本文为2022 Annual Fund Review Part 2，查看 part 1请点击 Annual Fund Review of Uniswap V3-Part 1

Introduction

In part one, we reviewed uniswap lp funds performance in detail. In this part, we will discuss some assumptions and arrangements we made and how to reproduce such stats in part one. You can find all scripts at https://github.com/zelos-alpha/research/tree/master/task_uniswap_matic_funds .

Our backtester, Demeter, is heavily used in this research. It will help you with quant research and strategy evaluation.

Pool selection

We choose the pool 0x45dda9cb7c25131df268515131f647d726f50608, eth-usdc-0.05% on matic with following reasons:

1. 1. Matic: faster evm chain and lower gas fee leads more active financial behaviors.

2. 2. Eth-usdc-0.05% pool have a huge amount of data and large open funds

3. 3. Eth and usdc are mature asset in crypto, there are a lot of holders and this pair will not be easily manipulated.

Data source

All data comes from event logs of pool-related transactions. You can find download detail and data cleaning in Demeter. We get all event logs belonging to eth-usdc pool and the additional event in this transaction to parse its behavior. It is worth mentioning that some funds, like gamma, interact with the pool directly without NFT minting. We can not trace their actions by position id.

The data range is from the pool creation date (2021/12/21) to the last day of 2022 (2022/12/31).

Fund selection

a Uniswap v3-lp fund is a fund that raises money on a chain, uses smart contracts to manage assets, and participates in the liquidity market making of Uniswap v3 to obtain profits.

Excluding the swap event, we got 830250 actions that interacted with usdc-eth pool last year. The "collection" event is helpful and can be used to distinguish fund managers. Combined with the Mint/Burn event, we got a cleaned table that includes liquidity providers and their behavior.

The next step is to select the substantial funds in the market with the following criterion:

1. 1. Considerable accumulated vol in liquidity providing.

2. 2. The fund uses smart contracts to manage assets.

3. 3. Frequent rebalancing

With the additional background check on Etherscan, we got six funds and 10 EOA addresses. They are the top 10 accumulated lp providers in all EOA users. They may be centralized funds, or there may be giant whale users. Their operations are more representative than ordinary users, and their capital is closer to funds. Their performance will be a good benchmark for public funds.

The addresses are as follow:

• • Polycat is inferred from the contract name.

Return Rate Calculation

The standard calculation of the yield is as follows:

Based on the characteristics of our funds, we need to consider two further issues. First, in public funds, investors' access to the fund and the fund manager's performance influences the fund's net value in parallel. Second, cryptocurrency markets are real-time clearing and have no breaks. Therefore, we propose a new yield calculation model based on one-minute granularity.

r’ = 1 r_minute1 * r_minute2 * ... * r_minute_n= NetValueEnd / NetValueStart*

We set a theoretical market close time at the end of each minute. That is, for each minute, we re-locate the investor's capital inflow and outflow behavior to the end of that minute.

The fund's return rate calculation steps are as follows: Point A is the beginning of one minute, B is the end of one minute, C is the beginning of the next minute, B and C are at the same time in a different order.

• • a: get net value Na by current eth price.

• • a-b: calculate fee income from a to b.

• • b: get net value Nb by the price of b.

r = Nb/Na. It is the rate we want, excluding join/quit impact.

• • b-c: collect fund mint/burn during a,b

• • c: The starting point of the next cycle. Calculate net value Nc with new capital change.

The influence of capital inflow and outflow is excluded through the above method, and only the impact of LP fee income and ETH price is counted in the return rate calculation. This return rate well reflects the manager's performance for each minute. By cumulative producing over time-series of these fund returns, we got the public fund performance of last year as follows.

Calculation error

Beware that any stats is an opinion of fact rather than the fact itself. We have to trade off accuracy and simplification.

The above method will underestimate fee income if one mints during one minute since it lost part of fee income before its lp position goes into the next minute. Versa vice, it will overestimate if one burns its position. Eventually, these coupled errors will be canceled out mostly. For this reason, we consider our stats an unbiased estimation and can ignore such errors.

Another source of error comes from our Demeter, a minute-based backtester. OHLC like data is sampled data from tick data which will lose some facts. For example, Demeter cannot capture JIT impact during backtesting. However, we believe this error is relatively small since we made a full test and detailed report for Demeter. Overall, Demeter is still an excellent tool for our research. You can find the detail report in https://medium.com/zelos-research/demeter-and-suggestion-about-handling-uniswap-v3-data-382f29aece10

Disclaimer

This is a working paper representing research in progress. The report is the production of a professional study, and its contents are intended to be informational only. This article is not and should not be considered as providing investment advice. No representation or warranty is made herein regarding the information's fairness, accuracy, reasonableness, or completeness.

推荐阅读

• Optimism Liquidity Mining Program Analysis

• User Behavior Analysis on Uniswap V3 Liquidity Positions