Randomness has two jobs

Randomness is everywhere and it is not one tool. A shuffled level, a drawn card, a loot table: the randomness happens, and then you decide what to do with the result. That is input randomness. You can plan. A shot that misses because a hidden number said so, after you aimed well, is output randomness. You already decided. The game overruled you.

Players forgive the first and resent the second, unless the game has taught them that the fog is the sport, as in a tabletop war game where the dice are the agreed fiction. Using output randomness to disguise a difficulty you did not want to design is how combat feels cheap. Using input randomness to deal a fresh problem is how a card game stays alive for a hundred hours. Name which one you are shipping. If you cannot, you are probably shipping the cruel one by accident.

Randomness is everywhere in games, and it is doing a different job in each place. Generated levels, wandering encounters, missed shots, drawn cards, rolled dice, spun wheels, loot tables, and the number generators underneath them.

What these systems share is that neither the designer nor the player fully decides the result. Something the game cannot predict steps in and redirects what happens next.

Sometimes that produces the best story a player will tell about your game. Sometimes it produces a loss they never had a chance to avoid. The tool does not change; the way it lands does.

The most obvious job randomness does is supply volume. A good generator turns out levels, worlds, characters, enemy mixes, and problems faster and in greater numbers than any team could build by hand.

The trade is craft. Generated content is seldom tuned as tightly as an authored level, but it delivers breadth. The orc captains of Shadow of Mordor, the size of a Minecraft world, the shared universe of No Man's Sky, and the room layouts of roguelikes all draw on randomness, each in its own way.

There is a second effect worth noticing. When the layout changes every run, memorizing a stage stops being a strategy. Players have to understand the systems well enough to answer whatever the generator hands them.

In multiplayer, randomness works as a leveler: it lowers how much raw skill decides the outcome. A fortunate roll or a bad draw keeps a weaker player in contention longer than pure execution would.

The effect is strongest when the luck is aimed at whoever is losing. Mario Kart's item boxes hand better items to racers at the back of the field than to whoever is leading.

That fits party games, family games, and anything played across mixed skill levels. In a competitive game built around mastery, the same dial turned too far hides who actually outplayed whom.

Uncertainty makes a reward hit harder. A rare drop in a looter shooter lands the way it does because the player knows the odds were against it. Low probability manufactures anticipation, then surprise, then a spike of feeling when the result finally arrives.

The same machinery is easy to point the wrong way. Tune a random reward into a Skinner box and it starts harvesting time, money, and attention instead of serving the game.

So the question to ask of any random reward is plain: does it make the game better, or does it mostly live off the player's urge to try once more?

Randomness also puts a ceiling on planning, and that ceiling is often the point. Planning runs on information: where enemies stand, how much health they have, what resources are on hand, sometimes what the opponent intends to do.

Give the player more information and their plans get better, up to a point. Past it, trouble starts: when every variable is on screen, players keep calculating deeper in search of the optimal move, and the game grinds down into analysis paralysis.

Into the Breach lives on that edge. Every enemy attack for the coming turn is laid out, which is exactly what makes it such a sharp strategy game, and also why one turn can eat a very long time.

Full information can also make plans too dependable. Knowing the exact monster lineup ahead in Plants vs. Zombies makes a solid defense easy to assemble. Without something to disturb it, the game flattens out.

It helps to think in terms of an information horizon: how far ahead of the current moment the player can see before things become unknown.

That horizon can be pulled in by several means. Chess pulls it in with sheer branching complexity. Action games pull it in with execution risk, since a plan is only as good as the hands carrying it out. Strategy games pull it in by hiding things, fog of war being the standard example.

Randomness is simply one more way to do it. If parts of the future are unknowable by design, no plan can be perfect, and the player's job shifts from solving to preparing, adapting, and recovering.

The split that explains the most is input versus output randomness. Input randomness resolves first, and the player decides afterward.

The roguelike builds a level, then you walk into it. The deckbuilder deals your hand, then you choose how to spend it. Dicey Dungeons rolls the dice, then you pick which slot each one goes in.

What this kind of luck hands the player is a situation. Chance writes the problem; the answer is still entirely the player's to choose.

Output randomness runs the other way round. The player commits first, and chance decides what the commitment was worth.

XCOM's hit chance is the case everyone reaches for: you tell the soldier to fire, and the game decides whether the shot connects. Enemy moves resolved after you end the turn, wandering encounters, and loot boxes belong to the same family.

This is the version that makes players angry, because it can break a plan that was already committed to. The loss reads less like a mistake in judgment and more like the game pulling the controls away.

Hence the usual verdict that input randomness sits better with strategy. It hands you a random situation and asks you to work with it. Output randomness can void the work after you have already done it.

