On May 27, 2026, the Department of Justice charged a Google engineer, Michele Spagnuolo, with making more than $1.2 million on Polymarket as the user “AlphaRaccoon” — allegedly by trading on confidential Google data. The charge made headlines for a day and moved on.
But the genuinely interesting part isn’t the charge. It’s that the entire alleged crime has been sitting in public this whole time — on-chain, immutable, fully reconstructable — for anyone who knows how to read it. Polymarket settles on Polygon, which means every bet AlphaRaccoon ever placed, the price he paid, the day he placed it, and exactly what it paid out is permanently visible. So we pulled his complete trading record.
#What he actually knew
Every December, Google publishes its “Year in Search” — the rankings of the most-searched people, shows, and topics of the year. Polymarket runs prediction markets on the outcome: “Will [X] be the #1 searched person on Google this year?”
A Google engineer with internal access doesn’t have to guess. He can see the real trend rankings — who actually topped the list — weeks before the public reveal. On those specific markets, he wasn’t forecasting. He was trading with the answer key already in hand.
That’s the edge. Not a model, not a hunch, not being early — just knowing the result before anyone else could.
#The complete record
We reconstructed all 92 markets this one wallet ever traded, split into the ones where he allegedly held Google’s internal answer key versus everything else. Here’s every bet, the day he placed it, and what it made or lost:
The Real Data Behind Google's Polymarket Insider
A Google security engineer allegedly read the company’s confidential “Year in Search” rankings — the data showing who was really the most-searched person of 2025 — then bet on Polymarket before the public knew. On the bets he held the answer key for, he was superhuman. On everything else, he lost money.
The two-faced trader
Polymarket records every trade on-chain. Split this one wallet’s bets into the markets where he held Google’s internal answer key vs. everything else, and two completely different traders appear.
One vertical line called d4vd
Running P&L across all 92 markets in chronological order. The account drifts sideways and underwater for most of its life — then the Year-in-Search markets resolve in late 2025 and it vertical-launches. Gold dots are insider-lane bets. Hover the line.
Every insider-lane market
He bet NO on the obvious favorites (Trump, Pope Leo XIV, Bianca Censori as the #1 search) because he knew the real #1 was an artist named d4vd — and bet YES on d4vd at 3¢. That single market turned ~$10.6k into ~$205k. Sort the cards below.
Found blind, then matched to the charges
How this wallet was identified
This account was surfaced from on-chain behavior alone — before any name was attached — by scanning Polymarket’s public trade ledger for the signature of an informational edge: tiny entry prices, an extreme win rate concentrated in one narrow market family, and conviction-sized stakes against the crowd. The Google “Year in Search” cluster lit up immediately.
Both figures reproduce the publicly stated charges to within ~1% straight from the chain — independent confirmation that this is the right wallet and the right set of bets. A key nuance: the whole account looks unremarkable (about +19% ROI overall). The insider signal only appears once you isolate the specific lane. That is the real lesson for spotting insiders — score the slice, not the wallet.
Research / educational. Built entirely from public on-chain Polymarket data and publicly reported charges. Identity per DOJ/CFTC public charging documents (2026-05-27). Allegations only; charges are not convictions.
#The tell
If you had to point at one number that exposes the whole thing, it’s not the $1.2 million. It’s the contrast.
On the 23 Year-in-Search markets — the ones where he allegedly held Google’s internal answer key — AlphaRaccoon won 22 of 23, a 95.7% win rate, +$1.2M profit, +44% ROI. On the other 69 markets he traded — crypto, sports, elections, geopolitics — he was a net loser: roughly a coin-flip win rate, −13.5% ROI, including a single −$276k loss on an Ethereum ETF market.
Same person. Same account. Same risk appetite. The only thing that changed was whether he already knew the answer.
The standout bets tell the story cleanly:
- He bet YES on an unknown artist named d4vd at 3¢ → roughly +$194k, about a 19× return, because he knew d4vd was the real #1 searched person — long before the public would have guessed it.
- He bet NO on the “obvious” favorites — Trump, Pope Leo XIV, Bianca Censori — at high prices, because he knew they weren’t actually #1.
- His lone insider-lane loss was Kendrick Lamar (YES @ 6¢, −$11.9k) — so 22 of 23.
That is what insider knowledge looks like in data: superhuman in exactly one narrow lane, and thoroughly mediocre everywhere else.
#How we found it
We didn’t start from his name. We started from the DOJ’s public facts — roughly $2.75M staked, ~$1.2M profit, October–December 2025, on markets tied to Google information — and went looking on-chain.
The approach: scan Polymarket’s full on-chain trade ledger and rank wallets by a behavioral fingerprint — total stake on the relevant lane, redemption-aware profit, win rate, account age, and concentration. One wallet matched all four DOJ numbers to within about 1%:
0xee50a31c3f5a7c77824b12a941a54388a2827ed6
— an account first active in May 2024.
The crucial nuance is in the methodology sidebar above: the whole account looked unremarkable, only about +19% ROI across everything. The insider signal does not appear at the wallet level. It only appears when you isolate the specific lane he had inside information on. That’s the real lesson for spotting insiders on prediction markets: score the slice, not the wallet.
#The trail doesn’t disappear
Prediction-market insider trading leaves a permanent, public, machine-readable trail. The DOJ needed Google’s internal records to prove intent — to show the engineer had access to the rankings and acted on them. But the footprint — the tiny entry prices, the lopsided one-lane win rate, the conviction stakes against the crowd — was visible on-chain the entire time, months before any charge was filed.
The forensics didn’t catch a clever crime. They caught an obvious one that simply hadn’t been read yet.
Research / educational. Built entirely from public on-chain Polymarket data and the publicly reported DOJ/CFTC charges of 2026-05-27. Michele Spagnuolo has been charged, not convicted; all characterizations of intent are allegations per the filing.