Risks and Returns of Uniswap V3 Liquidity Providers

Lioba Heimbach Eric Schertenleib Roger Wattenhofer

ETH Zürich ETH Zürich ETH Zürich
Switzerland Switzerland Switzerland
hlioba@[Link] ericsch@[Link] wattenhofer@[Link]
ABSTRACT the smart contract. For their service, the liquidity providers receive
Trade execution on Decentralized Exchanges (DEXes) is automatic transaction fees from the trades supported by their liquidity.
and does not require individual buy and sell orders to be matched. Most DEXes on the Ethereum platform implement a constant
arXiv:2205.08904v2 [[Link]] 21 Sep 2022

Instead, liquidity aggregated in pools from individual liquidity function market maker (CFMM) for automatic trade execution [2, 3,
providers enables trading between cryptocurrencies. The largest 6, 10, 15]. Uniswap [10] is currently the biggest DEX on Ethereum [9]
DEX measured by trading volume, Uniswap V3, promises a DEX in terms of volume. There are two actively used versions of Uniswap:
design optimized for capital efficiency. However, Uniswap V3 re- V2 [13] and V3 [14]. While Uniswap V2 implements a constant
quires far more decisions from liquidity providers than previous product market maker (CPMM), where the liquidity supplied by
DEX designs. any liquidity provider supports trading on the entire price range,
In this work, we develop a theoretical model to illustrate the Uniswap V3 utilizes a new CPMM design intended to optimize
choices faced by Uniswap V3 liquidity providers and their impli- capital efficiency. Uniswap V3 has acquired a significant market
cations. Our model suggests that providing liquidity on Uniswap share throughout the past year and has overtaken its predecessor
V3 is highly complex and requires many considerations from a to become the DEX with the largest trading volume [8].
user. Our supporting data analysis of the risks and returns of real Liquidity providers on Uniswap V3 specify the price range in
Uniswap V3 liquidity providers underlines that liquidity providing which they wish to supply their liquidity. Thus, the choices liquidity
in Uniswap V3 is incredibly complicated, and performances can providers face have dramatically increased. In addition to choosing
vary wildly. While there are simple and profitable strategies for the pool to provide liquidity (as in Uniswap V2), in Uniswap V3,
liquidity providers in liquidity pools characterized by negligible liquidity providers must also specify the position and width of the
price volatilities, these strategies only yield modest returns. Instead, price range for which they wish to supply liquidity. This choice
significant returns can only be obtained by accepting increased significantly impacts their expected returns, as well as the related
financial risks and at the cost of active management. Thus, provid- financial risks1 . The increased complexity of providing liquidity
ing liquidity has become a game reserved for sophisticated players begs the question of whether liquidity providing has become a
with the introduction of Uniswap V3, where retail traders do not game of sophisticated players or whether retail traders still stand
stand a chance. a chance in comparison. To phrase it more pointedly, are retail
traders looking for high returns as liquidity providers running a
KEYWORDS risk of losing it all?
In this work, we derive an analytical expression for the imper-
blockchain, decentralized exchange, constant product market maker, manent loss of a liquidity position. Impermanent loss is a liquidity
liquidity provider provider’s risk of a decrease in the value of their liquidity position
in comparison to the value of the initial assets. We find that the
1 INTRODUCTION impermanent loss increases at a faster pace with concentrated liq-
uidity. Furthermore, we present a theoretical model illustrating the
Since the inception of Bitcoin [24] in 2008 and later Ethereum [27] complexities liquidity providers face.
in 2014, the capabilities of blockchains have significantly evolved. In addition, we analyze the performance of liquidity positions
Most notably, the introduction of smart contracts by Ethereum in the largest Uniswap V3 pools and show that the returns and
enabled the blockchain to host an entire financial system com- risks of liquidity positions vary wildly. Both our model and the
monly known as decentralized finance (DeFi). DeFi has shifted the data highlight that due to the complexity of Uniswap V3, providing
widespread perception of cryptocurrencies being a tool for price liquidity in price volatile pools requires both active management
speculation to a technology with the potential to revolutionize the and high sophistication. Retail traders, unwilling to risk significant
future of finance. losses, should stick to simple strategies offering only low returns
Decentralized exchanges (DEXes) represent a DeFi cornerstone but also promising only small financial risks.
technology. DEXes allow users to exchange cryptocurrencies with- We further find that the interests of high profit seeking liquidity
out giving up custody of their assets. Users interact directly with providers and the protocol may be misaligned. While the protocol
the smart contracts that build the DEXes. Trading is enabled by seeks long term liquidity providers in order to offer traders sufficient
liquidity for each pair of tradeable cryptocurrencies reserved in a market depth at any time, liquidity providers maximize their profit
respective smart contract referred to as a liquidity pool. Thus, when by actively managing their positions. The later could lead to a
a user wishes to exchange 𝑋 -tokens for 𝑌 -tokens, the user interacts significant draining of liquidity in turbulent market situations.
with the respective liquidity pool, deposits 𝑋 -tokens in the pool,
and receives 𝑌 -tokens from the pool’s liquidity. Individual liquidity 1 We focus on the financial risks stemming from price fluctuations and analyze them
providers supply the pool’s liquidity by depositing both assets in using historical data. Other risk, e.g., risks related to the protocol, are not considered.
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

2 RELATED WORK
Several studies of the risks and returns of liquidity providers in the 𝑥 real
original CPMM exist. Evans [17] studies the returns of liquidity 𝑢
providers and shows that they are capable of replicating various
trading strategies and financial derivatives. In a further study, Evans
et al. [18] find the optimal transaction fee for liquidity pools to

