Since every deal here is solvable, a failed board is always a sequence problem. The tiles were removable; you just removed them in an order that stranded something. Good play is about the order.
Rule one: work from the top down
Tiles on upper layers block everything beneath them, and there are far fewer of them. Removing a top-layer tile typically frees one or two below; removing a bottom-row tile at the edge frees nothing at all.
On the turtle, the single tile at the peak and the four beneath it should go early. They unlock the whole central stack, which is where most late-game deadlocks happen.
Rule two: count what each move unlocks
Before taking a pair, look at what sits underneath and beside each tile. A pair that frees four new tiles is worth far more than a pair that frees none, even if both are equally available now.
The tiles at the far left and far right of the turtle — the three sticking out in the middle row — are the classic example. They are always free, so they can be taken at any time. Taking them early wastes the flexibility; they are best saved as a spare move when you need one.
Free tiles that stay free are a resource
A tile that nothing can block is a move you can make at any point. Keep at least one such pair in reserve rather than spending them early.
Rule three: handle the fours carefully
Most tiles exist in sets of four. If all four are free at once, take all four — two pairs, no risk. The danger case is two free and two buried in different places.
If you take the two free ones, the remaining two must both become free later, in different parts of the board. If instead you wait, you might match a free one with one that has just been uncovered. So when only two of a four are available, ask where the other two are. If they are both deep in the stack, taking the free pair now is usually right; if one is about to be uncovered, waiting is better.
Rule four: prefer pairs from crowded areas
Tiles in dense regions are the ones most likely to become blocked later. Tiles in sparse areas tend to stay accessible. Given a choice between an easy pair at the edge and an awkward one in the middle, take the middle one — the edge pair will probably still be there in ten moves.
Rule five: use the pairs counter as a warning light
The readout shows how many playable pairs exist. Early on this is typically ten or more. When it drops to three or four, the board is getting tight and each choice matters much more.
At that point, stop playing quickly. Look at every available pair and work out what each one unblocks before committing.
Thinking one move ahead
Full lookahead is impractical, but one move is achievable and it catches most errors. Before taking a pair, ask: after this, what pairs will be available? If the answer is 'fewer than now, and none of them unlock anything', look for a different move.
When to use the tools
- Hint when you cannot see any pair — it also tells you how many exist, which is the more useful information.
- Undo the moment you realise a move stranded something, rather than playing on hopefully.
- Shuffle only when there are genuinely no pairs. Three per board is not many.