Value Creation from Uniswap V2
Uniswap is depicted as a success story of financial innovation — but how much value does the protocol actually generate, and how successful are its liquidity providers
Written by Mathieu Amsler and Andrea 3arbon — May 1st 2021
Uniswap V2 is approaching the end of its life-cycle, as it will most likely be overshadowed by the introduction of Uniswap V3 on May 5th. It has been an amazing ride, with volumes rising to billions of dollars in less than one year, and total liquidity approaching 8 billion dollars. We believe this is a good time to look back and draw some conclusions about the achievements of such an innovative protocol. In particular, we are interested in the following question:
Did Uniswap generate added value for liquidity providers?
In other words, was the protocol actually profitable for LPs over its lifetime?
To answer this question, we queried TheGraph and gathered data for all transactions recorded on the Uniswap V2 smart contracts for 9 among the most traded currency pairs. The data covers swaps, mints, and burns for the following pairs, since the release of V2 in May 2020:
Returns From Liquidity Provision
Next, we define a meaningful measure of profitability from the point of view of LPs. Such a metric should consider the earned transaction fees while accounting for permanent and impermanent losses over the last year. Moreover, since we are interested in the value generated only by the Uniswap protocol, it should abstract from the overall increase of cryptocurrencies in dollar terms over this period. Hence, the returns outlined below are measured as the difference against a simple “buy and hold” strategy. In addition, the UNI rewards distributed as part of the liquidity incentive programs are excluded from the return computations, as the objective of the analysis is to assess whether the Uniswap protocol is value-generating at an aggregate level, and from a purely financial perspective.
Just to make it clear, we are not interested in the profits/losses made by single LPs but, rather, by the protocol as a whole. With this objective in mind, we calculate profits and losses of a representative LP, that is, we proceed as if all the liquidity positions belong to the same individual.
We compute the net return R(t) from liquidity provision on day t as
where 𝓛(t) is the total liquidity available for the pair at the end of day t, expressed in units of liquidity tokens. Recall that 𝓛(t) is by definition the square root of k, the invariant in the constant product formula x * y = k, at time t.
The first component F(t) represents the trading fee generated by the pool for every swap event of day t. To compute it, we rely on a dollar-agnostic method as to abstract from price changes in the underlying tokens in dollar terms, described in Adams et al. (2020). Recall that for any given trade on Uniswap, a flat fee of 0.30% is charged to the trader and added directly to the reserves as new liquidity. This new liquidity is distributed proportionally to all LPs participating in that pair. Hence, the accrued fees can be deduced from the change in total liquidity, using the expression
The second component is the Permanent Loss PL recorded on day t. It arises at each burn event whenever the price at the time of transaction is different from the liquidity-weighted average price of all prior mint events:
Where 𝛥p(t) is the ratio between the quoted price at the end of day t and the liquidity-weighted average price at which liquidity has been provided since inception of the pool up to day t-1, while B(t) is the amount burned (that is, withdrawn from the liquidity pools) on day t.
The third component is the total Impermanent Loss IL(t) as of day t. It is the unrealised loss relative to holding the tokens arising from the difference between the current price and the weighted average of the prices of all prior mint events, computed on the outstanding liquidity in the pool:
Where 𝛥p(t) is defined as above. Notice that the above expression measures the total impermanent loss accrued as of day t. In the above formula for daily returns R(t) we use its percentage change, denoted by 𝛥IL(t).
Cumulative Returns
The following plots display the cumulative net returns from providing liquidity, accrued over time from May 2020 to April 2021. These are obtained by compounding the daily net returns defined above.
These plots clearly show that the returns are positive and pretty high for all the pairs we consider. They correspond to Annual Percentage Yields (APYs) ranging from 10% for stable/stable pairs to 40% for those involving cryptos.
Notice that the stable/stable pairs deliver a much more stable stream of daily returns with respect to the more risky pairs. With this observation in mind, we move to a more meaningful measure of profitability.
Sharpe Ratios
A commonly used metric to assess the risk-return profile of an investment opportunity is the so-called Sharpe Ratio, which measures the average return of the investment per unit of risk involved in it. We compute the realised annualised Sharpe ratio from providing liquidity on a given pair as
where E[R] is the average of daily returns and V[R] is the standard deviation.
