Skip to main content

Prisme · blind Bitcoin Arena

Blind Bitcoin chart game: your forecast scored with the Brier score

The Prisme Arena is a free game, no account needed, that tests how well you read a Bitcoin chart blind. A 48-hour window drawn from the Binance BTCUSDT archive since 2017 is shown without a date, rebased to 100. You call rise or no rise with a confidence from 50 to 99%; your answer is locked, then scored with the Brier score.

Play the Arena (free, interface in French) →

Bitcoin only. Exploration, not a verdict.

Prisme blind Arena after the reveal: the T0 line, the move at the chosen horizon and the Brier score
After the reveal: the T0 line, the move at the chosen horizon and the Brier score.

How a round works

How the Brier score is computed

Brier = (probability of a rise − outcome)², with outcome 1 for a rise and 0 otherwise; an unchanged close counts as "no rise". The lower the score, the better.

Example: announcing a 70% chance of a rise scores 0.09 if the rise happens and 0.49 if it does not. Always answering 50% scores 0.250. Rounds overlap in time, so they are exercises, not independent observations.

Why play blind?

On a chart you recognise, memory does part of the work. Hiding the date and the price level removes that help, and the Brier score rewards calibrated confidence rather than bold calls: a confident miss costs more than a cautious one. Prisme treats what you think you see as a hypothesis; only a test on data the rule has never seen can say whether it holds.

Frequently asked questions

Is the Prisme Arena free?

Yes. It is free and needs no account. An anonymous session keeps your rounds; no one else's forecasts are shown to you.

Can I recognise the period on the chart?

The window is drawn at random from the Binance BTCUSDT archive since August 2017, with dates hidden and the price rebased to 100, so the period is harder to recognise. The continuation stays on the server until the reveal.

What is a good Brier score?

Lower is better. Always answering 50% gives 0.250, so a lower average means your stated probabilities matched the outcomes better than a coin-flip answer. Saying 70% for a rise scores 0.09 if the rise happens and 0.49 if it does not.

Is the Arena a trading signal or a skill ranking?

No. It is an exploration exercise, not a verdict: there is no leaderboard and no talent diagnosis, and rounds overlap, so they are not independent observations. To test a trading rule on data it has never seen, use Strategy Arena Lab.

Is the game interface in English?

The game interface is in French for now. The rules fit in one sentence: look at 48 hours of Bitcoin, choose rise or no rise with a confidence from 50 to 99%, then read your Brier score after the reveal.

Go further

Data: frozen Binance BTCUSDT archive in 15-minute candles, since August 2017. The game runs on a Rust/WebAssembly engine, with WebGPU and a Canvas fallback. Educational exercise, not investment advice.