𝑌 reserves
attract liquidity. On the other hand, we empirically analyze the
risks and returns of individual liquidity providers and illustrate the
complexity facing liquidity providers in the novel CPMM design 𝑥
utilized by Uniswap V3. 𝑚
Heimbach et al. [20] present an empirical study of the behavior 𝑦real
of Uniswap V2 liquidity providers, as well as general risk and return 𝑦
metrics of Uniswap V2 pools. Their work shows the performance of 𝑙
liquidity providers is largely driven by the pool’s price volatility. In
contrast, our work is focused on the risk and returns of Uniswap V3
liquidity providers. We find that providing liquidity in Uniswap V3
𝑋 reserves
is significantly more complex, exemplified by the stark difference
in liquidity position returns dependent on their strategies, even
within the same pool. Figure 1: CPMM price curve with virtual liquidity. The blue
Neuder et al. [25] study strategic liquidity provision in Uniswap price curve ensures the constant product 𝑥 · 𝑦. The current
V3 and evaluate three classes of strategies for liquidity providers. marginal market price 𝑆 is at point 𝑚 and is given by the
However, their analysis makes several strong assumptions and, for ratio of the virtual reserves, 𝑆 = 𝑦/𝑥.
instance, does not take into account the losses liquidity providers
can face due to changing asset prices – the sole driver of liquidity
provider losses. In a similar line of work, Fritsch [19] quantifies the for each tradeable cryptocurrency pair. We note that Uniswap V2
performance of a set of liquidity provider strategies by simulating and V3 allow for the creation of liquidity pools between any ERC-20
them with historical trading data. We model the considerations tokens – an Ethereum standard for fungible tokens. Many individual
liquidity providers must face when choosing the liquidity position liquidity providers supply the pool’s liquidity, and their aggregated
in our work. Further, we empirically analyze the risks and returns liquidity enables trading in the pool. The CPMM enforces that
of real Uniswap V3 liquidity providers. during trading, the product between the reserves of the pool’s two
In a series of blog posts, Lambert [21, 22] investigates several cryptocurrencies stays constant, i.e., the pool’s state moves along
challenges of providing liquidity on Uniswap V3 and makes the link the price curve drawn in Figure 1.
between a Uniswap V3 liquidity position to financial derivatives. Uniswap V3’s two predecessors, as well as SushiSwap [6], employ
The presented analysis makes several simplifications to obtain an the original CPMM design. To illustrate the pricing mechanism of
analytical solution. However, some of these assumptions are not the original CPMM, we consider a 𝑋 − 𝑌 pool between 𝑋 -tokens
supported by the actual data. We, on the other hand, generalize and 𝑌 -tokens. If there are 𝑥 𝑣2 𝑋 -tokens and 𝑦 𝑣2 𝑌 -tokens reserved
Lambert’s ideas in our model and highlight the complexities faced in the pool, then the pool’s marginal price is given by 𝑆 = 𝑦 𝑣2 /𝑥 𝑣2
√
by liquidity providers. and the pool’s liquidity is defined as 𝐿𝑣2 = 𝑥 𝑣2 · 𝑦 𝑣2 [13]. In this
A recent report presented by Loesch et al. [23] makes a first traditional implementation of CPMM, liquidity placed in the pool
attempt at analyzing the returns of real Uniswap V3 liquidity po- supports trading on the entire price range (0, ∞) (cf. Figure 2a).
sitions and concludes that around 50% of liquidity positions are With Uniswap V3, Adams et al. [14] introduce a novel CPMM
losing money. However, their report has several shortcomings. For design. Uniswap V3 liquidity providers specify the price range
instance, they only consider the position’s lifetime as a contribut- [𝑆𝑙 , 𝑆𝑢 ] in which they wish to supply liquidity (cf. Figure 2b). Their
ing factor to the liquidity position’s return in their analysis. In our liquidity then only supports trading within this price range. Thus,
work, we account for multiple factors that influence the returns Uniswap V3 has an increased liquidity concentration around the
of liquidity providers on Uniswap V3 both theoretically and em- current price and thereby increases the market’s capital efficiency.
pirically, finding that with appropriate considerations providing We note that liquidity providers can only choose the price bound-
liquidity on Uniswap V3 can be profitable. aries of their liquidity position from a predefined set of the pool’s
initialized ticks. There is a tick at every integer exponent of 1.0001,
and the price of tick 𝑖 (𝑖 ∈ Z) is given by
3 CONSTANT PRODUCT MARKET MAKER
𝑆 (𝑖) = 1.0001𝑖 .
Uniswap V3 [14] functions as an automated market maker (AMM),
i.e., trading is automatic, and a predefined algorithm controls the Consequently, each tick is 0.01% (one basis points (bps)) away from
cryptocurrency prices. More specifically, Uniswap V3, like its prede- each neighboring tick. Not every tick can be initialized in a pool,
cessors Uniswap V1 and V2 [13], utilizes the most widely adopted instead, only ticks with indexes divisible by the pool’s predefined
AMM subclass: constant product market maker (CPMM). tick spacing (𝑡𝑠 ) can be initialized. Figure 2c shows a schematic
Uniswap aggregates liquidity in what is known as a liquidity pool representation of a liquidity allocation across a liquidity pool’s
Risks and Returns of Uniswap V3 Liquidity Providers

We demonstrate Uniswap’s trading mechanism by, again, con-

liquidity

liquidity
sidering a liquidity pool 𝑋 − 𝑌 . The (virtual) reserves at the current
price are given by 𝑥 𝑋 -tokens and 𝑦 𝑌 -tokens. In Uniswap V2, a
trader wishing to exchange 𝛿𝑥 𝑋 -tokens for 𝑌 -tokens, will receive
0 ∞ 𝑆𝑙 𝑆𝑢 𝛿 𝑦 𝑌 -tokens, where
price price

(a) Uniswap V2 liquidity (b) Uniswap V3 liquidity position

𝑥 ·𝑦 𝑦 (1 − 𝑓 )𝛿𝑥
𝛿𝑦 = 𝑦 − = .
𝑥 + (1 − 𝑓 )𝛿𝑥 𝑥 + (1 − 𝑓 )𝛿𝑥
liquidity

Here, 𝑓 is the transaction fee charged relative to the trader’s input

amount 𝛿𝑥 [13]. As long as the pool’s price does not move across an
0 ∞ initialized tick, the same holds in Uniswap V3. Otherwise, the trade
price
executes at the available liquidity depth until it reaches the next
(c) collection of Uniswap V3 liquidity positions initialized price tick. Then the remaining trade is completed at the
available liquidity depth in the subsequent interval. This procedure
Figure 2: Schematic representation of liquidity allocation re-applies until the entire trade input is swapped.
on CPMMs. The points 𝑆𝑙 and 𝑆𝑢 specify the lower and up- The collected transaction fees are distributed pro-rata to the
per bounds of the liquidity position’s price range, respec- pool’s liquidity providers, who have deposited liquidity at the
tively. Unlike on Uniswap V2, liquidity is not distributed price the asset pairs are trading at. We note that while the fees
uniformly across the entire price range but is determined in Uniswap V2 were compounded, they are not in Uniswap V3.
by the choices of the liquidity providers. Further, observe that on Uniswap V3 the same asset pairs can be
tradable in different pools that differ by the fee that is charged to
traders. These are referred to as pool fee tiers. The possible tiers are
price range. The liquidity is no longer constant across the entire 𝑓 ∈ {0.01%, 0.05%, 0.3%, 1%}. Thus, the fees received by a liquidity
price range, instead, only between the pool’s initialized ticks. provider in the current price range are not only dependent on the
Uniswap V3 utilizes the concept of virtual reserves, adjusted pool’s volume and liquidity depth but also on its fee tier. Uniswap
larger reserves, to describe the pool’s behavior between two ad- suggests the usage of lower transaction fees in pools with low rela-
jacent ticks 𝑇𝑙 and 𝑇𝑢 . The virtual reserves act as if the liquidity tive price volatility between the two cryptocurrencies, such as two
in the entire pool matches that of the current price range. Thus, stablecoins. As suggested by the name, stablecoins are designed to
the virtual reserves of the current price range are a transformation have a stable price and are often pegged to the US$. Higher transac-
of the range’s real reserves that allow for the application of the tion fees are suggested in pools with a high relative price volatility
constant product formula. In the following we consider a pool with between the two cryptocurrencies.
reserves 𝑥 real and 𝑦real between two adjacent ticks. The price range
between ticks 𝑇𝑙 and 𝑇𝑢 is given by [𝑆𝑙 , 𝑆𝑢 ]. Instead of ensuring
that the product between the reserves stays constant, the proto- 3.1 Liquidity Provision
col ensures that the product of the virtual reserves 𝑥 and 𝑦 stays In the original CPMM design, liquidity providers only choose a pool
constant, i.e., and their liquidity supported trading on the entire price interval.
𝑥 · 𝑦 = 𝐿 2, In comparison to holding their assets, liquidity providers’ returns
where 𝐿 the liquidity reserved in between ticks 𝑇𝑙 and 𝑇𝑢 [14]. were positively influenced by transaction fees and negatively by
Further, the marginal price is given by 𝑆 = 𝑦/𝑥. The following impermanent loss. The latter describes the risk of a liquidity provider
relationship then holds between the virtual reserves, liquidity, and seeing the value of their reserved assets decrease in comparison to
marginal price: the value of the initial assets. In particular, price changes between
𝐿 √ the two reserved cryptocurrencies drive up a liquidity provider’s
𝑥= √ 𝑦 = 𝐿 𝑆. impermanent loss. Thus, the returns of liquidity providers were
𝑆
influenced by a pool’s volume and price volatility.
Thus, the virtual reserves behave according to the constant product
In Uniswap V3, liquidity providers must also specify the range
price curve as shown in Figure 1. Similarly, the real reserves can be
in which they wish to supply liquidity. It is, therefore, no longer
obtained as follows:
√ only a pool’s transaction fees and volatility but also the liquid-
𝐿 𝐿 𝐿 √︁ √︁
𝑥 real = 𝑥 − √ = √ − √ 𝑦real = 𝑦 − 𝐿 𝑆𝑙 = 𝐿 𝑆 − 𝐿 𝑆𝑙 . ity provider’s price range that influences the returns. A liquidity
𝑆𝑢 𝑆 𝑆𝑢 provider’s liquidity is only active when the pool’s price 𝑆 is within
Observe that the pool only needs to maintain sufficient reserves to her specified price range [𝑆𝑙 , 𝑆𝑢 ]. Whenever the liquidity is ac-
support trading within the price boundaries [𝑆𝑙 , 𝑆𝑢 ]. Thus, the real tive, it is facilitating transactions and, in turn, earns fees from the
reserves of 𝑋 -token shrink as the value of 𝑋 in terms of 𝑌 increases pool’s transactions. When, however, the price is outside the liquid-
and are fully depleted at the upper price boundary 𝑆𝑢 . The opposite ity provider’s price range, her liquidity is inactive. More specifically,
holds for the real reserves of 𝑌 -token. a liquidity provider that places (virtual) liquidity 𝐿˜ into the pool
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

