Rock Paper Neural is a free browser rock-paper-scissors duel against a small multilayer perceptron that trains on your move history. After a few rounds it tries to predict your next pick. About ~20 KB after Play, no install, no account.
Want a cannon duel vs a neural net instead? Try Bangbang Artillery. Want an optimal-stopping puzzle? Try Googol Stopping Game.
Pick Rock, Paper, or Scissors. A tiny neural net watches your sequence and tries to beat your next move. Keyboard: J / K / L. Adapted from Victor Ribeiro's jokenpo (MIT). About ~20 KB.
Works with click/tap or keyboard. Session-only - no save data.
Rock Paper Neural pits you against a learning opponent right in your browser - free, with no account and no install. The longer you play, the more the net learns your habits and starts countering them.
Key features
- A quiet break against a learning opponent - nothing to sign up for, and only about 20 KB loads after you press Play.
- The net predicts your next pick from your recent history, so you notice your own move habits the moment it starts countering you.
- Session-only by design: scores and training stay in this tab and clear on refresh, with nothing written to localStorage.
- Adapted from Victor Ribeiro's jokenpo under the MIT license.
When to play
Reach for it as a quick mental break, to spot the patterns in how you throw Rock, Paper, or Scissors, or to compare tiny client-side AI toys on this site.
Privacy
Everything runs in your browser tab. Your scores and the model's training stay on this page only until you refresh, and nothing leaves your device.
For more free browser games, open the games hub. Related neural-net toys on this site include Bangbang Artillery (an aim-and-power neural duel) and the decision puzzle Googol Stopping Game.
Frequently Asked Questions
What is Rock Paper Neural?
A free browser rock-paper-scissors game where a small neural network trains on your move history and tries to beat your next choice.
How do I play?
Click or tap Rock, Paper, or Scissors. On a keyboard press J for Rock, K for Paper, L for Scissors. Scores for player, computer, and ties update each round.
When does the AI start learning?
The first few rounds are mostly random. After enough history, the net predicts your next move and picks the counter. It only trains when it did not win the previous round.
Does it save my history?
No. Training is session-only in memory. Refresh clears the model and scores.
How big is the download?
About ~20 KB of HTML, JavaScript, CSS, and a small sprite sheet after you press Play.
Is this open source?
Yes. Adapted from jokenpo by Victor Ribeiro under the MIT license. This site build strips the service worker and demo GIF, adds noindex, and ships LICENSE plus CREDITS next to the engine.