Hi, I’m Jatin, CEO of BlockDog.xyz - an AI powered crypto credit score. Mango’s crypto lending could be safer & smarter with BlockDog’s financial insights. We’ve been building BlockDog for 4 months now and
- have developed machine learning models trained on Aave historical data that predict liquidations with 97% accuracy.
- Support blockchains - Ethereum, Polygon, BSC, Avalanche, Arbitrum, Cronos. Solana & more incoming.
- Currently, we look at lending history on Aave V2 & Compound. Working to include Solana lending protocols like Mango as well.
- Mango lenders can increase/decrease interest rates or collateral based on the BlockDog credit score
- We look at factors like transaction history, age of wallet, previous defaults, repayment regularities, existing loan positions and many more features to score a wallet.
We’d love to collaborate with Mango and any feedback from the community would mean the world to us.
Jatin: Thank you for this feedback request and information on BlockDog. It seems like an exciting and potentially value-adding product.
Two quick questions come to mind as I dig into the project:
- How are the credit scores benchmarked versus others across the ecosystem?
- Example: A credit score of 750 may be pristine if all others are at 500, but may be sub-par if the peer average is 900. Any color or examples here would be helpful.
- Is it possible to aggregate user wallets across applications and even chains to create a more comprehensive scoring methodology?
- Example: If the same human is liquidated frequently on an Ethereum wallet / protocol, it is likely that trend could continue when they are using a Solana wallet / protocol. The same would apply to dapps; frequent Aave liquidations may indicate frequent Curve liquidations for example.
- Admittedly, this may not be technically feasible, but I wanted to pose the question all the same.
Thank you again. I look forward to continuing to review.
Note: Written entirely in my personal capacity and not representative of any organization.
Appreciate your response, August.
How are the credit scores benchmarked versus others across the ecosystem?
Our credit score is proportionate to the percentile the wallet lies in for a particular ecosystem. We obtain individual percentiles/credit score from each of the lending protocol/wallet and weight them accordingly based on the total no. of wallets in each protocol. For ex. a BlockDog credit score for a wallet that has activity in Aave and Compound, we check its percentile in each and since Aave has more wallets than Compound, we assign more weightage to the Aave score accordingly before aggregating the final score.
Is it possible to aggregate user wallets across applications and even chains to create a more comprehensive scoring methodology?
Yes. Metamask gives you the same address for all EVM compatible chains. They are still different addresses though since they are located on different chains. But the good thing is, since BD is looking at multiple chains (6 currently and working on integrating top 20) and also different lending protocols on those chains, their credit score does get affected accordingly. So, we do aggregate history across chains and dapps.
For non-EVM wallets like Solana, besides looking at txn & liquidation history on Solana, we plan to check bridge interactions like Wormhole (along with the aggregation of the credit score for the corresponding EVM wallet)