Total Value Locked is one of the few quantifiable metrics one can look at when comparing a broad swath of crypto infra projects - primarily L1s. It represents the dollar value invested in a given ecosystem across different Defi apps and staking services like Lido (for Ethereum). While this number may give a gauge of interest and capital at risk on a given platform, is it actually useful from an investment perspective? Let’s take a look at some relationships between TVL and forward returns.
First, let’s import a csv from DefiLlama on the TVL of Ethereum since mid 2018:
Then let’s get pricing data from yahoo finance and merge the two datasets. We also calculate log returns in price to aggregate more easily across time later. We take the TVL in ETH terms as a less circular metric (since as ETH goes up in USD, TVL will also naturally go up in USD without reflecting any additional interest or willingness to risk capital by participants per se):
Let’s do some EDA and just plot what we have to look for anything anomalous:
You’ll notice when you put TVL in ETH terms, TVL doesn’t collapse quite as much compared to in USD. We calculate the log change in TVL in eth terms:
With log changes, we can more easily sum up our data into weekly or monthly changes. Let’s resample at weekly intervals:
Now we shift the data so we can see if weekly changes in TVL are correlated to next week’s log returns in price. After all, we are trying to see if more capital entering the ETH ecosystem necessarily translates to a higher price (over a weekly timeframe):
We notice an outlier in our data. When we pick out the row that has such a large change in TVL, we see it’s in the beginning of 2019. It’s possible that TVL pre-2020 was so low that it was relatively easy to increase drastically over a week. But for our purposes, let’s just remove it to better see the rest of the data.
Mostly we see a weak negative relationship. AKA, increases in weekly TVL generally led to negative weekly returns in price (and vice versa) on average. But again, the scatterplot is more indicative of a very weak or non-existent relationship.
Next, we take deciles of weekly changes in TVL and see what average log returns are by decile:
We see a similar relationship presented in a slightly different way. Low or negative changes in TVL one week tend to lead to higher returns the next week (and vice versa). Perhaps there is some mean reversion effect going on here similar to the analysis we did on BTC flows (high flows leading to negative subsequent returns on a weekly time frame). People might withdraw their ETH from the ecosystem when they’re fearful and end up setting a local bottom - leading to higher subsequent weekly returns. Again, not a strong relationship, but interesting nonetheless.
What else could we look at? Different time frames like daily or monthly might be insightful. We can also look at different chains and see if there is more of a relationship there. We might also do some testing to see if the relationship is causal in the other direction - i.e., do positive price returns lead to investors trusting an ecosystem more, which leads them to park more of their capital in native defi apps? I could see that being the case for a chain like Solana last year.