← blog · 2026-09-09 · 3 min read
What 30 days of fomo tape data say about following the crowd
We backtested every tracked fomo.family buy over 30 days on Robinhood Chain. Median outcomes are negative, crowding is the strongest negative signal, and the only consistent edge is being within two minutes of the first ranked buyer.
hoodwatch records every fill by roughly two hundred tracked fomo.family wallets, with the price of each token at one, six and twenty-four hours after every entry. That lets us ask a blunt question: if you copied what the tape showed, what would have happened? Below are the results from the 30 days ending in early September 2026. All numbers are medians unless stated.
The base rate is negative
Take every token at the moment the first tracked wallet bought it. Six hours later the median price was about 30 percent lower. Most launches fail. Any strategy that buys everything the tape shows loses money on median, and no amount of ranking the traders fixes that on its own.
Crowding is anti-predictive
The intuitive signal, several good traders in the same token, is the worst one in the data. Tokens with two or more ranked wallets already in, or three or more tracked buyers, went on to lose 60 to 75 percent on median. The mechanics are simple: the follower wave is the exit liquidity for the wallets that were first.
This also showed up in our own research scores. The first version of the language-model research desk gave its best verdict, "consider", to tokens with strong crowd signals. Live calibration over about 270 reports found "consider" was the worst bucket: down 57 percent on median with a 19 percent hit rate. We rebuilt the scoring around it.
The edge is at the first ranked buy
Measure from two minutes after the first ranked wallet's buy, restricted to tokens where the smart-money and entry components of our tape score were strong:
| metric | value |
|---|---|
| median 1h return | +34% |
| hit rate (positive at 1h) | 59% |
| reached 2x within 24h | 34% |
Two minutes is roughly the time it takes to see the fill, check the honeypot and liquidity flags, and act. After that the crowd arrives and the numbers above degrade quickly.
Fat tails pay for everything
About 15 percent of tokens reached 5x within 24 hours of the first tracked buy, and about 10 percent reached 10x. The median drawdown before a 5x was mild (0.95x), meaning a wide stop would not have shaken you out of the winners; it would only have cut the losers, which lose anyway.
That shaped the exit rules we now use for the paper desk: no stop, 24-hour maximum hold, sell half at 2x, then trail the rest at 40 percent below the peak with a floor at 1.3x entry. Across the backtest that beat every fixed stop-loss and take-profit combination we tried.
What we do with this
- The tape marks FIRST BUY only for ranked or alpha wallets, and shows run-up since that buy, so late is visible.
- The triage board sends tokens with three or more ranked buyers or a 2x run-up to a LATE lane, and reserves TICKET for pools under 30 minutes old where one ranked wallet is in below 2x, or at most two buyers are in below 1.3x with the crowd still buying.
- The research desk now asks the model one question about a ticket ("is this launch real?") instead of letting it score momentum.
Numbers will drift as the market changes; we recompute the calibration continuously and show it on the analytics page. The golden-ticket playbook that came out of this is in The golden ticket playbook.
hoodwatch is unofficial and read-only, not affiliated with fomo.family or Robinhood. Nothing here is financial advice. Open the live tape.