The "fresh wallet insider" signal
A brand-new wallet funds up and immediately places one big, focused bet. Surely that's informed money. Five thousand paper copies say: no.
The hypothesisInsiders don't use their main wallet. A freshly-created wallet making a large, concentrated first bet is the classic signature of someone who knows the outcome — detect it on-chain, copy it instantly.
The test
A daemon flags wallets whose first meaningful activity is a heavy, focused position, verifies wallet age on-chain via block explorer data, and paper-copies the bet at the live price. Every copy settles at resolution. The thesis is a staple of prediction-market folklore — actual insider cases have happened — so the question was never whether insiders exist, but whether the signature selects them at a rate that pays.
The result
Across more than five thousand settled paper copies, the fresh-wallet cohort is a net loser. The signature mostly selects gamblers: new accounts making confident, wrong bets look identical to new accounts making confident, right ones — and there are far more of the former. Filtering harder on bet size, focus, and wallet age changed the volume, not the sign.
Conclusion
Genuine insider trades exist inside this cohort — we have watched a few resolve implausibly well — but they are buried under so much fresh-wallet noise that the signal as a whole is unprofitable to copy. A narrower cousin (wallets dormant for months that wake up with one large bet) tests meaningfully better and remains an open experiment. "Follow the smart money" fails when the detector can't tell smart from loud.