Most prediction games ask you to pick a side. Leaderboard asks you something harder and more interesting: how sure are you?
Every event on your slate takes a probability from 0% to 100%. When the event resolves, you are scored on how close your probability was to what actually happened, using a rule borrowed from professional forecasting.
The formula
Your score on each event is:
100 × (1 − (p − o)²)
where p is your probability and o is the outcome: 1 if the event happened, 0 if it did not.
A few examples, assuming the event happened (o = 1):
- You said 90%: 100 × (1 − 0.01) = 99 points
- You said 70%: 100 × (1 − 0.09) = 91 points
- You said 50%: 100 × (1 − 0.25) = 75 points
- You said 20%: 100 × (1 − 0.64) = 36 points
If the event did not happen, the same table flips: saying 20% earns you 96 points, and saying 90% earns you 19.
Why this rule
This is a Brier score, rescaled so bigger is better. It has one property that matters more than any other: it is proper. Your expected score is highest when you report exactly what you believe. If you privately think an event is 70% likely, writing down 70% beats writing down 60%, 80%, or any strategic hedge, on average, every time.
That property is what makes the game about calibration. Over a 20-event slate, the players who rise are not the ones with hot takes on one headline. They are the ones whose 70%s come true about 70% of the time.
What it means for how you play
Two practical consequences:
- Extreme probabilities are earned, not brave. 99% pays almost nothing extra over 95% when you are right, and is catastrophic when you are wrong. Reserve the edges for things you would genuinely be shocked to see go the other way.
- Every event counts. There is no folding a bad hand. An event you know nothing about is not a skip, it is a 50%, and 50% is a perfectly honest answer that protects you from disaster.
One event on your slate scores double: THE CALL. It has its own rules, and its own article.