Dungeons & Dragons plays best when players can only retry ability checks if something changed after the first roll.

For many ability checks, this makes sense because failing makes trying the same thing again impossible. The Dungeon Master’s Guide gives this example: “If the rogue loses a contest of Charisma (Deception) against the guard’s Wisdom (Insight), the same lie told again won’t work.” Typically, Intelligence and Charisma checks only allow one attempt.

Still, the realization that DMs should avoid allowing rerolls surprised me because some checks naturally invite retries. Typically, Strength and Dexterity checks seem open to multiple attempts. Why not try to pick a lock again? The Dungeon Master’s Guide even includes advice for multiple checks. “In some cases, the only real cost is the time it takes. To speed things up, assume that a character spending ten times the normal amount of time needed to complete a task automatically succeeds at that task.”

That suggestion seems like advice for retrying checks, but at the core, it tells when to skip them. If a character can succeed and has time, then skip the roll. This guideline never explains how to determine whether characters have enough skill and talent to succeed. (Spoiler: I will recommend rolling a check.)

DMs aim to run an exciting game, and skipping rerolls helps. The outcome of a roll should launch—or at least nudge—the game ahead or into a new direction. Even a failure that blocks progress should inspire ingenuity. If a failure does nothing more than prompt a player to fish for a higher number, then the game stalls. Boring. Third edition even created rules to avoid such fizzles. A player could take 20 and spend 20 times as long in exchange for an automatic 20 on a check attempt.

For checks that seem to allow multiple retries, the take 20 rule answers the can-I question using addition. While fifth edition skips the take-20 rule, the math remains the same. If a character’s check bonus plus 20 meets or beats a task’s DC, then mathematically the character could succeed. But this take-20 math leads to unsatisfying play for two reasons:

• Take-20 math eliminates the fun and surprise of rolling a die to answer the game’s dramatic questions. See 3 benefits of letting die rolls shape your game world’s reality.
• When characters have plenty of time, take-20 math makes difficulty classes too easy.

If bending the bars of dungeon cells requires success against a DC 20, which the game calls “hard,” then anyone with an average Strength of 10 and time to spare gets loose. The same goes for a hard lock and an untrained thief with no aptitude. For dungeons to hold athletes and locks to deter thieves, the iron and the mechanisms need a DC 25 and talented beginners still get through. I think we can all agree that prisons work better than that. But we contrive DCs to create a fun and challenging game. D&D aims for heroics where the barbarian bends the portcullis and the rogue picks the lock, all before the approaching monsters arrive. So the game recommends DCs based on the assumption that characters only get one attempt.

Instead of using take-20 math to decide whether a character has enough talent and skill to succeed, the game usually plays better when one ability check decides. Knowledge checks already work that way. When players ask if their characters know the history of Netheril, a check answers without rerolls or taking 20. For any check where failure blocks a retry, one die roll decides whether a character can succeed. Can the paladin persuade the mayor? Can the wizard decipher the sigil?

Checks that decide what a character can accomplish date back to first edition and attempts to bend bars or lift gates. “The attempt may be made but once, and if the score required is not made, the character will never succeed at the task.” Does a failed check reveal thicker bars? Maybe. Such checks fit D&D’s long tradition of rolling to learn things about the game world, things like whether a door is locked or if the skies are clear.

Die rolls bring more fun than letting take-20 math or a DM’s ruling decide between success and failure. Plus, you only roll once so the game never stalls while someone fishes for a higher number.

Now that you know to never allow a character to repeat a check, I will weaken that principle by revealing exceptions.

One obvious exception: Sometimes a change in approach allows a reroll. If a character failed to move a stone, perhaps a lever will help. Gain another attempt.

The second exception requires a DM’s judgement, something right in the DM job description.

When characters risk paying a price for failure, allowing retries can create a more exciting game. If a climber can fall or if someone attempting to disarm a trap can suffer the device’s effects, the character can retry and bear the risks again. Often the price of failure comes from losing time against a ticking clock—a draining hourglass in most D&D worlds. In classic D&D games, every moment wasted raises the risk of wandering monsters. When attempting to break a door, the price of failure is often noise, whether it alerts wandering monsters or nearby guards.

Fifth edition allows for setbacks more interesting than damage or delays. Failure can mean that a character “makes progress combined with a setback determined by the DM.” For example, if a rogue blows a roll to pick a lock, the DM can decide that the lock opens but the attempt made a noise that leads to complications. Gamers call this failing forward.

The Call of Cthulhu roleplaying game includes a “pushing the roll” rule that lets players agree to pay a price for a second failure in exchange for a retry. Players describe the extra effort or time taken to justify a second chance. Plus, they give the game master “permission to bring dire consequences” if the next attempt fails. You may steal this rule for your D&D game. The idea works especially well for checks with a potentially harsh penalty for failure. For example, a character who fails a first climb attempt realizes the cliff seems too treacherous and makes no progress. If the climber pushes on despite the risk, and then fails again, they fall.

Related: In D&D, Letting Everyone Roll Certain Checks Guarantees Success, So Why Bother Rolling?