When the clues disagree, do not guess
Separate position clues from relation clues, write them down, and change one part per attempt.
A good lock box gives you more than one clue, and almost always one of those clues is wrong. Not wrong because the design is broken, but because some clues are deliberately written to mislead. Players who fail are usually not the ones who are not clever enough; they are the ones who accept the first clue as true and build the whole answer on that assumption. The Locks & Combinations section keeps titles that let you try again, so a mistake carries no penalty — but your time is still spent.
The first step is to split clues into two kinds: clues that describe position, and clues that describe relation. Position clues tell you where something is: on the left, above, across. Relation clues tell you order: before, after, next to. The two cannot disagree if the box was designed properly, so when they do disagree, one of the two kinds has been written as a trap.
The second step is to write. Lock boxes are the one puzzle type that genuinely needs paper. Write each clue as a line, then mark which ones you have proved and which are still assumptions. In most boxes two or three clues are enough to open the lock, and the rest only add confidence. When you are stuck, look at the lines that are still assumptions: that is usually where the error began.
The third step is to test with one small change, not by rebuilding the whole answer. If the lock rejects your combination, change one part only and try again. Changing three parts at once tells you whether the answer is right or wrong, but not which part was wrong. Boxes that allow unlimited attempts suit this method; boxes that punish the second attempt do not, which is why they stay off the shelf.
Finally, stop when you start guessing at random. Random guessing means you have lost the trail of clues, and continuing only adds time. Close the tab, reopen it later, and read the clues from the first line. The answer often looks obvious after a short break, because you stopped defending an old assumption. A lock box is not a test of intelligence; it is a test of whether you are willing to read again.
One kind of clue gets ignored most often: the one that mentions a count. Clues like three keys, two chests or one door are usually the easiest to test, because you can count for yourself whether the number matches what is visible on screen. When the number in the clue does not match the objects, you already know the clue is wrong without trying a combination. Use these clues as your first filter, and save the position clues for the last step.
Besides counts, two other kinds of clue are easy to test: colour and material. A clue that names a colour usually points at an object that exists only once in the room, so one look confirms it. A clue that names a material tests whether you can tell metal from wood, or glass from stone. When two clues describe the same object with different colours or materials, one of them is almost certainly a trap. Keep both, mark them as conflicting, and solve the part of the lock that does not depend on that object first. In most boxes that part is enough to open the lock, and the conflicting clue never has to be resolved at all.
Besides writing what each clue says, write the order in which you found it. A good lock box hides the decisive clue in a place you visit later, and that order is itself information: the clue found last usually locks the answer, while the first clue often only points a direction. When you are stuck, reread your list from the bottom up rather than the top down. The method reverses a normal habit and often exposes an assumption you built early in the game. In a box with many clues, reading from the bottom up cuts the time you spend stuck by more than adding another attempt does.

Formula Car Stunt
Endless Path: Twist & Turn