How to Play Rock-Paper-Scissors at 6proclub: A Cryptographic Take on the World's Oldest Bluffing Game

Rock-Paper-Scissors looks like a coin flip until you play it five times against the same person. The rules, our best-of-5 format, the commit-reveal protocol that makes online RPS actually fair, and the strategy behind a game that is really about people.

Rock-Paper-Scissors is the game everyone learned before they could read. That is exactly why it is interesting. A game that simple, played by that many people, for that many centuries, has been pressure-tested by humanity itself — and it still works. The question is why it works.

This post covers the rules of RPS as we ship it on 6proclub, the cryptographic protocol that makes online RPS playable in the first place, and the strategy you can actually apply once you understand that the game is not about your hand — it is about the person across from you.

The basic rules

Each round, both players simultaneously choose one of three options: rock, paper, or scissors.

• Rock beats scissors (rock crushes scissors)
• Scissors beats paper (scissors cut paper)
• Paper beats rock (paper covers rock)

If both players choose the same option, the round is a tie and is replayed. The round counter does not advance on a tie — only on a decisive round.

That's it. The entire ruleset.

Our match format

At 6proclub, RPS is played as best-of-5. The first player to win three decisive rounds takes the match.

A few specifics:

• Ties replay. A tie does not count as a round won for either side. The pair simply resets and you both choose again.
• Hard cap at 9 rounds. In the rare pathological case where players keep tying — five, six, seven times in a row — the match is hard-capped at nine total rounds (decisive + tied). If the cap hits without a winner, the side with more decisive wins takes it. We have never seen the cap fire in production, but it is there.
• No timeout punishment for thinking. The bidding window is generous. Take your time. Choosing fast does not change anything except how fast you lose.

The cryptography problem (and how we solved it)

Online RPS has a fundamental problem that does not exist face-to-face: how do both players choose simultaneously over the internet?

If we just send the chosen move to the server, the server (or an attacker) sees both moves before declaring a winner. If the server sends one player's move to the other before the second player has chosen, that player wins every time. There is no naive way to do this safely.

The standard solution is a commit-reveal protocol, and it is what we ship on every RPS game:

Phase 1 — Commit

When you make your choice, your client computes a SHA-256 hash of (your-choice, random-nonce) and sends only the hash to the server. The hash is a 64-character hex string that reveals nothing about your choice — but it is cryptographically bound to it. You cannot change your mind after submitting the hash without producing a different hash.

The server holds both players' hashes until both have arrived. Neither player has seen the other's choice. Neither has the server.

Phase 2 — Reveal

Once both commits are in, both clients send the actual (choice, nonce) to the server. The server hashes each one and confirms the result matches the original commit. If it does, the choices are accepted as binding. If a player's reveal does not match their commit, they forfeit the round.

Now both choices are known and the round is resolved.

This is mathematically equivalent to playing RPS in person with a transparent box. Neither player can change their move after seeing the other's. Neither can the server. The fairness is verifiable: any third party with access to the commit hashes and the reveal data can re-derive every round of every match we have ever shipped.

Strategy: the game is not about hands

Here is the surprising thing about RPS that most people never learn: it has a known, provably-optimal strategy, and almost nobody uses it.

The Nash equilibrium

The game-theoretic optimal strategy for RPS is the uniform mixed strategy: choose rock with probability 1/3, paper with probability 1/3, scissors with probability 1/3, completely independently every round. If you do this perfectly, your opponent cannot beat you over a long series — and you cannot beat them. The expected outcome of every round is a 1/3 win, 1/3 loss, 1/3 tie.

This is also the strategy that is impossible to exploit. It is the only strategy that guarantees you a 50% win rate against a perfect opponent. It is also boring.

Why nobody plays the Nash equilibrium

Two reasons.

One: humans are bad at being random. When you try to "just choose randomly," you cluster — you avoid choosing the same thing twice in a row, you cycle through R-P-S-R-P-S more than you would by chance, you bias toward whichever option you played last losing. World RPS championship analysis shows that humans deviate from uniform random in measurable, exploitable ways.

Two: even if you could play perfectly random, you would have no advantage. The Nash equilibrium guarantees you 50%, but nobody ever made money by guaranteeing 50%. The prize in RPS is identifying when your opponent is not playing optimally, and exploiting the deviation.

What to actually look for

Patterns to watch for in your opponent:

• Rock bias on round 1. New players choose rock disproportionately on the opening throw. Throw paper.
• Loss → switch. When a player loses a round with X, they tend to switch to whatever beats X next round. If they just played scissors and lost, they will likely play rock next. Throw paper.
• Win → repeat. When a player wins with X, they tend to repeat X (it worked, why change?). Throw whatever beats X.
• Streak avoidance. Most players will not throw the same option three times in a row, even if doing so is correct. After two of the same, expect a switch.

These are called conditional response patterns in the academic literature and they are why best-of-5 RPS is more strategic than people realize. You only need to read your opponent correctly twice in five tries to gain an edge — the other three you can play optimal random.

The "I know that you know" regress

The classic RPS meta-game is the infinite regress: I know you bias toward rock, so I throw paper. But you know I know that, so you throw scissors. But I know you know I know, so I throw rock. And so on.

In practice, this terminates after two or three levels. Most players reason one level deep. A player reasoning two levels deep wins more often. A player reasoning three levels deep wins more often than that, but at three levels you are mostly inventing the opponent rather than reading them — so the returns drop fast.

The pragmatic rule is: pick a level slightly deeper than what you think your opponent is reasoning at, and stay there.

The verification

Every RPS match on 6proclub is fully reconstructible after the fact. We persist:

• Both players' commit hashes for every round
• Both players' reveals (choice + nonce) for every round
• The round-by-round outcomes

You can verify any past match by hashing the reveal data and confirming it matches the commits. If we ever broke the protocol, the chain would no longer verify, and anyone with a transcript could prove it. Provably fair is not a marketing line — it is a property of the system.

Where to play

Head to /rps/setup to host a private table or queue into a public match. Best-of-5 takes about three minutes if both players are fast and a bit longer if either of you is enjoying the third level of recursion.