𝑋 − 𝑌 in the price range [𝑆𝑙 , 𝑆𝑢 ] will have

 


 ˜ · √1 − √1
𝐿 𝑆 < 𝑆𝑙
  𝑆𝑙

 𝑆𝑢
𝑥˜ real = 𝐿˜ · √1 − √1 𝑆𝑙 ≤ 𝑆 < 𝑆𝑢 (1)

 𝑆 𝑆𝑢

0
 𝑆 ≥ 𝑆𝑢
𝑋 -tokens reserved in the pool when the pool’s price is 𝑆, as well as

 0 𝑆 < 𝑆𝑙
 ˜ √

 √︁ 

𝑦˜real = 𝐿 · 𝑆 − 𝑆𝑙 𝑆𝑙 ≤ 𝑆 < 𝑆𝑢 (2)
 √ √︁ 
𝐿˜ · 𝑆𝑢 − 𝑆𝑙

 𝑆 ≥ 𝑆𝑢

𝑌 -tokens [14] 2 . Thus, when the price is within her price range, she
holds both tokens in the pool, and her liquidity is active, i.e., she is
earning fees. When, however, the price is outside the boundaries,
she only holds one of the tokens. More specifically, when 𝑆 < 𝑆𝑙 (the
price of 𝑋 with respect to 𝑌 decreases and is outside the boundary), Figure 3: Simulation of the impermanent loss of a liquidity
her liquidity only consists of 𝑋 -tokens. On the other side of her position on Uniswap V2 and Uniswap V3 for different liq-
price boundary, the opposite holds true. Thus, both the location uidity position widths. We set 𝑆𝑙 = 1/𝛼 · 𝑆 0 and 𝑆𝑢 = 𝛼 · 𝑆 0 .
and width of the price range starkly influence a liquidity provider’s
return. While large price ranges decrease a liquidity provider’s
capital efficiency, the liquidity is likely active for longer periods.
Small price ranges, on the other hand, increase a liquidity provider’s is given by
capital efficiency, i.e., she earns more fees relative to her liquidity
size when the price is in her range. At the same time, the price is ˜
˜ 𝑆 0, 𝑆 1 ) = 𝑆 1 · 𝑥 0 + 𝑦0 = √𝐿 𝑆 1 + 𝐿˜ 𝑆 0 .
√︁
generally more likely to fall out of her range more quickly. Thus, 𝑉v2,hold (𝐿,
𝑆0
liquidity providers often readjust their liquidity positions following
price changes. In the following, we provide a thorough analysis and Thus, we can obtain the impermanent loss as follows:
evaluation of the risks and returns faced by Uniswap V3 liquidity
providers. 𝑉v2,pos − 𝑉v2,hold
ILv2 (𝑆 0, 𝑆 1 ) =
𝑉v2,hold
4 ANALYSIS √ 
˜ √ 
In this section, we discuss the factors influencing the performance of 2𝐿˜ 𝑆 1 − √𝐿 𝑆 1 + 𝐿˜ 𝑆 0
𝑆0
= √
liquidity positions on Uniswap V3. Liquidity providers face several ˜
√𝐿 𝑆 1 + 𝐿˜ 𝑆 0
decisions when choosing their liquidity position. We provide a √︃
𝑆0
theoretical discussion of the implications of each of these choices. 𝑆1
© 2 · 𝑆0 ª
= ­­ − 1®® .
4.1 Impermanent Loss 𝑆1
1 + 𝑆0
« ¬
We start by analyzing the main risk faced by liquidity providers:
impermanent loss. The impermanent loss describes the loss in value We note that a liquidity providers impermanent loss is zero when
of a liquidity position in comparison to holding the original assets 𝑆 0 = 𝑆 1 , i.e., the price is the same as at the initial time of liquidity
as the price changes. We start by deriving the impermanent loss of injection. Otherwise, the impermanent loss is always negative3 .
a Uniswap V2 liquidity provider, which corresponds to a Uniswap We plot the impermanent loss as a function of the relative price
V3 liquidity provider setting her price range to (0, ∞). Consider a change for a liquidity provider on the entire range, the equivalent
liquidity provider that places 𝐿˜ liquidity into a pool 𝑋 − 𝑌 when of a Uniswap V2 liquidity provider, in yellow in Figure 3.
the pool’s
√ marginal price is 𝑆 0 . Thus,
√ the liquidity provider places In the following, we repeat the same steps to obtain the imper-
˜ 𝑆 0 𝑋 -tokens and 𝑦0 = 𝐿˜ 𝑆 0 𝑌 -tokens in the pool. The
𝑥 0 = 𝐿/ manent loss of a liquidity provider that supplies liquidity 𝐿 into
value of the liquidity provider’s position at a later point, when the the pool 𝑋 − 𝑌 in the price range [𝑆𝑙 , 𝑆𝑢 ]. The liquidity provider
pool’s price is 𝑆 1 , is given by: inserts the liquidity into the pool, when the pool’s marginal price
˜ is 𝑆 0 and we, again, derive the impermanent loss at a later point
˜ 𝑆 1 ) = 𝑆 1 · 𝑥 1 + 𝑦1 = √𝐿 𝑆 1 + 𝐿˜ 𝑆 1 = 2𝐿˜ 𝑆 1,
√︁ √︁
𝑉v2,pos (𝐿, in time, when the pool’s marginal price is 𝑆 1 . We start by obtain-
𝑆1
ing the position value at price 𝑆 1 with the help of Equation 1 and
where 𝑥 1 and 𝑦1 are the liquidity provider’s assets in the pool. The
value of the liquidity provider’s original assets, on the other hand,
2 The virtual liquidity 𝐿˜ can be expressed in terms of the initial deposited value 𝑉 3 Byconvention the impermanent loss is smaller or equal to zero, thus, a non-zero
0
using 𝑉0 = 𝑆 0 · 𝑥 0 + 𝑦0 . impermanent loss is detrimental to the liquidity provider.
Risks and Returns of Uniswap V3 Liquidity Providers

