Method
Cryptomental follows the same rules as Quantimental, applied to Base tokens.
- Nothing is invented.
- Missing data is shown as missing. A token too new to have a typical hour is called too new, not calm.
- Unusual is measured.
- A move is called unusual only against the token’s own median hour over the past day.
- ETH is the market.
- Each move is printed beside ETH’s move and the median tracked Base token’s, so you can do the subtraction.
- Events are adjacency, not cause.
- Large trades, liquidity removals and early-holder sales are listed with their timing and transaction. The copy never says “because”.
- It describes, it does not forecast.
- A test fails the build if any generated sentence uses a forecast word.
1
2
3
4
5
Risk scores
A rug pull risk score is a forecast, so it is held to a higher bar. No score appears on a token page until its backtest is published here: precision, recall and how many minutes before a rug it raised the flag, measured on tokens it was never trained on.
Quantimental once had buy and sell ratings. A five-year backtest showed they did worse than a coin flip, so they were deleted and the test was published. The same test decides whether risk scores ship.
Coverage
Uniswap V3 pools on Base paired with ETH, discovered from the past 24 hours of launches. Uniswap V4 and Aerodrome pools are not covered yet.