EVM Log Correlation Pattern Detection

Problem

With only a known EVM address, raw log rows are often fragmented: it is hard to tell which contract_address and topic0 event combinations are strongly associated with that address, or whether interactions are concentrated in a small set of counterpart wallets and contracts. EVMLOG turns that investigation into a repeatable statistical workflow: start from one address, then surface likely event patterns and candidate transaction paths.

How It Works and Where It Applies

The skill first pads the address to a 32-byte topic form, then queries a Dune log table in two directions: topic1=A* and topic2=A*. It aggregates results by (contract_address, topic0), counting hits, top counterpart addresses, top1_ratio, and gap_ratio to separate strong patterns from weak ones. Weak patterns can be expanded by increasing top_n and checking from/to context; strong patterns are followed by sample transaction hashes and event-sequence examples.

It fits cases where the address appears in topic1 or topic2, such as ERC20 Transfer, Approval, NFT transfers, DeFi custom events, and bridge events. It does not decode ABI by default, so topic0 to event-name mapping may need separate enrichment; very large ranges should be narrowed or run with a larger query tier.