Equation 2:
𝑉pos (𝐿,˜ 𝑆 1, 𝑆𝑙 , 𝑆𝑢 ) = 𝑆 1 · 𝑥 1 + 𝑦1
 
𝐿˜ · 𝑆 1 √1 − √1

 𝑆 1 < 𝑆𝑙
 𝑆 𝑆
  √

 𝑙
√︁ 𝑢 𝑆 1 
= 𝐿˜ 2 𝑆 1 − 𝑆𝑙 − √ 𝑆𝑙 ≤ 𝑆 1 < 𝑆𝑢

 √ √︁  𝑆𝑢
𝐿˜ · 𝑆𝑢 − 𝑆𝑙 𝑆 1 ≥ 𝑆𝑢 .



Here, 𝑥 1 𝑋 -tokens and 𝑦1 𝑌 -tokens are the liquidity position’s real
reserves in the pool. We obtain the value of the original inserted
liquidity similarly:
˜ 𝑆 0, 𝑆 1, 𝑆𝑙 , 𝑆𝑢 ) = 𝑆 1 · 𝑥 0 + 𝑦0
𝑉hold (𝐿,
 
𝐿˜ · 𝑆 1 √1 − √1


 𝑆 0 < 𝑆𝑙
  𝑆 +𝑆 𝑆𝑙 √︁ 𝑆𝑢 𝑆 



˜ 0
= 𝐿 √ − 𝑆𝑙 − √ 1 1
𝑆𝑙 ≤ 𝑆 0 < 𝑆𝑢

 √𝑆 0 √︁  𝑆𝑢 Figure 4: Simulation of the impermanent loss of a liquidity

 ˜
𝐿 · 𝑆𝑢 − 𝑆𝑙 𝑆 0 ≥ 𝑆𝑢 ,
 position minted outside the current price on Uniswap V3.
where 𝑥 0 𝑋 -tokens and 𝑦0 𝑌 -tokens are the reserves initially placed The respective price ranges are set [𝑆 0, 2 · 𝑆 0 ] and [0.5 · 𝑆 0, 𝑆 0 ].
in the pool. The impermanent loss can then be obtained from the
two preceding expression as follows:
position (including the earned fees). This allows the performance
𝑉pos − 𝑉hold comparison of owning the two tokens and providing liquidity vs.
𝐼𝐿(𝑆 0, 𝑆 1, 𝑆𝑙 , 𝑆𝑢 ) = . just owning the two tokens. This performance measurement is
𝑉hold
We simulate the impermanent loss of a liquidity provider with more suited than comparing it to a fixed currency, for example,
price range [1/𝛼 · 𝑆 0, 𝛼 · 𝑆 0 ] for 𝛼 ∈ [1.1, 4, 20] in Figure 3. Notice US$, as it is not dominated by the cryptocurrencies price evolution
that the smaller 𝛼, i.e., the tighter the price range, the faster the im- compared to this fixed currency, but rather allows to pinpoint the
permanent loss increases. Thus, not only do liquidity providers run return that stems from the actual decision of providing liquidity.
a higher risk of their liquidity becoming idle, when the pool’s price The fees collected are distributed pro-rata to the pool’s liquidity
moves outside their interval, but their impermanent loss increases providers, who have deposited liquidity at the price the asset pairs
more quickly as well. The liquidity of liquidity providers with a are trading at. We note that on Uniswap the same asset pairs can be
wider price interval, on the other hand, is less capital efficient. They tradable in different pools that differ by the fee that is charged to
earn less fees for their liquidity, when the price is in the interval. traders. These are referred to as pool fee tiers. The possible tiers are
Figure 3 shows the impermanent loss of a liquidity provider who 𝑓 ∈ {0.01%, 0.05%, 0.3%, 1%}. Thus, the fees received by a liquidity
enters the pool when the price is in the middle of her interval. In provider in the current price range are not only dependent on the
Figure 4, on the other hand, we show the impermanent loss of a pool’s volume and liquidity depth but also on its fee tier.
liquidity provider that enters the pool, when the price is on the Additionally, to garner fees, the selection of the liquidity posi-
edge of her interval: 𝑆 0 = 𝑆𝑙 (drawn in green) and 𝑆 0 = 𝑆𝑢 (drawn tion’s price range is crucial. For pairs of stablecoins, the acquired
in violet). Notice that when the price remains outside the interval, fees are directly determined by the price range and trading volume,
i.e., 𝑆 1 ≤ 𝑆𝑙 for 𝑆 0 = 𝑆𝑙 , the impermanent loss is zero. However, as the impermanent loss is insignificant. Thus, liquidity providers
unless the price moved in and out of the interval in the meantime, deposit in a narrow price range near the stable ratio. However,
the liquidity provider also did not earn any fees. As soon as the in volatile pools, like Ether in US$, selecting a suitable liquidity
price moves into the interval and beyond, the impermanent loss position is more challenging as the price is more likely to exit the
builds up quickly. chosen price range leading to a higher impermanent loss and no
When choosing where to provide liquidity, Uniswap V3 liquidity further fees being collected.
providers must account for risks stemming from the impermanent The feature that fees are only collected as long as the price re-
loss, which are only negligible when little to no price movements mains within a certain bandwidth, necessitates the liquidity provider
are expected for the pool. to gauge the probability thereof. This requirement naturally sug-
gests adopting methods used for pricing financial derivatives where
4.2 Selection of Liquidity Position short- and medium-term price predictions are essential. The most
In addition to the negative influence of the experienced imperma- well-known model for modeling the future development of the
nent loss, a liquidity provider’s return 𝑅 is positively influenced by price of a risky asset 𝑆 (𝑡) is the Black-Scholes market model, where
the fees earned 𝐹 and given by the price is represented as an Itô process satisfying the stochastic
differential equation
𝑉pos + 𝐹 − 𝑉hold
𝑅(𝑆 0, 𝑆 1, 𝑆𝑙 , 𝑆𝑢 , 𝐹 ) = . (3) 𝑑𝑆 (𝑡) = 𝜇𝑆 (𝑇 )𝑑𝑡 + 𝜎𝑆 (𝑡)𝑑𝑊 (𝑡),
𝑉hold
We note that the return compares the value of the initial assets where 𝜇 is called the drift of the asset price and determines the
deposited into the liquidity position to the value of the liquidity expectation value of the future price, i.e., E[𝑆 (𝑡)] = 𝑆 0𝑒𝑥𝑝 (𝜇𝑡) [16].
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

Figure 5: Simulation of daily asset price paths over 30 days Figure 6: Probability that the asset price remains within the
relative to initial price 𝑆 0 . The shaded area shows 2-𝜎 level of liquidity position as a function of time for different 𝛼. For a
the asset price distribution. Note that the width of this area volatile asset the probability that the asset moves out of the
grows with the square root of time (geometric Brownian mo- selected range increases over time.
tion). In this plot only 600 of 40’000 paths used in total are
shown.

