XXXX ETH transferred from an unknown wallet to Coinbase Institutional. Can we gain more insights beyond just 'unknown wallet'? As a data science engineer, I've embarked on a solo project to analyze the correlation between whale transactions and subsequent market dynamics. I am not a financial and crypto market expert, so I post my approach here for constructive suggestions, and once the project is completed, I will make the application publicly available.
My initial analysis focuses solely on the ETH mainnet (with plans to expand to other chains and DEXs if the approach proves meritorious). I examine how large transfers correlate with market movements after a certain time delay.
The approach delivers market insights in two ways: visualization and alerting:
figure 1: the lower left plot of figure 1 shows a potentially monotonic relationship between a wallet’s transactions and the market log return after a 5-hour lag.
- Visualization: Analyzing and plotting a wallet's transaction history against market responses over specific time delays (ranging from 1 hour to several days, selected using a cross-correlation algorithm to identify the most significant correlation delay).
- For example, for a given whale address, the lower left plot of figure 1 reveals a potential monotonic relationship between the wallet’s transaction volume and the market log return after a 5-hour lag. Inflow transactions may precede market gains, whereas outflow transactions could indicate losses.
- A color-coding scheme highlights the temporal nature of transactions, with yellow-ish dots for recent transactions and purple-ish dots for earlier ones.
- Alerting System: Working with services like Chainbot or Whale Alert, my system aims to notify users of significant fund movements, linking historical transaction data with market changes. For example, “On XX-Jan-2024, an outflow of XXXX ETH from the address preceded a market change of XXX% after a delay of XXX hours”. Determining which transaction events are significant enough to report, however, remains a challenge.
There are several limitations to this project:
- I've only analyzed ETH mainnet on-chain data since January 1, 2021, focusing on transactions equal to or exceeding 1 million dollars. I treat all transactions the same, but when involving smart contract interactions, the impact can vary significantly. Additionally, significant ETH-related activity on other chains and decentralized exchanges (DEXs) is not captured.
- I'm still in the process of gathering labels for whale addresses to determine whether they belong to an organization or exchange.
- The historical balance of wallets has not been retrieved yet.
I welcome constructive suggestions and criticism of my approach. If it proves valuable, I will regularly report on development progress and eventually make the analysis tool publicly available.
