Skip to main content
Each round is pari-mutuel: your payout comes from everyone else’s stakes. Every round has a median error cutoff that splits participants into winners and losers.
  • You win if your error is below the round’s median error. You get your stake back, plus a share of the prize pool by accuracy.
  • You lose if your error is at or above the round’s median error. Your stake funds the prize pool, you get nothing back.

​
How much you can get

Among winners, the most accurate prediction in a round can turn a stake into several times its value. A prediction that barely beat the median gets a modest slice on top of its stake back. A small edge near the top of the pack can be worth a lot more than the same edge further down. Losers get nothing back, no matter how close they were to the cutoff. For participant ii with stake sis_i and error eie_i, weigh accuracy relative to the round’s median error: ri=eimedian error,ai=(11+ri)5r_i = \frac{e_i}{\text{median error}}, \qquad a_i = \left(\frac{1}{1+r_i}\right)^{5} If you win, your payout is your stake back, plus your share of the prize pool in proportion to your weight wi=ai×siw_i = a_i \times s_i against everyone else’s: payouti=si+prize pool×wi∑jwj\text{payout}_i = s_i + \text{prize pool} \times \frac{w_i}{\sum_j w_j} If you lose, payouti=0\text{payout}_i = 0.
The cutoff is relative, not absolute: it’s your error against that round’s median error, not a fixed dollar or percentage bar. So the same error that would win one round could lose another, if the rest of the field was tighter or looser that time.

​
Where the take goes

Losing stakes fund the prize pool, but a 20% take comes out first: 75% funds Trepa revenue, referral incentives, and partner integrations; 25% goes to the streak accumulator, which pays back to players.
Last modified on August 4, 2026