The volatility of the asset price is denoted as 𝜎, and𝑊 (𝑡) is a Wiener

process. Note that the term risky asset is adopted from finance,
but in our case, the choice of which coin in a pool is considered
the risky asset is arbitrary. Rather, it just represents a choice of
measuring the price of that coin 𝑆 (𝑡) with respect to the other
coin. Throughout this section, we pick an annual volatility of 70%,
which is reasonable for an asset like Ether measured in US$ (see for
example [5]). Furthermore, we choose the drift to be zero, as the
short-term price movements are mainly governed by the volatility.
Finally, we use the terminology from finance and say that a liquidity
position is in the money (ITM) if the asset price is within the price
range the liquidity provider chose. In the opposite case, we say the
position is out of the money (OTM).
Figure 7: Expected proportion of time the asset price lies
Given the initial asset price 𝑆 (0) the above differential equation
within liquidity position for different 𝛼. A wider position is
has the formal solution
expected to remain in the money, and hence collecting trans-
𝜎2
 
action fees, for a longer fraction of time.
𝑆 (𝑡) = 𝑆 (0) exp 𝜇𝑡 − 𝑡 + 𝜎𝑊 (𝑡) . (4)
2
As the liquidity provider only collects fees if the price remains
price spends ITM as, during this time, she profits from each trade
within its selected price range, we are interested in a suitable selec-
that occurs in the pool. Given the probability that the price is ITM
tion of the liquidity range. To this end, we numerically calculate
at time 𝑡𝑖 , 𝑃 ITM (𝑆𝑙 < 𝑆 (𝑡𝑖 ) < 𝑆𝑢 ), we can compute the expected
40’000 possible future daily asset price paths using Equation 4 (cf.
time the position is ITM relative to the total time passed 𝑡𝑛
Figure 5) and compute the probability that after time 𝑇 the asset has
remained in the liquidity range determined by 𝛼, where as in Fig- 𝑖=𝑛
1 ∑︁
ure 3, we assume that the initial price 𝑆 0 is related to the lower and E[𝑇ITM (𝑡𝑛 ; 𝛼)] = 𝑃 ITM (𝑡𝑖 ; 𝛼)Δ𝑡,
𝑡𝑛 𝑖=0
upper bound of the liquidity position by 𝑆𝑙 = 𝑆 0 /𝛼 and 𝑆𝑢 = 𝛼𝑆 0 ,
respectively. where Δ𝑡 is the discrete time step (in our case one day). Figure 7
This probability 𝑃ITM (𝑡; 𝛼) = 𝑃 (𝑆 0 /𝛼 < 𝑆 (𝑡) < 𝛼𝑆 0 ) is shown shows the fraction of time the price is in ITM as a function of time
in Figure 6 as a function of time for different 𝛼. As expected, it passed. Again, the relative expected time ITM decays over time and
becomes more likely that the position is OTM with passing time is smaller for narrower position
and a larger 𝛼 means that the position is likely to remain ITM for Figures 6 and 7 illustrate the relevance of both the time the liq-
longer. uidity provider keeps the position active as well as the width of the
While the probability that a position remains ITM is important to position’s price range. While a large width reduces the probability
gauge for a liquidity provider, she is more interested in the time the that the position drops out of its price range, the fees collected per
Risks and Returns of Uniswap V3 Liquidity Providers

Uniswap V3 liquidity positions to understand the impact of these

considerations.

5.1 Data Collection

We analyze Uniswap V3 data to measure and understand the na-
ture of the risks and rewards awaiting liquidity providers. The first
Uniswap V3 pool was launched in block 12’369’739. Thus, we an-
alyze the data beginning with block 12’369’739 (May 4, 2021) up
until block 14’497’033 (last block on March 31, 2022). We collect
data from the Ethereum blockchain by launching an erigon client.
More specifically, we filter the event logs for all events related to
Uniswap V3.
Our data analysis focuses on the four biggest cryptocurrencies
on Uniswap V3 in terms of total value locked: WETH, WBTC,
USDC, and DAI [10]4 . Not only are the pools between these four
Figure 8: Simulation of the proportion of time the asset price cryptocurrencies amongst the largest in terms of volume and total
remains within the liquidity position relative to the liquid- value locked [10], but they also allow us to investigate varying
ity width 𝛼 after time 𝑇 . As the collected fees grows with patterns between different types of pools.
the time ITM and decays with position width the ratio of With the introduction of Uniswap V3, currency pairs are clas-
the time ITM relative to 𝛼 is indicative for the collected fee. sified as stable, normal or exotic depending on the relative price
Thus, the longer a liquidity provider intends to keep her liq- volatility of the assets [1]. Stable pairs are characterized by little to
uidity position in a volatile asset active, the larger the width no price changes between the pair’s two cryptocurrencies, while
should be chosen. we can expect significant price volatility between the two assets of
a normal pair. Finally, for exotic pairs, at least one asset is not an
established cryptocurrency, and the relative price between the two
trade are adversely affected by a widely spread liquidity position.
assets can fluctuate wildly. In their study of Uniswap V2 liquidity
To illustrate this trade-off, we can consider the following simplified
pools, Heimbach et al. [20] highlight the differences in the returns
model. Assume that near the current price, the liquidity distribu-
and risks of liquidity providers depending on the pair’s category,
tion of the pool is uniformly distributed, i.e., the same amount of
i.e., the pair’s price volatility. These findings demonstrate the ne-
liquidity in each tick. Then the fees collected by a position ITM are
cessity of comparing the performance of liquidity providers based
inversely proportional to 𝛼. On the other hand, the fees collected
on the price volatility of the respective asset pair rather than over
are proportional to the trading volume multiplied by the time in the
the entire protocol. Our selection of pools covers the largest pools
money 𝑇ITM . Thus, in this model, the total fee a liquidity provider
both with respect to trading volume and total value locked. They
collects is proportional to the time in the money divided by 𝛼
make up approximately a third of the total liquidity on Uniswap.
𝐹 ∝ 𝑇ITM /𝛼 . The performance of the liquidity position in these pools is thus in-
dicative of the typical risks and returns faced by liquidity providers
This quantity is depicted in Figure 8 as a function of 𝛼 at different
on Uniswap V3. We further note that our analysis is on the level of
times 𝑇 . The plot shows that there is an optimal 𝛼, which depends
individual liquidity positions, as opposed to wallets or entities.
on time, illustrating that the liquidity provider should carefully
select their liquidity position width.
5.2 General Liquidity Pool Statistics
We conclude this section by noting that the liquidity providers
in pools not consisting of two stablecoins face a complex problem. We commence the data analysis by extracting general statistics of
The liquidity provider should first have a prediction of future price liquidity positions in the three Uniswap V3 pools with the highest
and volatility developments of the selected pool. Therewith, she total value locked at the time of writing [10]: USDC-WETH (f=
chooses the price range of her liquidity position, bearing in mind the 0.3%), WBTC-WETH (f= 0.3%) and DAI-USDC (f= 0.01%). USDC-
time she intends to keep her funds locked, as well as the expected WETH and WBTC-WETH are normal pools, as at least one of the
impermanent loss suffered. Additionally, she should also consider pools’ assets is subject to significant price movements. As opposed
the liquidity distribution of the whole pool. Once the position is to exotic pools, all pool assets are established cryptocurrencies.
open, the liquidity provider must constantly monitor the position, DAI-USDC, on the other hand, is a stablecoin pool, as both the
update her predictions for the new market conditions and decide pool’s tokens are pegged to the US$ and, thus, only small price
when it is best to withdraw the funds. This complexity illustrates fluctuations are expected. The transaction fees levied by normal
that successfully providing liquidity in a volatile pool requires a pools tend to be higher than the transaction fees charged by stable
high level of sophistication. pools due to higher risks involved for liquidity providers stemming
from the impermanent loss. We further note that for each of the
5 REAL-WORLD MEASUREMENTS three token pairs, multiple pools with different fee tiers exist that
With the complexities faced by liquidity providers when selecting 4 Note that wrapped assets, e.g., WETH, by design, have the same value as the underly-
a liquidity position on Uniswap V3 in mind, we analyze the historic ing, e.g., Ether. For our purpose, we can therefore consider them equivalent.
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

