How to use randomness without frustrating players

1 The bottom line

Put luck before the player decides, so it feeds their plans instead of wrecking them. Players forgive randomness that gives them something to work with and resent randomness that blocks what they already know.

2 Key takeaways

• Roll the dice before the player chooses, not after. Luck they can see is luck they can plan around.
  • Subset's FTL was packed with swingy output randomness, while Into the Breach almost exclusively uses input randomness and is fairer and more strategic Mark Brown, Game Maker's Toolkit7:32.
• Never let luck stop a player from using what they have learned. Knowing the answer but being unable to act on it feels like a wall, not a challenge.
  • In Blue Prince you might get a safe code, but the safe won't appear for three or four runs Mark Brown, Game Maker's Toolkit3:29.
• Put bounds on random draws. Limits keep luck exciting without letting a bad streak ruin a session.
  • Modern Tetris deals a shuffled bag of all seven pieces, which caps the drought between I-blocks at twelve pieces Mark Brown, Game Maker's Toolkit10:57.
• Limit how far ahead the player can see. Perfect information invites paralysis and airtight plans, and drama needs plans to be disrupted.
  • Into the Breach shows every enemy plan, so one turn can take 10 to 20 minutes of staring Mark Brown, Game Maker's Toolkit4:06.
• Spend randomness on variety and memorable moments. That is where it adds something that fixed rules cannot.
  • Maxis removed The Sims' urinal one-gap rule before launch because random toilet choice was funnier than predictable, dull behavior Mark Brown, Game Maker's Toolkit16:08.

the fundamental difference between randomness that supports strategy, and randomness that undercuts strategy

Mark Brown, Game Maker's Toolkit

2.1 Decide when the roll happens

Luck that arrives before a choice gives the player material to plan with, and luck that arrives after can undo the plan.

• Roll random events before the player acts. Mark Brown, Game Maker's Toolkit7:32
  • Let players draw a hand in a deckbuilder, or roll dice in Dicey Dungeons and then choose where to spend them.
  • Expect most of the anger to come from output randomness, which lands after the choice and can wreck a plan through sheer bad luck.
• Reveal the random result before the player acts when you can. Mark Brown, Game Maker's Toolkit8:25
  • Early builds of Slay the Spire hid what enemies would do until after you ended your turn. Making the random choice at the start of the turn let players strategise around their foes.
  • The game became far more fun. In Brown's words, "Output became input".
• Keep output randomness only where it helps the player or teaches contingency planning. Mark Brown, Game Maker's Toolkit17:35
  • Into the Breach's only output randomness is a small chance that an attack on a building misses and saves you from certain defeat. It is too small to rely on, but it feels great when it lands.
  • When shots can miss, players must prepare backup plans. The miss becomes input for the next turn, and the best XCOM players are those with a backup plan Mark Brown, Game Maker's Toolkit14:16.
  • Co-designer Justin Ma says players never complain when they expect something bad and get something good.
  • If you need a hit roll, use partial outcomes. Phoenix Point simulates each bullet, so some hit and some miss, which is far less annoying than XCOM's all-or-nothing shot Mark Brown, Game Maker's Toolkit14:29.

2.2 Never let luck block what the player knows

If the random draw can hide the lock for a clue the player already holds, the player waits instead of thinking.

• Make sure the clue and its lock can both be reached. Mark Brown, Game Maker's Toolkit3:29
  • Adventure puzzles boil down to finding a clue in one place and applying it in another. Random drafting can leave you with "A" while your house has no "B", or the reverse.
• Let players test theories on demand. Mark Brown, Game Maker's Toolkit4:24
  • Testing a hunch is a core pleasure of puzzle games. When each test needs a specific combination of rooms, the guess, test, learn loop breaks.
  • Brown spent hours in Blue Prince knowing what he wanted to do but blocked by bad rolls, asking where the pump room was.
• Don't pair random draws with puzzles that demand single-thread focus, and count the memory load first. Mark Brown, Game Maker's Toolkit5:07
  • Random draws mean you can't follow one rabbit hole, so methodical players get knocked off their thread. Brown contrasts this with chasing the quantum moon in Outer Wilds or following the Golden Path in Tunic, and admits he kept re-rolling.
  • Blue Prince asks players to juggle storylines, a family tree, dates, codes, languages, paintings, chess pieces and locked doors. The more threads a game has, the more random interruption hurts Mark Brown, Game Maker's Toolkit5:42.
• Give players tools to bend the pool toward a goal. Mark Brown, Game Maker's Toolkit10:20
  • In a deck-building style game, players should be able to make certain rooms likelier and remove unwanted ones.
  • Blue Prince offers rarity changes, re-rolls, forced room colours and next-day pool shaping. Brown finds them limited, since you can't easily remove rooms or boost a very specific one.
  • Done well, arranging rooms near each other becomes a puzzle in itself. It is "a way of proving knowledge to the game", like being in a specific place at a specific time in Outer Wilds Mark Brown, Game Maker's Toolkit7:52.

2.3 Shape the draw so streaks stay bounded

Constrain how bad or good a run of luck can get, and show players the odds in a form they trust.

