Accuracy weight - Trepa Documentation

Among winners, the closer you were to the outcome, the larger your share of the prize pool. The system compares your to the and assigns a weight; better predictions get a higher weight. The curve is steep—a small edge among winners can move a lot of dollars. Dollar tables: Payout overview; then Capped proportional payout applies the per-round profit ceiling.

Line chart titled Accuracy-weighted pool allocation: accuracy weight on the vertical axis vs outcome on the horizontal axis, with a peaked green curve highest at the true outcome and annotation that most accurate predictions capture larger shares of the pool.

Formulas (for verification)

Normalized error (your error vs median): $r_i = e_i / m$ , where $e_i$ is your error and $m$ is the median error.
Accuracy weight: $a_i = \bigl( \frac{1}{1 + r_i} \bigr)^\gamma$ with exponent $\gamma$ . Currently $\gamma = 6$ . Closer predictions have smaller $r_i$ and higher $a_i$ .

Last modified on May 4, 2026

Capped proportional payoutHow the 100× per-round profit cap works after accuracy-weighted splits.

⌘I

Formulas (for verification)