TVL for uncollateralized lending protocols

[0xngmi]

Uncollateralized lending protocols such as Maple work by allowing everybody to lend money to a pool, and then a whitelisted set of entities, usually market makers, can borrow from that pool without requiring collateral by taking loans that are enforced through legal contracts.

The problem of counting TVL comes from looped borrowing. These are just illustrative numbers but Maple was offering 5% APR on their pools + 15% APR in token incentives to lenders of the pool, while borrowers paid 10% interest and got 5% in token incentives. With these APRs, if a borrower took a loan and deposited the money in the same pool they would be paying 5% in interest while earning +20% in tokens, thus earning a net of +15%.

Because of how uncollateralized lending protocols are structured, users could deposit 1M, then a borrower could just sign a legal contract, get approved for a 10M loan ceiling, and then just proceed to borrow 1M and deposit 1M, repeating this process 10 times until they reach their borrow ceiling. By doing this they'll earn 1.5M/year (15% apr on 10M) without having to provide/lock any capital of their own nor take any risks, so it's very profitable for borrowers, however in doing so they are inflating the TVL by 11x (it will look like 11M have been deposited by users but only 1M has actually been deposited).

This is not an hypothetical, there's been two cases where borrowers on Maple have been found to do this, and in private talks with other market makers I've been told that others are engaging in this too, it's just that they don't just send the funds to another address and loop from there (what happened in the cases that got independently discovered). Furthermore, this has even happened in other protocols that didn't have token incentives, presumably because the borrowers were speculatively farming airdrops.

A simple solution to this would be to add a clause to the legal contracts that forbids borrowers from depositing the money into the protocol from which they borrowed from. The whole protocol relies on those legal contracts, so you could use them to enforce that condition as well. I've discussed this with the Maple team and they don't want to add this.

So the situation is that this group of protocols make it possible to easily inflate their TVL, TVL inflation has already happened before multiple times across multiple protocols in the group, some protocols heavily incentivize it (by offering free money to do it), and it's not possible for either defillama nor outside observers to verify how much TVL is actually real vs inflated because we don't have any visibility into the finances of borrowers.

Users come to DefiLlama to find metrics that they can trust, and our job is to vet them and make sure that our data is correct and can be trusted. For this reason we chose not to to alter our definition of TVL for this group of protocols: we report TVL as money in the smart contracts of the protocol, excluding borrowed coins. This is the same methodology we use for all other protocols on defillama, including other lending protocols such as aave or compound, and we (and anybody else) can clearly verify that the money in the smart contracts and it matches our numbers.