● Reference · Methodology 2.0.0
Trust Score
ten signals, each with its evidence.
How PARACHUTE scores a token: what it measures, how the signals are weighted, and when it refuses to give a score. Everything is read live from the chain in the trader's browser. A score is information, not advice.
01Two speeds
- Quick Trust on every feed row, computed from the one multicall the page already makes.
- Deep Trust on the token page: 10 signals, each with its evidence (the numbers, the block range scanned, the transactions). It is cached per block range. On the public RPC it takes 10 to 40 seconds.
02The ten signals
| Signal | Weight | What it measures |
|---|---|---|
| Can you sell? | 15 | A simulated sell with eth_call: does it revert, and what does a small buy-and-sell-back cost in fees, taxes and price impact. |
| Holder concentration | 20 | Share of supply held by the top 10 wallets and the largest wallet, outside protocol contracts, from Transfer logs. |
| Creator and team selling | 15 | How much of what creator and team wallets took in they sold or moved to other wallets. |
| Launch snipers and bundles | 15 | Supply bought by team and other wallets in the first 10 seconds after launch. |
| Creator track record | 15 | The creator's earlier pons v2 launches: how many, how many graduated, stalled or were rescued. Serial launchers (20+) score low. |
| Bonding curve and graduation | 5 | On the curve, ready to graduate, swept, graduated, or rescued. |
| Trading fees and taxes | 5 | Protocol fee plus creator tax on every trade. |
| Uniswap v4 pools and hooks | 5 | Whether the graduated pool is confirmed and whether any v4 pool uses an unknown hook. |
| Liquidity lock | 3 | Whether the pool position sits in the pons launch locker. |
| Creator fee recipient | 2 | Changes of the creator fee recipient, and whether one is pending. |
Each signal scores from 0 to 1 using measured numbers only, with a verdict of good, warning or bad. The weights and scoring functions are in src/features/trust/scoring.ts and are unit-tested.
03From signals to a score
- Unknown is not guessedA signal that could not be measured is "unknown" and left out; coverage drops instead.
- Not applicable is left outSignals that do not fit the token's phase (for example the liquidity lock before graduation) count neither in the score nor in coverage.
- Weighted meanThe score is the weighted mean of the measured signals, from 0 to 100.
- Minimum coverageBelow 60% of the applicable weight, no score is shown ("not enough data"). Coverage is shown as a percentage.
- Hard capsA sell simulation that reverts caps the score at 25. A rescued graduation caps it at 30.
- GradeA from 80, B from 65, C from 50, D from 35, F below.
04The maturity gate
Absence of bad evidence on a new token is not evidence of safety, so young tokens are treated with care:
| State | When | Effect |
|---|---|---|
| Too early | Younger than 15 minutes, or fewer than 10 holders | No score at all ("too early to judge"), with the reasons |
| Young | Younger than 6 hours, fewer than 50 holders, holder count unknown, or under 85% of signals readable | Score capped at 64 (grade C at best), with the reasons |
| Mature | Otherwise | No cap |
05Limits
- The score describes on-chain facts at the time of reading. It cannot see intentions, off-chain agreements or wallets that are not linked on-chain.
- Holder scans are bounded; the evidence states the range scanned.
- A high score is not a recommendation and does not mean a token will keep its value. Memecoins are highly risky; most lose value.