Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.trepa.io/llms.txt

Use this file to discover all available pages before exploring further.

Each round, you and everyone else pay the same and submit a price . After the is known, your is the absolute distance from your estimate to the outcome. Everyone is ranked by error; the (middle value when errors are sorted) is the cutoff.

The median rule and ~50% baseline

In a typical round with many players, about half end up below the median error and about half at or above it, so your baseline expected win rate is roughly ~50% before your own edge. Small pools or duplicate errors can shift the exact split slightly.
You win if your error is strictly less than the median error. If your error equals or is worse than the median, you lose under the standard rule.
The best-coalition exception only matters when half or more of the field shares the same best (lowest) error so the plain median rule would leave nobody strictly below the cutoff. That situation is extremely rare in live play; most rounds are decided only by the rule above.

What winners and losers receive

Winners get their entry fee back in full, then a share of the prize pool funded by losers’ fees (after the platform take). Among winners, the closer you were to the outcome, the larger that share (accuracy weight). Losers forfeit their entry fee; those fees fund the prize pool and platform take. Prize flow and take: Payout overview.

Formulas (for verification)

  • Error (player ii, estimate xix_i, outcome yy): ei=xiye_i = |x_i - y|
  • Median error: the kk-th smallest error with k=(N+1)/2k = \lfloor (N+1)/2 \rfloor (e.g. N=10k=5N=10 \Rightarrow k=5).
  • Win condition (standard): player ii wins if ei<median errore_i < \text{median error}. Tie at the shared minimum error for half or more of the field: best-coalition exception (rare).

Example

Setup: 5 players, outcome = 97,100.
PlayerEstimateErrorResult
P195,0002,100Loss
P296,500600Win
P397,000100Win
P498,2001,100Loss (equals median)
P599,5002,400Loss
Median error is 1,100 (3rd smallest). Only errors strictly below 1,100 win, so P2 and P3 win; P1, P4, and P5 lose.
Last modified on May 4, 2026