend a move, a token, time, attention, or the chance to learn what went wrong. A reset is also not free: it removes local progress but can restore a clearer starting state.
For this model, an attempt has three undo credits. One undo reverses one move. One reset also costs one credit, clears the current arrangement, and returns to the starting state. Those numbers describe the example, not every game.
Compare the next five moves, not the last mistake
A local undo is better when one identifiable move caused the problem and the rest of the plan remains sound. A reset is better when several uncertain moves have replaced the original plan or when you can no longer explain why the current state is safe.
The relevant question is not “can I still undo?” but “what is the cheapest way to reach a state I can reason about?” Sometimes the cheapest route is a complete reset.
| Situation | Better action | Why |
|---|---|---|
| One clear mistake and a stable plan | Undo once | Removes the error without losing tested progress. |
| Several unexplained trades | Reset | Restores a state simple enough to reason about. |
| Rule understanding just changed | Reset | The old plan was built on a different rule. |
| Input misfire on a known-good plan | Undo or correct forward | Avoid punishing the plan for an execution slip. |
Four required undos exceed three credits
If reaching the intended branch requires four undos and the attempt has only three, the local repair cannot fit. In that case, reset instead of spending every credit and improvising afterward.
The arithmetic is intentionally blunt. Four required undos exceed three credits, so the plan needs a different branch, not one more optimistic reversal.
Credits: 3 Undos required: 4 Decision: reset Reason: the local plan no longer fits the allowance.
Prefer a reversible probe when the outcome is unknown
When two branches look possible, avoid committing your remaining resources immediately. Look for a move that can be reversed cheaply and reveals one new fact: whether a color becomes accessible, whether a path opens, or whether a piece can settle safely.
Write down what the probe is testing. If it succeeds, continue. If it fails, undo without calling the attempt bad; the probe did its job by removing a wrong assumption.
- Before probing, state the fact you expect to learn.
- Reserve at least one credit for the next uncertainty.
- Do not chain probes after their meaning has become unclear.
Turn a reset into information
Before resetting, write one sentence: what state you want to preserve, which move created doubt, and what you will check earlier next time. Without that sentence, a reset can become a loop.
If you repeatedly reach the same state and feel uncertain, the problem is usually representation rather than reflex. Count the resources, name the top accessible pieces, or draw the remaining choices before continuing.
