ArcadeWhale
RNG stands for random number generator. It is the part of a game that decides everything left to chance: whether your attack is a critical hit, what drops from a chest, which piece falls next in a puzzle game, how a dungeon is laid out. When players say “the RNG hates me” or “that run was pure RNG”, they are using the word to mean luck itself.
That much is simple. What is less known is that game randomness is almost never truly random, that true randomness feels unfair to most people, and that a lot of games quietly bend the odds to make up for it. Below are two small experiments you can run right here, plus the tricks designers use.
Computers cannot flip a coin
A computer follows instructions exactly, which makes it bad at surprises. So games use pseudo-random number generators: formulas that take a number, scramble it, and hand back the result, which then becomes the input for the next number. The output looks random, but the whole sequence is fixed by the very first number, called the seed.
Some older games went even simpler. The original Doom (1993) did not calculate random numbers at all; it read them in order from a fixed table of 256 values and started again from the top when it ran out.
Seeds are why the same Minecraft world can be shared as a string of digits, and why games like Spelunky 2, Slay the Spire and Balatro let you play a “seeded run”: everyone who enters the same seed gets the same layout and the same draws. It is also what makes RNG manipulation possible. In some games the next random number depends on exactly which frame you press a button on, so speedrunners and Pokémon shiny hunters time their inputs to the sixtieth of a second to get the outcome they want.
Experiment 1: a 25% critical hit
Imagine a weapon with a 25% chance to land a critical hit. Below are forty attacks, rolled two ways. Pink squares are crits.
True random
Pseudo-random distribution (how Dota 2 does it)
Both rows land a crit about 25% of the time in the long run. Orange squares mark the longest dry spell in each row. Roll a few times: the top row keeps setting new records, the bottom row never goes past 11.
In the true-random row you will regularly see long droughts, and just as often three or four crits bunched together. That is what randomness actually looks like. It is also what players call broken.
We simulated 5,000 sessions of 200 attacks each. With true 25% odds, the typical longest dry spell was 14 misses in a row, and the worst one we saw was 39. With the pseudo-random version, the typical longest dry spell was 7, and it can never go past 11, because the chance climbs with every miss until a hit is guaranteed on the twelfth attack.
That second system is called a pseudo-random distribution, and it comes from Warcraft III and Dota 2. The first attack only has about an 8.5% chance to crit. Each miss adds another 8.5%, and a crit resets it. The average stays at 25%, but streaks in both directions become rare. Players get the luck they expected instead of the luck the maths would have given them.
Why fair luck feels unfair
People have strong, wrong intuitions about randomness. We expect it to alternate: hit, miss, miss, hit, miss. Real randomness clumps. After three misses we feel a hit is “due”, which is the gambler’s fallacy: the dice do not remember.
This is not only a games problem. Music players ran into it with shuffle. Listeners complained that a random shuffle kept playing the same artist twice in a row, so services like Spotify changed their shuffle to spread songs out on purpose. It became less random so that it would feel more random.
Sid Meier described the same thing in a well-known 2010 talk about Civilization Revolution. Players who attacked with odds of three to one expected to win, every time, and felt cheated when they lost one battle in four, even though that is exactly what those odds mean. The team adjusted the combat so outcomes matched what players felt was fair.
Experiment 2: waiting for the long piece
Anyone who has played a falling-block puzzle game knows this pain: you have built a perfect well for the long I-piece, and it simply does not come.
Pure random: every piece is a 1-in-7 roll
7-bag: shuffle all seven pieces, deal them out, repeat
Blue squares are I-pieces, grey squares are the other six, and orange squares mark the longest wait for an I-piece. Deal a few times: the top row keeps finding longer waits, the bottom row never goes past 12.
With pure random, a long wait is not bad luck. It is guaranteed to happen eventually. We dealt 1,000 pieces 10,000 times: the typical longest wait for an I-piece was 34 pieces, one game in ten had a wait of 46 or more, and the worst drought was 87. About one wait in twenty-two lasted 20 pieces or longer.
With the 7-bag system, the longest possible wait is 12 pieces, and it cannot be longer, ever. The game puts one of each of the seven shapes in a bag, shuffles it, deals them out, and only then refills the bag. Official Tetris games have used this rule since the early 2000s, and plenty of other puzzle games use a version of it. The pieces still arrive in a random order; they just cannot abandon you.
The other ways games bend the odds
The number on screen is not always the real one. Several Fire Emblem games, starting in the Game Boy Advance era, rolled two random numbers and averaged them to decide a hit. That makes high percentages more reliable and low percentages less likely than they look. A displayed 70% hit really lands about 82% of the time, and a displayed 30% only about 18%.
The game is on your side. Firaxis designer Jake Solomon has talked about how the easier difficulty settings in the modern XCOM games quietly tilt the odds toward the player after a run of misses, precisely because a 95% shot missing feels like a bug even when it is not.
Pity timers. Games that sell or hand out random rewards often promise a guaranteed prize after a certain number of misses. Genshin Impact publishes its rule: at most 90 pulls on a character banner before a five-star appears. Without a cap, some players would go hundreds of attempts with nothing.
Catch-up luck. Racing games that give better items to players at the back of the pack, like the Mario Kart series, use randomness to keep races close. It is still a random draw; it is just weighted by your position.
Good RNG and bad RNG
Game designers sometimes split randomness into two kinds, and it explains why some random games feel great and others feel cruel.
- Randomness before you decide (sometimes called input randomness): the game deals you a random hand, map or set of choices, and then you make the call. Deckbuilders and most roguelikes work this way. When you lose, you usually know which decision cost you.
- Randomness after you decide (output randomness): you make the right choice, and then a roll decides whether it worked. A 90% shot that misses is the classic example. It creates big dramatic moments, but losing to it can feel like the game overruled you.
Most games mix the two. The ones people call “fair” tend to give you more of the first kind and cushion the second with the tricks above.
RNG slang, quickly
- RNG carried: won mainly because of good luck.
- Bad RNG / RNG screwed me: lost to bad luck, or at least that is the story.
- RNGesus: the joking deity players pray to before a big roll.
- RNG-dependent: a strategy or build that only works with lucky draws.
- Low-RNG / no-RNG: a game or mode where skill decides nearly everything.
So the next time a game hands you a miracle or a disaster, it is worth wondering which one you got: real chance, a bag, a hidden bonus or a pity counter quietly ticking toward zero. Very often, the luck in a game is designed just as carefully as everything else.
