How CobWeb Works
Every score on this site is a heuristic over public on-chain data. Here is exactly how each one is computed, so you can judge them yourself.
Early Buyers
A wallet counts as an early buyer only if it received the token AND spent SOL in the same transaction — this filters out airdrops, LP seeding and DEX vault legs. For tokens covered by our real-time index, the order and launch slot are exact; otherwise they are reconstructed from up to 3,000 transactions of history (oldest-first).
Bot Score
Based on how many blocks after launch the wallet bought: under 3 blocks scores 85 (near-certain automation), under 10 scores 65, under 30 scores 35, later entries score 12. Buying within seconds of a launch is mechanically impossible for a human reading a ticker.
Smart Money Score
Starts at 20. Organic timing (30+ blocks after launch) adds 15. Conviction adds up to 30: a position at least 2x the median of fellow early buyers signals independent research, not noise. Independence from any detected cluster adds 10. Wallets flagged as bots score 0; members of suspicious clusters cap at 15 — coordinated buying is not "smart", it is inside knowledge. The Smart Money tab shows wallets scoring 55+.
Cabal Clusters
Three connection layers: temporal (3+ wallets buying within the same 60-second window), common funder (wallets whose first SOL came from the same address — known exchange hot wallets like Binance or Bybit are excluded, since a shared CEX is not a shared owner), and direct transfers between early buyers. Suspicion score grows with cluster size and connection density.
Dev Risk
The deployer wallet is profiled for prior token launches and quick-sell behavior. A deployer with a history of launching and dumping scores HIGH regardless of how the current token looks.
Lifecycle Stage
From the latest ~100 transactions: buy ratio ≥62% reads as markup, ≤38% as dump; the balanced middle is accumulation when unique buyers outnumber sellers and distribution otherwise. Fewer than 5 classified trades returns unknown — we do not manufacture a verdict from noise.
Market Data
Listed tokens are priced via DexScreener. Fresh bonding-curve launches not on any DEX get a volume-weighted price derived from their actual recent on-chain trades, with outliers beyond 4x of the median trimmed (MEV, dust). If nothing trades, we show a dash — never an invented number.
PnL & Copy-Trade Simulator
Realized PnL is computed from actual SOL flows across the last 500 transactions, parsing both standard swap events and raw pump.fun transfers. Unrealized PnL prices open positions via DexScreener; positions without a price are reported as unpriced rather than counted as zero. The simulator replays trades within the chosen window and reports realized ROI — no slippage modeling, open positions excluded. None of this is financial advice.