You can watch developers travel in that direction. FTL swings hard on output randomness, while Into the Breach shows almost every enemy intention before you commit. Slay the Spire improved when enemy plans became visible at the top of your turn rather than only after it ended.

None of which reduces to input good, output bad. They are both instruments. Input randomness left unchecked can wreck a game, and output randomness handled with care can carry one.

Left alone, input randomness can hand out openings so unequal that the player can no longer read their own performance. An early shotgun or jetpack in Spelunky changes a run enormously. The peaks are thrilling, but the unlucky runs start to feel like they did not count.

The predictable response is the reroll: quit and restart until the opening looks worth playing. It is a rational habit, and it is a warning sign, because the random setup is now talking players out of playing the game in front of them.

Slay the Spire answers this by granting a starting bonus only when the previous run made it to the first boss. The incentive points at finishing an awkward run instead of farming the perfect first hand.

Plenty of games constrain their randomness without draining it. Pandemic shuffles the epidemic cards into separate piles and stacks the deck from those. The procedure is fiddly, but it stops every epidemic from landing at once at the start, or arriving so late the game is already decided.

Diablo 3's smart loot biases drops toward the class you are playing, so fewer rewards are wasted. Modern Tetris shuffles all seven pieces into a bag, deals the bag out, then shuffles a fresh one, which guarantees variety while leaving the order genuinely unknown.

None of this removes luck. It bends the distribution so the exciting outliers survive while the runs that break the game outright become rare.

Timing matters as much as amount. Drop a fresh random event at the top of every single turn and the information horizon closes in so tightly that planning past the immediate move stops being worth doing.

The better pattern is usually spiky. XCOM gives you a stretch of turns to work a plan, then upends it when a new pod comes into view or reinforcements land.

Both halves of that rhythm are doing work. The quiet stretch of known information lets the player act deliberately. The occasional jolt makes them stop, regroup, and think again.

Output randomness earns its place too. In an abstracted combat system it stands in for error, imprecision, and the gap between intention and execution. Take it out and a tactics game where nobody ever misses starts to feel like a machine.

It is also how a game tests risk management. Playing XCOM well is not just picking the shot with the best number on it; it is knowing what you do when that shot misses.

Seen that way, output randomness converts into input randomness one turn later. The roll leaves a new board state, and the player has to plan from it.

One fix is to stop making results binary. Phoenix Point runs its shots through a ballistic simulation, bullet by bullet, so a burst can partly connect. A partial hit stings far less than a clean miss.

Publishing the odds matters as well. Once players can see the chance of success, they can decide what risk they are willing to carry, and understand what a step closer actually buys them.

The catch is that people are poor at probability. A 90 percent shot gets read as a sure thing, and a 33 percent shot as one that ought to land every third try. Plenty of games quietly bend the math behind the display so the number matches what players expect from it.

Familiar physical objects make chance easier to accept. A six-sided die is something a player already has intuitions about, in a way a hidden calculation is not. Cards carry the same familiarity, and they behave differently besides, because taking one out changes what is left.

That is part of why dice games like Tharsis and card games like Slay the Spire read so clearly. Players can reason about the randomness because its shape matches something they already know.

Cards add dependent probability on top. Every draw alters the odds of the next one, and that is what makes long-term deck reasoning and cascading synergies feel so good to build.

More on this: Card game maker: the rules engine is the hard part

The safest way to use output randomness is to let it break in the player's favor only. Into the Breach gives buildings a small chance to shrug off an attack that would otherwise have connected.

Because nobody can count on it, no plan is ever built on it. When it does fire, it feels like a gift: the outcome you had already accepted as bad turns out fine.

That asymmetry is the lesson. Luck that pulls a player out of disaster reads as generosity. Luck that drops them into one reads as the game working against them.

Luck can supply variety, close skill gaps, sharpen rewards, cap perfect information, manufacture surprise, and turn players into people weighing odds.

It can also wreck fairness, muddy who is responsible for a result, push players into rerolling, exploit the reward centers it touches, and make defeat feel arbitrary.

Input and output is the distinction that clarifies the most, but neither one is correct on its own. The design work is in the details: when chance resolves, how far it can swing the game, how much the player can see, and whether they get a real decision before it or after. Most of that is answerable only by building it, playing it, and trusting what the losing turn felt like.

In the Flockbay app, build one fight where the enemy pattern is shuffled before the fight, and you can see it. Then a version where your hits randomly miss. Play both. Keep the one you would replay. The other is the fog.

An AI deckbuilder maker is the honest home of a shuffle. A tactical RPG template should show you the roll if you insist on dice.