Results, presented in Table 2, show that the Sharpe ratio is highly positive for all the pairs we consider. These numbers are large compared to the usual Sharpe ratios achievable by investment in standard financial assets (for instance, the annualized Sharpe of the S&P 500 over the last year is 1.6 and that has been an exceptionally profitable year).
The Sharpe ratios for Ether/Stable and Ether/Crypto pairs lie within the same range and average at roughly 5, which is a solid performance. Moreover, the Sharpe ratios for pairs involving only stable coins are astonishing, averaging at 20. This is more than 12x the Sharpe ratio generated by the S&P 500 over the same period! This surprising result comes from the fact that prices of stable/stable pairs are extremely stable, indeed. Hence the permanent and impermanent losses tend to zero, while the trading volume is pretty high, generating significant fees revenues.
We thus conclude that, at least for the pairs we are considering, Uniswap v2 has been highly profitable for LPs, both in terms of net returns relative to holding and of the associated Sharpe ratios.
Are Lower Fees Possible?
As we showed in the previous section, liquidity provision has been highly profitable. As these profits arise from the protocol’s transaction fees, they correspond to costs for traders swapping tokens. As stated, Uniswap v2 sets a flat transaction fee of 30 bps (0.3%) of the swapped amount. A question then naturally arises:
Could these 0.3% fees be lowered, while still making it profitable for LPs to provide liquidity?
To shed light on this question we run a data-driven experiment: we compute the hypothetical profits that LPs would have earned if fees were set to different levels. In doing so, we keep fixed the traded amount over our sample period and the levels of liquidity in the pools.
Figure 2 displays the results from this exercise, plotting the resulting hypothetical Sharpe ratios as a function of the fee levels.
As expected, the plots clearly show that Sharpe ratios are declining linearly as transaction fees are set to lower levels. Most interestingly, they show that Sharpe ratios would be positive in a wide range of values below the current figure of 30 bps. In particular, most of the pairs would deliver a Sharpe ratio higher or equal to that of the S&P 500 even if fees were to be halved and set at the level of 15 bps.
It is worth noticing here that there are many other risks beyond impermanent loss involved with being an LP (such as smart contracts bugs), which are not captured by returns volatility. One could thus argue that these high Sharpe Ratios properly compensate LPs for those risks and depend on the aggregate risk aversion. If this view is correct, we may see lower transaction fees in the future as the protocol matures, smart contracts become more solid and the overall cyber-security of the environment improves. It will be interesting to observe the impact of the new fee structure introduced in the Uniswap V3 whitepaper (1.00%, 0.30% and 0.05% depending on the pair type) and whether the Uniswap governance will allow for new fee structures, for instance, determined by market forces.
Conclusion
Uniswap is a prime example of financial innovation. While being highly interesting from a conceptual perspective, the protocol also generates significant positive returns for its liquidity providers. As the platform will mature and reach mainstream adoption, the overall returns and Sharpe Ratios observed in our analysis are likely to be affected downwards — unless volume continues on its impressive growth rate trajectory.
Beyond financial returns, Uniswap proves that financial fairness is possible on a large scale. The protocol does not discriminate against any actors, does not require market makers to purchase any heavy hardware nor to develop their own set of liquidity provision algorithms, contrary to their counterparts in traditional trading venues based on order books. In addition, such markets are censorship-resistant and, therefore, cannot be halted by a centralised decision. Even in the eventuality of maintenance of the Uniswap website, the market still runs through its smart contracts.
Our results also indicate that fees could be substantially reduced without impeding the ability of liquidity providers to generate solid risk-adjusted returns. Lower fees would achieve lower transactions costs and better price efficiency. In the long-term, we see a flexible fee structure determined by market forces as one of the key components that will allow AMMs to definitely overtake centralised structures. Better capital efficiency, through the price range liquidity provisions implemented in the upcoming Uniswap upgrade, will also remedy one of Uniswap’s V2 pitfall.
Looking forwards to V3!
