Skip to main contentTrepa’s payout system handles special scenarios that require exception rules to ensure fair distribution. These edge cases occur when standard payout calculations would produce unexpected or unfair results.
Case 1: Best predictors with majority stake
What happens: When the best predictor (or a group of best predictors with the same minimum error) has collectively staked at least 50% of the total stake volume, their minimum error represents the stake-weighted median error by definition.
The problem: Winners are normally defined as players whose errors are smaller than the stake-weighted median error. However, in this scenario, no other players exist with smaller errors than the best predictors.
The solution: Trepa implements an exception rule that declares those players with the minimum error as winners under these circumstances, so they receive payouts from the remaining players even though they technically represent the median themselves.
Case 2: Fee waiver when winning stakes dominate
What happens: In a sub-case of Case 1, when the minimum error winners have collectively staked way more than 50% of the total stake volume, the winning stakes are so large relative to the losing stakes that the remaining losing stakes (after the protocol fees/takeout is deducted from them) are insufficient to cover the gains of the winners.
The problem: Normally, winners pay their portion of the protocol fee from the dividend pool. However, in this scenario, the dividend pool (which comes from remaining losing stakes) cannot cover the protocol fee for winners, so winners would receive less than their stakes back.
The solution: Trepa waives the protocol fee on winners when the remaining losing stakes (after the protocol fee is deducted) are insufficient to cover the protocol fee for winners. In this scenario, only losing stakes pay the protocol fee, so winners are guaranteed to have a non-negative (positive) ROI.
Case 3: Refund when losing stakes dominate
What happens: When losing stakes are much larger than winning stakes, only a portion of the losing stakes is used for payouts. The excess losing stakes beyond this limit are refunded to losing players.
The problem: Normally, all losing stakes contribute to the dividend pool. However, when losing stakes far exceed winning stakes, the effective prize pool (the losing capital available for payouts) is capped to prevent excessive payouts relative to winning stakes. We essentially cap the capital-at-risk in such scenarios.
The solution: When losing stakes exceed the cap, the surplus losing stakes (the difference between total losing stakes and the effective prize pool) are refunded to losing players pro-rata based on their stake share relative to total losing stakes. 
