$50.59B moved through prediction markets in a single month this summer. One venue took $37.7B of that. The category arrived.
80-90% of it is sports, per a February federal court filing. That volume funds the rest, and the rest is worth funding: a Fed working paper found macro event markets well calibrated, with a perfect record the day before FOMC decisions and CPI errors almost identical to the Bloomberg consensus. A central bank documented that an event venue prices the central bank’s own decisions about as well as professional tools do.
So the forecasting value is real. The open question is who is standing on the other side of the trade.
Right now nobody can answer that cleanly. Sethi’s wash-trade work found periods where up to 60% of volume was traders moving across wallets they controlled. Roughly 3% of traders drive most price discovery. Across $67B of volume, the top 1% of profitable users captured 76.5% of all profits. Every one of those figures is an identity question wearing a volume costume.
The category currently picks between two bad doors.
Door one is full KYC. It gives you surveillance that works, and it also means handing your passport to a venue whose resolution disputes have already produced death threats against a reporter naming where his family lives /// personally, I hate this.
Door two is pseudonymity, which buys privacy and hands you sybils, wash trading, and volume prints nobody can audit /// authorities don’t like this.
Think of zk proofs as a third door. You prove what a regulator actually needs (one human, one account, eligible jurisdiction, not sanctioned) and prove it cryptographically, without the venue ever holding the document. Sybil wash trading gets expensive because uniqueness is enforced at the credential layer. Volume becomes auditable because wallet clusters can be bounded. Concentration stats become checkable rather than inferred /// smartest approach imo.
Being honest about the limits: this does not catch the Special Forces sergeant or the teleprompter operator. Insider cases need identity unsealed after the fact, so you want selective disclosure with a court-order key, not pure anonymity. Credentials can also be rented unless they are bound to a device or a liveness check. And it does nothing for the oracle problem, where a $242M market got settled by a token with a $95M market cap and a $750 proposal bond.
My honest takeaway from putting this report together is that the forecasting value lives in a much smaller room than the valuations imply, and identity infrastructure is what makes that room bigger.
The sector is growing, but I hope it doesn’t turn into another pure gambling narrative - because prediction markets have so much more to offer than that.