• Constrain random draws. Mark Brown, Game Maker's Toolkit10:57
  • Pandemic splits the deck into four piles and shuffles one epidemic card into each, so epidemics spread evenly.
  • Diablo 3's smart loot favours items for your class.
  • Two epidemics in a row is still possible in Pandemic, so the tension survives.
• Soften random starting conditions so players don't restart until they get lucky. Mark Brown, Game Maker's Toolkit9:08
  • In Spelunky a crate can give a jetpack or shotgun at the start. That makes skill hard to tell from luck and tempts restarts.
  • Slay the Spire gives starting bonuses only if you reached the first boss last time, so players work with what they are given.
  • Spelunky skips its dark levels if you cleared the previous stage in under 20 seconds, as mercy to speedrunners.
• Show the odds, and make them match how chance feels. Mark Brown, Game Maker's Toolkit14:46
  • XCOM displays hit percentages that change as you move closer, so players weigh which risks to take.
  • Use familiar randomizers. Zach Gage says "we understand things that we can hold in our hand", so Tharsis uses dice, and Tharsis and Armello add physics to their digital dice Mark Brown, Game Maker's Toolkit16:04.
  • Cards add dependent probability, because drawing one changes the deck. That enables synergies in Slay the Spire, while dice rolls are independent Mark Brown, Game Maker's Toolkit16:04.

2.4 Set how far ahead the player can see

The information horizon decides whether players can plan at all, so set it on purpose.

• Choose your tool for the information horizon. Mark Brown, Game Maker's Toolkit5:19
  • Keith Burgun defines it as the distance between the current turn and the point where information becomes known. You can push it out with complexity, execution uncertainty, hidden information or randomness.
  • Chess uses exponential complexity, Starcraft uses fog of war, and Into the Breach uses not knowing enemy actions next turn.
• Don't give players perfect information. Mark Brown, Game Maker's Toolkit4:06
  • With full transparency, players calculate far ahead and optimise the fun out of the game. In Plants vs Zombies the upcoming wave is visible, so a perfect defence is easy and play turns flat.
  • Drama comes from plans getting disrupted. Brown notes there has never been a good movie where the heroes' scheme perfectly works as intended.
• Release new random information in spikes, not every turn. Mark Brown, Game Maker's Toolkit13:17
  • Fresh input randomness every turn pulls the horizon so close that players cannot make lasting plans.
  • Ethan Hoeppner suggests high-impact information arriving in discrete spikes at regular intervals, with a slow, regular flow between them.
  • In XCOM, plans stay viable for several turns until a new pod of enemies appears and forces a regroup.

2.5 Spend randomness on variety and discovery

Randomise the parts that keep repeat play fresh, and leave fixed the parts players need to learn.

• Randomise small elements inside a fixed layout. Mark Brown, Game Maker's Toolkit4:41
  • Prey: Mooncrash keeps the same floor plans but randomises hazards, collapsed tunnels or staircases, offline tram stations, escape pod and closet locations, enemy spawns and item spots. Players keep learned shortcuts, while a broken route forces them to think on their feet.
  • Deathloop repeats areas almost identically, down to turrets and Fiz-Pops, so by the sixth visit the surprise is gone.
• Draw enemies from a pool so memory alone cannot win. Mark Brown, Game Maker's Toolkit6:05
  • Baroness Von Bon Bon's first three phases draw sub bosses at random from five: Candy Corn, Gobstopper, Cupcake, Chocolate and Gumball Machine.
  • Each covers the screen differently and demands a different style of dodging, so players must react each attempt.
• Regenerate content so players master mechanics instead of memorising levels. Mark Brown, Game Maker's Toolkit2:19
  • In roguelikes, regenerating content each run forces players to master the underlying mechanics.
  • Rogue used procedural generation for an infinite supply of dungeons. Koichi Nakamura later paired random generation with permadeath in the Mystery Dungeon series Game Maker's Toolkit14:22.
  • Procedural levels are rarely as good as handmade ones.
• Use luck for discovery, side risks and fresh scares. Mark Brown, Game Maker's Toolkit7:35
  • A huge possibility space makes a secret feel special. In Blue Prince's drafted mansion, stumbling on a secret room feels like a genuine discovery.
  • Put random rewards off to the side. The Space Invaders UFO gives a random number of points, and the player chooses whether to spend shots and attention on it Mark Brown, Game Maker's Toolkit1:09.
  • Animal Crossing's turnip prices change daily. Lucky players profit hugely and unlucky ones watch turnips drain away, so the exploit becomes a decision about how much to risk Mark Brown, Game Maker's Toolkit13:54.
  • Brown's hypothetical Until Dawn redesign would be shorter, replayable with new instructions each time, and add random events so scares don't turn blase Mark Brown, Game Maker's Toolkit8:38.

2.6 Where designers disagree

Some luck costs players time or trust, and what pays off depends on your game's pace and audience.

• A fresh random layout can free players, or it can waste their hours.
  • Blue Prince dissolves dead ends, since you "just start a new day" with a whole new set of things to check Mark Brown, Game Maker's Toolkit8:39.
  • The cost is time. Brown stopped after room 46 because the massive endgame means hours of wrangling with RNG, and "The game doesn't really respect my time" Mark Brown, Game Maker's Toolkit12:55.
  • A prominent day counter suggests time pressure and stresses "wasted runs", even though there is no limit. Messaging should match the real rules Mark Brown, Game Maker's Toolkit11:12.
  • Brown still likes the single high-stakes final run, as in The Forgotten City and Prey: Mooncrash, where "You gotta nail everything in a single run" Mark Brown, Game Maker's Toolkit8:20.
• Honest odds or fudged odds.
  • Displayed odds turn luck into a decision. But humans are bad at understanding odds, so designers often lie Mark Brown, Game Maker's Toolkit15:12.
  • Fire Emblem quietly makes a 90% chance to hit behave more like 99%. Civilization makes the third of three 33% battles succeed after two losses, and Hearthstone allegedly has a pity timer for legendary cards Mark Brown, Game Maker's Toolkit15:12.