(a) USDC-WETH (f= 0.3%) and WBTC-WETH (f= 0.3%) (a) USDC-WETH (f= 0.3%) and WBTC-WETH (f= 0.3%)

(b) DAI-USDC (f= 0.01%) (b) DAI-USDC (f= 0.01%)

Figure 9: Median (lighter lines) and mean (darker lines) po- Figure 10: Number of active positions over time in three
sition size over time in three Uniswap V3 pools. Observe the Uniswap V3 pools. Note the low number of active liquidity
large difference (factor ten) between the median and mean. positions compared to the pool size for DAI-USDC.

hold significant liquidity on Uniswap V3. We only plot the general between the median and mean liquidity position size is also around
liquidity position statistics for the pools with the highest liquidity a factor of ten but then increases to a factor of 100, indicating a
for better visibility. In the later performance analysis (cf. Section 5.3), pronounced discrepancy in the distribution of liquidity. We note
we will include additional fee tiers for each token pair. that both the median and mean position sizes in the DAI-USDC
Figure 9 shows both the median and mean position sizes over pool are significantly larger than in the other two analyzed pools.
the pools’ lifetime, while Figure 10 shows the number of active Especially, the DAI-USDC pool thus appears to be in the hands of
positions in each pool over time. Note that as the DAI-USDC (f= large liquidity providers: underlined by the extremely small number
0.01%) was only created in late 2021 following the introduction of of active liquidity positions in the pool, as well as the presence of
the new fee tier, the data set is significantly smaller (cf. Figure 12b). single liquidity positions worth more than US$ 100’000’000. There
When considering the position size statistics in the USDC-WETH are only around 60 active liquidity positions in the pool (cf. Fig-
pool and WBTC-WETH pool (cf. Figure 9a), we notice that, apart ure 10b), while the pool holds around US$ 300’000’000 after its
from an initial growth phase, both the median (lighter lines) and initial liquidity growth.
the mean (darker lines) show little fluctuations over time. The me- In Figure 11 we plot the mean position lifetime (darker lines), as
dian and mean position sizes are quite similar between the two well as the mean time spent by a position ITM (lighter lines) for
pools. We also find that the number of active liquidity positions the three analyzed pools. The mean position age increases linearly
in the USDC-WETH pool is approximately double the number of with the pool’s lifetime at around half the rate for the three pools,
active positions in the WBTC-WETH pool at all times. The number indicating that a significant proportion of liquidity positions are
of liquidity positions in both pools’ stabilizes at a couple of thou- active over a long period. For the two normal pairs (cf. Figure 11a),
sand after an initial growth period of around two months after the we observe a significant difference between a position age and the
pool’s creation (cf. Figure 10a). Finally, we observe that the mean time the position was ITM, i.e., the position was active and earn-
is significantly (around ten times) larger than the median in both ing fees. This difference is most pronounced for the USDC-WETH
pools, indicating a highly unequal distribution of liquidity provider pool. In the pool, the mean of the time a position was active is only
funds. When turning to the DAI-USDC pool (cf. Figure 12b), we only around half of the total position lifetime. Thus, on average, liquid-
observe this trend magnified. Until February 2022, the difference ity positions only earn fees during half the time. This difference
Risks and Returns of Uniswap V3 Liquidity Providers

(a) USDC-WETH (f= 0.3%) and WBTC-WETH (f= 0.3%) (a) USDC-WETH (f= 0.3%) and WBTC-WETH (f= 0.3%)

(b) DAI-USDC (f= 0.01%) (b) DAI-USDC (f= 0.01%)

Figure 11: Mean time position was ITM (lighter lines) and Figure 12: Median position width over time in three Uniswap
total position lifetime (darker lines) over time in three V3 pools.
Uniswap V3 pools.

factor of around 1000. We further observe that while initially, the

is also present for the WBTC-WETH pool but less pronounced. In mean position size in the USDC-WETH pool was smaller than in
the WBTC-WETH pool, liquidity positions are, on average, active the WBTC-WETH pool, the trend reverses over the pools’ lifetimes.
for more than two-thirds of their lifetime. We presume that this Thus, the market learns that it must select larger price ranges for
difference stems from the relative price between the two cryptocur- liquidity positions in pools where the volatility of the relative price
rencies, Bitcoin and Ether, having a higher correlation with each is larger. We further notice that, especially in the USDC-WETH
other than with the US$ (cf. Figure 13). The less volatile relative pool, we observe an increase in the median position width, indi-
price makes it easier to determine the price range of a liquidity po- cating that liquidity providers are becoming more familiar with
sition and makes it less likely for the price to fall out of the liquidity Uniswap V3. Within a couple of months, liquidity providers as a
position’s price range. In the DAI-USDC pool, where both tokens whole appear to learn that they must select bigger price ranges if
are pegged to the US$ and, thus, intended to have the same value they want to hold their liquidity position for longer times, as we
at all times, the difference between the mean of a position lifetime show in Section 4.2.
and the mean of the time spent ITM is practically in-existent (cf. To conclude the general analysis of liquidity positions and pools
Figure 11b). on Uniswap V3, we consider both the pool’s realized volatility as
In close relation, we consider the median position width, mea- well as its mean daily trading volume (Figure 13). The realized
sured in bps, of liquidity positions in the three pools in Figure 12. volatility measures the assets historic volatility. More specifically,
Note that we consider the median as a sole position with an infinite the realized volatility 𝜎𝑟 is given by:
price range that would have a too significant impact on the mean v
u
t 𝑇
position width. We observe that for the stable pair (cf. Figure 12b), 365 ∑︁ 𝑆𝑖
ln ,
the median position size is tiny with 4bps and almost constant soon 𝑇 𝑖=1 𝑆𝑖−1
after the pool’s creation. The small price movements in the pool
make it easy for liquidity providers to choose a capital-efficient, where 𝑆𝑖 is the pool’s price, 𝑇 is the number of days over which
i.e., small, price range without the risk of the pool’s price falling 𝜎𝑟 is measured, and the factor 365 scales to volatility to one year.
out of their price range. For the two normal pairs (cf. Figure 12a), Thus, for each 30 day window in our data set, we plot the realized
the median width of a liquidity position is significantly larger by a volatility and the mean daily trading volume. We observe only small
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

(a) USDC-WETH (f=0.3%) and WBTC-WETH (f=0.3%) (b) DAI-USDC (f= 0.01%)

Figure 13: Mean daily trading volume and realized volatility for each 30 day window in three Uniswap V3 pools.

