Rock Paper Scissors

Rules (in case you forgot)

Features

The math behind RPS

Against a truly random opponent, your expected win rate is exactly 33.3% — you should win, lose, and draw in equal thirds over a long enough session. If your win rate is significantly above 33% after 100 rounds, you might be getting lucky; if it's significantly below, the same applies. This is a basic property of a zero-sum game against a random strategy.

The formal concept behind this is the Nash equilibrium. In a symmetric zero-sum game like RPS, the unique Nash equilibrium is for both players to randomize uniformly — playing each option with exactly one-third probability. Any deviation from this distribution, by either player, creates an exploitable tendency. If you throw Rock 40% of the time, an opponent who notices will simply throw Paper more and gain an edge. The math is unforgiving: the moment your distribution is non-uniform and your opponent detects it, they can profit from that asymmetry.

This is why the computer in this implementation uses a purely uniform random distribution. Against a computer playing true randomness, no strategy — no counting, no streaks, no patterns — can produce a long-run win rate above 33.3%. The expected value of every choice is identical. Playing against the computer is therefore a useful calibration exercise: it tells you exactly what 33% feels like over many rounds, which turns out to be a useful baseline for understanding variance.

Why humans are not random

The core strategic insight of competitive RPS is that humans are genuinely bad at generating random sequences. Even when people consciously try to randomize, their choices show statistical regularities — patterns that a careful observer can identify and exploit. This is not a weakness unique to novices; professional poker players, professional athletes, and people with high general intelligence all show measurable non-randomness in RPS studies. The brain simply does not have a random number generator.

One of the most consistently observed tendencies is a bias toward Rock on the opening throw. Rock is described by many players as feeling "strong" or "solid," and beginners in particular default to it when they have no other information. Among more experienced casual players this bias diminishes, but it rarely disappears entirely. A player who knows their opponent tends to open with Rock has a slight first-round edge by choosing Paper. This is a real exploitable tendency, not a manufactured one.

A second well-documented pattern is that players tend to avoid throwing the same move three times in a row. After two consecutive Rocks, the pressure to switch feels strong — most players will change to Paper or Scissors on the third throw. This is sometimes called "pattern fatigue" and it creates a predictable shift that an opponent tracking throws can anticipate.

Players also tend to cycle through all three options in clusters. Rather than truly mixing randomly, many people unconsciously ensure that Rock, Paper, and Scissors all appear within any short window of throws — a kind of self-imposed balance that makes the sequence far less random than it appears. Knowing this, an opponent can sometimes infer which throw is "overdue."

Win-stay, lose-shift: the most common unconscious strategy

Perhaps the single most important human tendency to understand in RPS is the win-stay, lose-shift heuristic. After winning a round, players disproportionately repeat the throw that won. After losing, they tend to switch away from the losing throw. This behavior is observed across many competitive contexts — it appears to be a general reinforcement-learning pattern that the brain applies automatically, even to simple games.

Here is how an opponent can exploit win-stay, lose-shift in practice. Suppose your opponent just won with Scissors, beating your Paper. Win-stay predicts they will throw Scissors again. You should throw Rock. If your opponent just lost with Rock to your Paper, lose-shift predicts they will switch — but to what? Lose-shift players often shift to the move that would have beaten the opponent's last throw. In this case, your opponent lost to Paper, so they might shift to Scissors to beat Paper. You should therefore throw Rock to counter the predicted Scissors. The shift-to-counter pattern turns lose-shift into a somewhat predictable sequence if you track carefully.

After a draw, behavior is less consistent, but many players shift to the move that beats what was just thrown. If you both drew with Rock, your opponent may throw Paper next (to beat Rock). Knowing this, you would throw Scissors. None of these predictions are certain, but each represents a slight probabilistic edge over guessing uniformly.

Conditional reads and second-level thinking

Against a perceptive opponent, simple tendency exploitation breaks down quickly. If your opponent knows you know they tend to throw Rock first, they will throw Scissors to beat your anticipated Paper. This escalation of predictions — thinking about what the opponent thinks you think — is called second-level or third-level thinking, and it is where RPS strategy becomes genuinely interesting.

A useful framework is to assign a "level" to yourself and your opponent. A level-0 player throws randomly or based on gut feeling. A level-1 player exploits basic tendencies. A level-2 player predicts that the opponent is level-1 and plays to counter their exploitation. A level-3 player accounts for this, and so on. The best counter to a very high-level opponent is often to revert to random play, since at some point the modeling becomes sufficiently uncertain that pure randomness is again the optimal baseline.

