Inside Autonomous Resolution: Predictions That Settle Themselves
For the first time, a supported prediction pool can be resolved by the stream itself — and refunds rather than guesses when it isn't sure.
Every prediction pool eventually needs to answer one question: what actually happened? For as long as PLLAY has run pools, a person has answered it — usually the creator, sometimes an admin stepping in on a dispute. This month, for a first supported case, that answer can come from the stream itself.
We built this in two deliberately separate phases. Phase one was a resolution-spec model plus vision-generation work that watches a stream and labels what it detects — for the first supported case, CS2's bomb-plant moment — without touching a single payout. We ran that phase against real settlements to see whether its outcome calls matched what a human reviewer would have said, before it was allowed near any money.
Only once that validation held up did phase two turn on: a worker that takes the vision system's outcome call and actually settles the pool — or, if the system isn't confident enough in what it detected, refunds every entry instead. That refund path isn't a fallback we hope never triggers; it's the design. An autonomous system that's occasionally wrong about a game state and pays out anyway is worse than one that's occasionally wrong and returns everyone's stake — so it's built to prefer the refund over a bad guess every time.
Autonomous doesn't mean unsupervised. Every automatic settlement is visible through a Resolution Console, so the team can review what the system decided and why, and the House Account view keeps the ledger it's touching fully reconciled and auditable. The goal was never to remove people from the loop — it's to remove them from the critical path of every single pool, so a creator's audience gets paid out in the time it takes a vision model to watch a replay, not the time it takes a human to be available.
CS2's bomb-plant detection is the first supported market, not the only one we're building toward. The pattern — vision detection validated payout-neutral first, then a settle-or-refund worker, then a review console watching over both — is the template for extending autonomous resolution to more games from here.