A hint system is a negotiation between two goals. The player asking for help wants to make progress. The designer wants to preserve the moment of realisation that makes the game worth playing. Most hint systems fail by picking one side.

The system I ended up with is graded. The first hint is about the route, not the word: it tells you which move you are stuck on and what kind of change that move probably involves. The second level narrows the target position. The third gives you a letter. You have to ask for each one, and the game records that you did.

The important design choice is that no level reveals the answer. At the deepest level you are still solving; you have simply replaced some of the search space with information. That keeps the player in the loop and prevents today's puzzle from turning into a thing you watch rather than a thing you do.

There is a second-order benefit to grading, which is that people use hints differently. Some players take one hint almost every day and treat it as part of the ritual. Others take none for weeks and then spend all three in a single bad sitting. A single all-or-nothing hint optimises for neither.

The layout matters more than I expected. Hints have to be visible where the player is stuck, not behind a settings menu, and they have to be cheap to dismiss. A hint the player did not want is worse than no hint at all, because it removes the sense that the board was theirs.

One thing I would not do again: hiding the hint counter. Players who know the count is tracked behave differently than players who do not, and not in a good way. Making the count explicit made the system feel fair.

If you are copying this, the graded approach is the part to keep. It fits naturally with unlimited practice, where the cost of a hint is only the time you spend, and it gives you a simple rule for every level above the first: never give information the player could have derived from the board.