Yuma Consensus: turning many judgments into one payout

Many judges grade the same competitors differently; this lesson explains how the network turns that pile of separate opinions into one payout.

Every validator on a subnet submits its own weights for the same set of miners, and they rarely agree exactly. Yuma Consensus is the network's rule for turning that pile of separate opinions into one official ranking, run automatically at the end of every judging period. [source]

Not every opinion counts equally: a judge's say is scaled by how much stake stands behind it, so someone heavily backed moves the outcome more than someone lightly backed. [source]

The network also guards against any one judge inflating a favorite: for each contestant it finds the middle grade that a majority of backing supports, and pulls any grade above that back down to it — so a single opinion cannot drag a payout up on its own. [source]

What survives that process becomes each contestant's rank, and normalized ranks decide their actual share of the contest's payout to the side that does the work. [source]

Judges are paid the same way, from the other side: one earns more by recognizing good work before the rest of the network catches up, and less by lagging or by grading out of step with where the majority lands. [source]

If a contest somehow ends up with no valid grades at all, payouts do not simply stop — the network falls back to splitting that period's payout by how much backing each participant has instead, so nobody is left with literally nothing. [source]