Rock paper scissors is supposed to be random. Three equal options, no information, no strategy. The theory says each choice should appear about one third of the time and players should be impossible to predict.

The data from 7,934 real rounds played on TicToe.org over two months says otherwise.

The choice breakdown contradicts the research baseline

The World Rock Paper Scissors Association, which has studied competitive play across thousands of tournament matches, reports a consistent baseline: rock appears most often at 35.4%, scissors next at 35.0%, and paper least at 29.6%.

Our data tells a different story.

Choice breakdown across all rounds (7,934 rounds):

Scissors is the most common choice in our dataset, not rock. And the three options are far more evenly distributed than tournament research suggests, with less than 2.2 percentage points separating the most and least common choice.

The online multiplayer data specifically, which covers 4,661 rounds between two real human opponents, is closer to the research baseline with rock at 35.8%, scissors at 32.5%, and paper at 31.7%. That gap between the aggregate data and the online-only data suggests players behave differently depending on who they think they are playing against, a human or a computer.

Against a computer, people may feel freer to experiment since losing feels lower stakes. Against a real opponent, the familiar patterns from competitive research reassert themselves. Rock climbs back to the top when the pressure is real.

Winners repeat. Almost always.

The most consistent finding in rock paper scissors research is that winners tend to repeat their winning throw on the next round. Our data confirms this pattern with striking consistency.

After winning (probability of repeating the same throw):

Across all three throws, roughly three quarters of winners play the same thing again. This is not rational behaviour in a random game. It is a psychological pattern, the winning move feels lucky or powerful, so players stick with it.

The practical implication is straightforward. If your opponent just beat you with rock, there is a 74% chance they will throw rock again. Throw paper.

Losers are more complicated

The theoretical prediction for losing behaviour is that players switch to the throw that would have beaten what just beat them. Lost to rock? Switch to paper. Lost to scissors? Switch to rock. The theory calls this the "losing switch" pattern.

Our data shows this is partially true but weaker than expected.

After losing (probability of switching to the throw that would have won):

The most common response to losing is actually to repeat the same throw, happening 68 to 73% of the time. Players are more stubborn after losing than the theory predicts. Only about one in seven losers makes the theoretically optimal switch.

This is a more exploitable pattern than pure switching would be. If someone just lost with rock and you expect them to switch to paper, you throw scissors. But if they are more likely to throw rock again, you should throw paper. The data says stick with paper.

Draw behaviour is the least studied pattern

After a draw, players are also more likely to repeat than switch.

After drawing (probability of repeating the same throw):

Draw behaviour has been studied less than win and loss behaviour in the existing research, which makes this finding genuinely novel. Players treat a draw more like a win than a loss in terms of their next throw. The draw feels like partial success, so the throw that produced it feels worth keeping.

What this means if you are playing right now

The exploitable patterns in order of reliability:

Most reliable: If your opponent just won, they will almost certainly repeat. Play what beats their last throw.

Second most reliable: After a draw, your opponent will probably repeat. Same strategy applies.

Least reliable but still useful: After losing, most players repeat despite the theory saying they should switch. Play what beats their last throw again, but be aware this pattern is weaker than the others.

The catch is that a good opponent knows these patterns too and will try to break them. The moment you establish a pattern yourself you hand them the same advantage. The optimal strategy in a long match is to exploit your opponent's tendencies while actively varying your own.

Against the computer on this site, none of this applies. The computer picks genuinely at random with no pattern to exploit. Play online against a real opponent where the psychology actually matters.

How this compares to the tic tac toe data

We ran a similar analysis on tic tac toe earlier this year using 26,776 games, finding that the most popular opening move has the lowest win rate and that going first wins nearly four times as often as going second. The rock paper scissors findings follow a similar pattern: real human behaviour diverges from the theoretical optimum in consistent and exploitable ways.

The full tic tac toe analysis is here if you want to compare: First mover advantage in tic tac toe: data from 26,776 real games

About this data

All figures are drawn from completed games played on TicToe.org over two months. The choice breakdown covers 7,934 total rounds across all game modes. The online multiplayer subset covers 4,661 rounds between human opponents. The throw sequence data covers rounds where the preceding round outcome was known. Only completed sessions are included.

One limitation worth flagging: we cannot guarantee that all rounds come from unique players. Someone playing multiple sessions would contribute more than once to the dataset. This means the data reflects how people actually play on TicToe.org rather than a perfectly controlled sample of unique individuals. The patterns are consistent enough across the dataset that we are confident they are real, but they should be treated as observational rather than experimental.

Data collection is ongoing. This post will be updated as more rounds accumulate and the patterns either strengthen or shift.

Sources