In practice, most recreational RPS interactions stay at levels 1 or 2. Reading the opponent means paying attention not just to their throwing history but to their body language, their response to your last outcome, and any tells they exhibit before revealing their throw. In-person RPS players sometimes observe the way an opponent positions their hand during the countdown — a fist forming loosely may indicate Rock, fingers extending suggest Scissors. These are conditional reads based on physical observation rather than throw history.

A worked example: you are in round four of a best-of-seven series. The sequence so far has been: you threw Rock and won (opponent threw Scissors), then you both drew with Paper, then you lost with Paper to Scissors. Applying the analysis: your opponent threw Scissors, then Paper, then Scissors. They have thrown Scissors twice in three rounds. You have thrown Rock and Paper twice each, which an attentive opponent might note. Win-stay says they repeated Scissors after a draw (round 2 Paper draw, round 3 they shifted to Scissors, which won). Now for round 4: lose-shift predicts you will switch from Paper. If your opponent thinks you will switch to Rock, they throw Paper. If you predict that, you throw Scissors. This is the kind of live reasoning that competitive RPS players actually run through.

The unbeatable random baseline

Understanding all of the above tends to produce a counterintuitive conclusion: the correct long-run strategy against any opponent whose level is uncertain is to play as close to uniformly random as possible. This is the Nash equilibrium strategy, and its guarantee is simple — it cannot be exploited. A player who genuinely randomizes one-third on each option will win exactly 33.3% over the long run regardless of what the opponent does. No strategy can beat it consistently.

The paradox is that humans cannot actually achieve this. True randomization requires a mechanism outside the human mind. Competitive RPS players who want to approach random play use physical methods: rolling a die before the match to bias their sequence, or using a prepared random sequence written on paper before the match begins. These methods genuinely approach the Nash equilibrium strategy and are difficult to counter.

For casual play, the practical takeaway is different. Against a casual opponent who is not actively tracking your throws, first-level tendency exploitation (throw Paper against beginners, watch for win-stay repeats) provides a genuine small edge. Against a careful opponent who is reading you back, mixing more aggressively and avoiding obvious patterns — not repeating the same throw more than twice consecutively, not cycling predictably — reduces your exploitability even if it does not achieve true randomness.

Competitive RPS: gambits and the organized game

Competitive RPS, as formalized by the World RPS Society and various international tournaments, adds a structured layer to the game through the concept of gambits. A gambit is a pre-planned sequence of three throws intended to create a specific psychological situation. The "Avalanche" gambit (Rock, Rock, Rock) applies pressure through sheer repetition — many opponents cannot resist switching to Paper eventually, but by the time they do, one gambit has already ended. The "Bureaucrat" gambit (Paper, Paper, Paper) mirrors this with Paper. The "Scissor Sandwich" (Rock, Scissors, Rock) conceals the Scissors in the middle position where it is least expected.

Gambits work because they exploit human discomfort with repetition. When an opponent throws the same move twice, most players feel an overwhelming urge to predict a change and act on it. The gambit player who sticks to the repeated throw can punish this predictable response. Against an opponent who recognizes gambits, the second-level counter is to identify the gambit early (usually after the first two throws confirm the pattern) and adjust on the third throw.

Tournament RPS typically uses best-of-three or best-of-five formats, with elimination brackets. Players often conduct pre-match "trash talk" as a deliberate psychological technique — claiming they will throw Rock is sometimes true (a double bluff), sometimes false, and sometimes the claim itself is the move (a triple bluff where the opponent who dismissed the claim now doubts their dismissal). The game at the highest levels is almost entirely psychological.

Practical takeaways for casual play

Here is a summary of the most actionable knowledge for someone who wants to win more often against casual human opponents — not guarantees, but genuine probabilistic edges:

• 1.Open with Paper — Rock is the statistically most common opener among casual players, making Paper marginally the best opening throw.
• 2.Watch for repetition — if your opponent wins with a throw, check whether they repeat it. Many do.
• 3.After a loss, opponents often shift to the move that would have beaten your last throw. If you threw Paper and won, they may shift to Scissors. Counter with Rock.
• 4.Avoid throwing the same move three or more times consecutively — most opponents will predict the continuation and exploit it.
• 5.Against an observant opponent, mix deliberately — choose throws that break the patterns you yourself would predict.
• 6.Against any opponent who might be tracking you, occasionally reset to a short burst of true-feeling randomness to erase established patterns.
• 7.In a best-of series, pay more attention during rounds 2 and 3 than round 1 — you will have actual data on your opponent's tendencies by then.

None of these tips change the fundamental math against a random computer. Against the implementation on this site, which uses Math.random() uniformly, your long-run win rate will be 33.3% regardless of strategy. But understanding why these tendencies exist in human play makes the game meaningfully more interesting when you bring it into the real world, and gives you an appreciation for why a deceptively simple game has sustained organized competition for decades.

Frequently Asked Questions