先找最受限制的位置
選擇最少的格子、線段、容器或路徑最適合先處理，因為錯誤分支較少。
區分確定與猜測
把已由規則證明的資訊和暫時假設分開。遊戲支援標記時，使用叉號、旗幟或候選記號。
每一步都重新掃描
一個確定動作常會改變多個交叉條件，完成後要重新檢查整盤，而不是只盯著同一區。
使用矛盾反證
若某個假設會使數量超標、路徑斷裂或留下無法完成的區域，該假設就可以排除。
分解大問題
把棋盤拆成邊角、封閉區與連接區，先解獨立的小區域，再整合全局。
先建立一套自己的標記習慣
邏輯類謎題的難度多半不在推理本身，而在於「我剛才是怎麼推出這一步的」很快就忘了。固定使用一組符號：確定為真用實心標記、確定為否用叉、暫時假設用問號或淺色。標記一致之後，回頭檢查會快得多，而且能立刻看出哪些結論是證明來的、哪些只是假設。
卡住時的四個提問
- 整盤最受限制的那個位置是哪裡？（選項最少的地方通常還有沒榨乾的資訊）
- 有沒有哪個數量或總和已經接近上限？逼近邊界的條件最容易產生確定結論。
- 我有沒有把某個「假設」誤當成「已證明」在用？
- 有沒有整個區域從剛才起就沒再看過？
假設法要怎麼用才安全
當純推理走不下去時，選一個只有兩種可能的格子做假設，然後往下推到出現矛盾或明顯完成為止。關鍵是先把當前盤面記下來（截圖或另存），否則推錯之後很難回到分岔點。每次只做一層假設；套疊兩層以上時，人腦幾乎一定會弄混哪一步屬於哪個分支。
Start with the most constrained position
Cells, edges, containers or paths with the fewest choices are best to solve first because fewer branches can be wrong.
Separate certainty from assumption
Keep proven information distinct from temporary hypotheses. Use crosses, flags or candidate marks when the game supports them.
Rescan after every move
One confirmed action often changes several crossing constraints, so review the whole board rather than staring at one area.
Use contradiction
If an assumption exceeds a count, breaks a path or leaves an impossible region, that assumption can be rejected.
Decompose the problem
Split a large board into corners, closed regions and connectors. Solve independent sections before combining them.
Build a marking habit first
The difficulty in logic puzzles usually is not the reasoning itself but forgetting how you arrived at a step. Settle on one set of symbols: a solid mark for proven true, a cross for proven false, a question mark or light shade for a working assumption. Once the notation is consistent, rechecking becomes much faster, and you can see at a glance which conclusions were proved and which were merely supposed.
Four questions when you are stuck
- Where is the most constrained position on the board? The place with fewest options usually still has unextracted information.
- Is any count or total already close to its limit? Constraints near a boundary produce forced conclusions most readily.
- Have I been using an assumption as though it were proved?
- Is there a region I have not looked at since several moves ago?
Using hypotheses safely
When pure deduction stalls, pick a cell with exactly two possibilities, assume one, and follow it until a contradiction or a clear completion appears. The essential step is recording the current board first — a screenshot or a save — because returning to the branch point is otherwise difficult. Work one level of assumption at a time; nested hypotheses are where human players reliably lose track of which deduction belongs to which branch.