Skip to main content
A “Sybil attack” is the oldest game in on-chain markets: one person, many wallets, one disproportionate share of the prize pool. Most bracket pools, prediction sites, and tournament products have no defense against it. Prediction Arena does.

​
In tournaments

Tournament entrants commit their entire bracket upfront, anchored by a single sealed cryptographic commitment, before the first game is played. After the tournament starts, no entrant — across any number of wallets — can change their picks based on new information. Combined with path-dependent scoring, this breaks the wallet-spam economy:
  • Spinning up 50 wallets costs 50 entry fees. Sybils pay for their own attack.
  • Each of those 50 wallets must commit a complete bracket before round one. You cannot wait to see how round one plays out and then “decide” your round-two picks across 50 wallets — your picks across all wallets are locked the moment the tournament starts.
  • Random “spread” brackets score badly. Path-dependent scoring means your later picks only count if your earlier picks were also right. A scattershot strategy across many wallets produces many wallets with mostly-broken brackets, not many wallets with high scores.
The dominant strategy in an Prediction Arena tournament is be good at predicting the bracket. Spamming wallets is not just unhelpful — it is actively unprofitable.

​
In asset markets

Asset markets are inherently more Sybil-resistant because they are pari-mutuel: the pot is the pot, and a Sybil’s stake on side A is just as valuable to the winning side as anyone else’s. There is no “rebate,” no “first-mover bonus,” no “loyalty tier” that a Sybil could farm by splitting one stake across many wallets. We deliberately do not run any user-level loyalty program, points program, or first-N-bonus that would create a Sybil incentive in the first place.

​
In general

Our north star for any new feature: does this give a 100-wallet user any advantage over a 1-wallet user with the same total stake and the same skill? If the answer is yes, we don’t ship the feature.