realized volatilities in the DAI-USDC (cf. Figure 13b). In the USDC- to erroneous returns due to the finite precision of ERC-20 tokens.
WETH and WBTC-WETH pools, on the other hand, we observe a In Figure 14, we plot the mean daily return as well as the width
significant realized volatilities reaching 140%. Notice that the real- of the position’s price range for all liquidity positions that were
ized volatility in the USDC-WETH pool is significantly larger than active for at least a day. Note that we calculate a position’s daily
in the WBTC-WETH pool. While this finding might appear counter- return according to Equation 3 at the end of each whole day during
intuitive initially, it stems from the prices of Bitcoin and Ether being the position lifetime. A position’s width represents a position’s
correlated, and, therefore, leading to a less volatile relative price. In price range: the larger the width, the larger the price range. While
general, the higher the pool’s price volatility, the higher the pool’s vastly different patterns appear for normal pairs and stable pairs,
acquired fees must be to compensate for the impermanent loss. In we observe similar patterns within a category. Figure 14a shows
Figure 13 we observe that for all three pools there is a correlation the average daily return and position width of individual liquidity
between the realized volatility and the mean daily trading volume positions for the normal pairs. We notice that the magnitude of the
over the same 30 day period. This correlation, exceeding 0.75 in mean daily returns can be significantly larger for small position
each pool, is promising for liquidity providers. However, while we widths than for large position widths, consistent with our prediction
find this correlation within a pool, it is not reflected between pools. in Section 4.2. While the magnitude of the mean daily returns is
While liquidity providers must hope for higher trading volumes in significantly larger for small position widths, liquidity positions
more volatile liquidity pools, we find that the mean daily volume exhibit both positive and negative returns. The magnitude of the
in the DAI-USDC pool is largely similar to that in the significantly mean daily returns tends to be smaller in the slightly less price
more price volatile WBTC-WETH pool. Finally, it must be noted volatile WBTC-WETH pair pools, as well as in the respective pools
that while higher volatility may lead to higher trading volume and with the smaller fee. In particular, we want to point out that there
more fees collected, as seen in the previous section, the probability are a few liquidity positions with daily returns of around -20% in
that the price drops out of the liquidity position also increases with the USDC-WETH pools. These positions were only active for a little
volatility. Furthermore, volatility will also increase the probability longer than a day in mid-May of 2021. During this time, the price
of larger losses due to impermanent loss. Thus, high volatility is of Ether dropped by around 30% on a single day [4]. For large price
not per se in the interest of the liquidity provider. ranges the mean daily returns are very close to zero. Thus, liquidity
providers can earn significant returns with small width liquidity
positions in normal pairs, but at the same time, the risks are higher.
5.3 Performance Statistics of Liquidity Choosing larger liquidity positions minimizes risks, but the mean
Positions daily returns are unlikely to be significant.
We continue the analysis with an investigation of the performance For the stable pair a wildly different pattern appears (cf. Fig-
of individual liquidity positions in six Uniswap V3 pools: USDC- ure 14bz. While the magnitude of the daily returns also decreases
WETH (f∈{0.05%,0.3%}), WBTC-WETH (f∈{0.05%,0.3%}) and DAI- with the position width, they are never significantly negative. Due
USDC (f∈{0.05%,0.01%}). These pools include those with the highest to the negligible price volatility in both stable pair pools, there is
volume as well as those with the highest liquidity on Uniswap no impermanent loss, the driving factor of negative returns. Ad-
V3 [10]. Further, we include all fee tiers of a given pair if the tier ditionally, while the mean daily returns are more significant for
holds significant liquidity, allowing for additional comparisons. liquidity positions with small price ranges, they only reach about
Throughout this section, we analyze all liquidity positions with a 0.3% in the most extreme cases and are, thus, significantly less than
liquidity deposit in excess of US$ 0.0001. Smaller positions can lead for normal pairs.
Risks and Returns of Uniswap V3 Liquidity Providers

(a) USDC-WETH (f∈{0.05%,0.3%}) and WBTC-WETH (f∈{0.05%,0.3%}) (a) USDC-WETH (f∈{0.05%,0.3%}) and WBTC-WETH (f∈{0.05%,0.3%})

(b) DAI-USDC (f∈{0.05%,0.01%}) (b) DAI-USDC (f∈{0.05%,0.01%})

Figure 14: Mean daily returns of individual liquidity posi- Figure 15: Mean daily returns of individual liquidity posi-
tions, depending on the position width, in six Uniswap V3 tions as a function of the position lifetime in six Uniswap
pools. Observe the significant spread in daily returns for nar- V3 pools.
row positions. The mean of each of the data series in Fig-
ure 14a is negative, thus, on average, liquidity providers lose
money in comparison to holding the assets and are hence most liquidity positions exhibit similar daily returns independent
not compensated for the additional risk of providing liquid- of their lifetime. As liquidity provider returns are mainly influenced
ity. by the pool’s volume and available liquidity, daily fluctuations of
returns are less significant.
The negligible volatility of the daily returns received by liquidity
The difference between normal and stable pairs is as apparent positions in the two stable pair pools is further highlighted in the
when plotting the mean daily returns of individual liquidity posi- risks and return analysis of liquidity positions in Figure 16. We
tions against the lifetime of a position (cf. Figure 15). While the plot the daily mean returns against the volatility of the returns for
magnitude of the daily return tends to decrease with the lifetime each liquidity position in Figure 16 in order to analyze the risks and
of a position, this trend is more apparent in the four normal pair returns of liquidity positions. Note that we only include positions
pools (cf. Figure 15a). Only liquidity position with a short lifetime whose lifetime exceeds 30 days to allow for more representative
tend to experience stark positive and negative daily returns. Thus, calculations of return volatility. Thus, positions with short lifetimes,
garnering significant profits as a liquidity provider requires active that generally experience the most extreme daily returns (cf. Fig-
management, indicating that providing liquidity in Uniswap V3 is ure 15), are not in the data set. Generally speaking, higher volatility
a game reserved to professional traders. In Figure 15b, we observe in the returns of an investment suggests greater risks. We find that
that more extreme values for the mean daily return values are only in the two stable pair pools (cf. Figure 16b), the daily volatility of the
present for liquidity position’s with a shorter lifetime. These out- daily returns is incredibly small. Additionally, all liquidity positions
liers stem from variations in the pool’s daily volume. In general with lifetimes exceeding 30 days exhibit mean daily returns rang-
 Lioba Heimbach, Eric Schertenleib, and Roger Wattenhofer

(a) USDC-WETH (f∈{0.05%,0.3%}) and WBTC-WETH (f∈{0.05%,0.3%}) (a) USDC-WETH (f∈{0.05%,0.3%}) and WBTC-WETH (f∈{0.05%,0.3%})

(b) DAI-USDC (f∈{0.05%,0.01%}) (b) DAI-USDC (f∈{0.05%,0.01%})

Figure 16: Mean daily returns and volatility of daily returns Figure 17: Mean daily returns of individual liquidity posi-
of individual liquidity positions for six Uniswap V3 pools. tions and 5% CVaR of daily returns for six Uniswap V3 pools.

