How Scrynx works

Read-only truth score. Starts at 100, then subtracts wash trading, linked funding, bundles, sniper timing, thin turnover, and holder concentration. Missing signals keep fixed weights (not reweighted). Confidence shows how much of the model had data. Low score = warning, not proof of fraud.

Labels

70–100 Likely real · 40–69 Mixed · 0–39 Likely fake

A severe signal sets a floor. Below ~50% confidence, “Likely real” is withheld. Use Deep scan for more tape + funding.

Signals and weights

Wash trading 35%

Multi-flip wash or a single flip that closes ≥90%. Round-trip SOL-eq ÷ total = fake-volume share (USDC legs counted).

Shared funding 20%

Same funding ancestor within 3 hops and 10 minutes. Known CEX/bridge funders ignored.

Bundled launch 15%

Linked wallets buying in the first 3 blocks (~8s by timestamp), plus creator sells. If launch isn’t in the fetched tape, scores the earliest observed buys instead.

Bot timing 15%

Buy volume piled into one slot, or many wallets buying in the first seconds after launch.

Volume vs liquidity 10%

24h volume far above pool depth — a common wash / recycled-flow pattern.

Holder concentration 5%

Top-10 wallets holding a large share of supply after launch.

Not in the number

X / Telegram / Discord hype
Opinion or third-party “trust scores”
Anything that needs signing or sending txs

Data coverage

Live checks pull DexScreener + Helius under a short budget. Missing launch/funding history is marked not scored in the coverage watermark.

← Back to scoring