Methodology

• Base mainnet logs via free public RPC endpoints (primary mainnet.base.org) — every Transfer, Swap, Initialize/PoolCreated, and liquidity event since the token's creation block. No third-party analytics API sits between the chain and these figures.
• Chainlink ETH/USD on Base (aggregator ↗) — the only fiat input; carried last-observation-forward for USD marks.
• A self-hosted Postgres database holding the raw event tables, keyed and ordered by (block_number, log_index) — the chain's only total order.

Cadence: pools and trades ingest continuously; candles refresh ≈15 seconds behind head, trader statistics ≈60 seconds. Every panel carries an "as of" stamp with the exact block.

Basecat is a B20 native token: balances live in native state, not contract storage, so storage-enumerating explorers under-count holders. This site reconstructs every balance by folding the complete Transfer log from block zero of the token's life. Two standing proofs keep that honest:

• Fold == balanceOf, wei-exact. A pinned sample (top-10 balances plus hash-selected wallets) is compared against live balanceOf at the cursor block on every verification run.
• Supply closure. The sum of all folded balances must equal exactly total supply minus burns — an equality, not a tolerance.

House rule: internal checks are equalities wherever the data admits one. A tolerance would swallow a real accounting bug forever.

Each hour is priced by the highest-USD-volume eligible pool that hour; sub-hour candles inherit the hour's winning pool, so a minute-level rollup reproduces the hourly series exactly (asserted on every verification run). Volume sums all pools. The headline spot price is the dominant pool's last print — a $2 swap in a dust pool cannot set the headline. USD uses the Chainlink feed, last-observation-forward. Unpriceable pools contribute no USD by refusal rather than by guess. As a cross-implementation sanity band (not an equality — different fx and bucket edges), the hourly series is compared against GeckoTerminal's independent OHLCV for the launch pool.

• Wallet attribution. A swap's sender is always a router, so the real wallet is recovered by folding the trade transaction's own transfers: routing hops cancel, and the largest surviving mover on the trade's side is the wallet. Smart-wallet (ERC-4337) bundles can settle several users in one transaction — attribution is "the biggest mover in the tx", and is labeled as such wherever shown.
• P&L is flow-basis. Net token delta per wallet-block, valued at that minute's price, average-cost basis. Coins received by plain transfer count as entries at market. This is a consistent, reproducible convention — not a claim to know anyone's true cost.
• Clusters are evidence, not proof. Wallets linked by large swapless transfers (hub-guarded) are grouped as one probable owner; an OTC deal between strangers would draw the same edge.

Certainty grades used across the site: exact (an equality against raw logs), estimated (a stated model over exact inputs), projected (extrapolation, rarely used and always labeled).

• Liquidity for v4 pools is partly estimated (swap-flow folds cannot see LP deposits directly); the pools page labels which legs are live reads and which are estimates.
• USD marks depend on the Chainlink cadence; sub-minute fx moves are invisible.
• Attribution inside ERC-4337 bundles is the largest mover, not a notarized identity.
• Wallet names are curated social handles or reverse-ENS — labels, never identity claims.

Corrections policy: a demonstrated error is fixed and the fix is stated where the number appears. See about for contact.