ing from 0% to 0.04%. Thus, while liquidity position’s on Uniswap we also consider the conditional value at risk (CVaR) of the ana-
V3 experience little to no risks of losing money, the returns are lyzed liquidity positions (cf. Figure 17). The 5% CVaR represents the
generally small. For the four normal pair pools (cf. Figure 16a), we, expectation value of the return of an investment in the 5% worst
as expected, observe more significant mean daily returns also for cases [26]. The CVaR is one of the most frequently used risk mea-
positions with lifetimes exceeding a month. However, this comes sures, as it is not influenced by higher than average returns and
at the cost of a higher volatility of the daily returns, suggesting best reflects the tail-risk behavior of investments. It is therefore
greater risks. Note that this is in line with our expectation, as the sometimes referred to as the average worst case loss. We observe
risks presented to liquidity providers stem from the impermanent that the CVaR is positive for the vast majority of liquidity positions
loss, which is driven by price fluctuations between the pair’s two in the two stable pair pools (cf. Figure 17b). Thus, even in the 5%
assets. We also observe that liquidity positions in the USDC-WETH worst cases, the expected daily return of the liquidity positions is
experience more extreme volatility in their daily returns than in still positive, again suggesting a very moderate financial risk for
the WBTC-WETH pools. This observation is in line with both the liquidity providers in these pools. For the four normal pair pools, a
higher fluctuation in price between US$ and Ether, as well as with different picture paints itself (cf. Figure 17a). Individual positions
the higher fluctuations in volume (cf. Figure 13). We further no- experience CVaRs worse than -10%, exemplifying the risk related to
tice that for normal pairs liquidity positions that have the highest providing liquidity in non-stable pools. While individual positions
volatility have the poorest performance. The opposite is true for sta- experience these extreme CVaRs, the 5% CVaR of most positions
ble pairs where the positions with the highest volatility experience is better than -5%, but these higher risks experienced by liquidity
the most significant daily returns. providers do not appear to be rewarded with high returns. Less than
To further study the risks associated with providing liquidity, 30% of the liquidity positions in the four pools are rewarded for
Risks and Returns of Uniswap V3 Liquidity Providers

the added risks they shoulder in comparison to providing liquidity [2] 2022. Balancer. [Link]
in the stable pools, signaling that achieving high returns is not a [3] 2022. Curve. [Link]
[4] 2022. ETH/USDT. [Link]
simple undertaking as a Uniswap V3 liquidity provider. [5] 2022. GVOL Analytics. [Link]
We note here that while we measured low historical volatilities [6] 2022. Sushiswap. [Link]
[7] 2022. TerraUSD Price. [Link]
in the stablecoin pools in our analysis, this must not hold true in the [8] 2022. Top Cryptocurrency Decentralized Exchanges. [Link]
future. The low volatility in stablecoin pools relies on the continued rankings/exchanges/dex/.
confidence investors place in the respective stablecoins. The price [9] 2022. Top Decentralized Exchanges on CoinGecko by Trading Volume. https:
//[Link]/de/dex.
turbulences of UST in May 2022 [7] only highlight that providing [10] 2022. Uniswap. [Link]
liquidity should not be viewed as a passive investment – even in [11] 2022. USDC / UDT. [Link]
stablecoin pools. Both the market design of Uniswap V3 and the 0x92995d179a5528334356cb4dc5c6cbb1c068696c.
[12] 2022. USDC / USDT. [Link]
necessary decision-making from liquidity providers may pose a 0x3416cf6c708da44db2624d63ea0aaef7113527c6.
challenge to the protocol. Sudden and unexpected price changes, [13] Hayden Adams, Noah Zinsmeister, and Dan Robinson. 2020. Uniswap v2 Core.
(2020).
such as seen in May 2022 for UST or USDT, can cause the price to [14] Hayden Adams, Noah Zinsmeister, Moody Salem, River Keefer, and Dan Robinson.
move such that little or no liquidity depth is available at the market 2021. Uniswap v3 core. Technical Report. Tech. rep., Uniswap.
price, causing trading on Unsiwap V3 to cease [11, 12]. [15] Guillermo Angeris and Tarun Chitra. 2020. Improved price oracles: Constant
function market makers. In Proceedings of the 2nd ACM Conference on Advances
in Financial Technologies. 80–91.
6 CONCLUSION [16] Fischer Black and Myron Scholes. 1973. The Pricing of Options and Corporate
Liabilities. The Pricing of Options and Corporate Liabilities 81, 3 (1973), 637–654.
Providing liquidity in traditional finance markets is generally an [17] Alex Evans. 2020. Liquidity provider returns in geometric mean markets. arXiv
investment form reserved to professional traders and institutions. preprint arXiv:2006.08806 (2020).
[18] Alex Evans, Guillermo Angeris, and Tarun Chitra. 2021. Optimal fees for geomet-
The decentralized nature of the blockchain, on the other hand, ric mean market makers. In International Conference on Financial Cryptography
allows for many individual liquidity providers to join together to and Data Security. Springer, 65–79.
facilitate trustless cryptocurrency exchanges on the blockchain [19] Robin Fritsch. 2021. Concentrated Liquidity in Automated Market Makers. In
Proceedings of the 2021 ACM CCS Workshop on Decentralized Finance and Security
while earning fees. Previous works have shown that providing (DeFi@CCS), Virtual Event, Republic of Korea.
liquidity on DEXes utilizing the original CPMM design can be a [20] Lioba Heimbach, Ye Wang, and Roger Wattenhofer. 2021. Behavior of Liquid-
profitable investment, which is accessible to retail traders and only ity Providers in Decentralized Exchanges. In 2021 Crypto Valley Conference on
Blockchain Technology (CVCBT), Rotkreuz, Switzerland.
requires a few simple considerations from their side. [21] Guillaume Lambert. 2021. A Guide for Choosing Optimal Uniswap V3 LP Po-
In contrast, our work shows that obtaining high returns as a sitions, Part 1. [Link]
optimal-uniswap-v3-lp-positions-part-1-842b470d2261.
liquidity provider on Uniswap V3 is a highly complicated under- [22] Guillaume Lambert. 2021. A Guide for Choosing Optimal Uniswap V3 LP Po-
taking requiring active management and a good know-how. The sitions, Part 2. [Link]
introduction of Uniswap V3 has thus turned liquidity providing into optimal-uniswap-v3-lp-positions-part-2-4a94b0a12886.
[23] Stefan Loesch, Nate Hindman, Mark B Richardson, and Nicholas Welch. 2021.
a playing field for sophisticated investors where retail traders must Impermanent Loss in Uniswap v3. arXiv preprint arXiv:2111.09192 (2021).
be wary to avoid risking significant losses. Retail traders, unwilling [24] Satoshi Nakamoto. 2008. Bitcoin: A peer-to-peer electronic cash system. Decen-
to risk these losses and unable to perform the resource intensive tralized Business Review (2008), 21260.
[25] Michael Neuder, Rithvik Rao, Daniel J Moroz, and David C Parkes. 2021. Strategic
active management, should therefore restrict themselves to simple Liquidity Provision in Uniswap v3. arXiv preprint arXiv:2106.12033 (2021).
strategies that yield only small returns. [26] R Tyrrell Rockafellar, Stanislav Uryasev, et al. 2000. Optimization of conditional
value-at-risk. Journal of risk 2 (2000), 21–42.
[27] Gavin Wood et al. 2014. Ethereum: A secure decentralised generalised transaction
REFERENCES ledger. Ethereum project yellow paper 151, 2014 (2014), 1–32.
[1] 2021. Provide Liquidity on Uniswap V3. [Link]
5391541-provide-liquidity-on-uniswap-v3.