Overview

The following charts show the total and average daily volume in the v2 and v3 pool. The difference in volume is calculated as v2 minus v3 volume. Although total volume difference between the two (turquoise bars) looks quite volatile at first sight, this is mainly due to a few outliers in the v2 pool. In May, the v2 pool has around 1.4 billion USD more total volume than v3. This difference keeps decreasing and by August it is the complete opposite - v3 total volume is 1.4 billion more than v2.

When looking at the difference in average volume, this is more obvious. The v3 pools have consistently generated more volumes than v2 since mid-May (excluding 5 outlier days). The average daily volume in v3 is around 20-25k USD more than v2, except for July it is only around 10k USD more.

Top 5 total volume pools

It is also interesting to see if the difference in v2 and v3 volumes has stabilised for the top LP pools with the highest total volume.

For both v2 and v3, USDC-WETH pool and WETH-USDT pool come to the top 5 by total volume. So the daily trade volume difference between v2 and v3 is also analysed below for these two most popular pools.

WETH-USDT LP pool

The v3 pool has generated more total trade volumes than v2 pool from early on, but the difference has significantly increased since the end of July, where v3 has on average around half a billion more volume than v2.

USDC-WETH LP pool

On the other hand for USDC-WETH LP pool, the total daily volume has been significantly higher in v3 than in v2. The difference between the two seems quite volatile and it's not clear if the difference will stabilise at around 1 billion.

Conclusion

By looking at the total daily volume and average volume between Uniswap v2 and v3, and testing statistically if the volume difference has stabilised in terms of total volume, the following conclusions can be made:

• v3 has been generating more volumes, especially in the recent 1 month
• the total volume difference between the two has not yet stabilised and is still quite dynamic and time dependent
• for the two largest total volume LP pools, one of them (WETH-USDT) shows a stationary and stabilised volume difference between v2 and v3; whereas the other (USDC-WETH) does not > Therefore, it is still a bit early to conclude whether the difference in total volume for v2 and v3 has stabilised. Across all LP pools, the difference at least shows some statistical stationarity, which means the difference in volumes in the future can be predicted using time-series models; whereas when looking at different LP pools, the story could be different and some of them might not be predictable.

Has the difference stabilised between v2 and v3?

To answer this question, purely using graphs and eyeballing is difficult and could be inaccurate. The statistical definition of stationarity could be of good use here.

Stationarity in time-series means the individual data points do not depend on time - the mean and standard deviation do not change over time. It does not mean the values do not change at all, but only the statistical properties embedded in the time-series do not change, so we can discover and make use of these properties to predict future trends.

Augmented Dickey-Fuller (ADF) test is a common way to test if a time-series is stationary. The hypothesis of the test are: >- Null Hypothesis: the time-series is non-stationary (unit-root) >- Alternative: the time-series is stationary > More details of the test can be found here https://en.wikipedia.org/wiki/Augmented_Dickey–Fuller_test

The reason why a stationary process is a good way to assess if the difference between v2 and v3 volume has stabilised is because it is a stable time-series that moves around a constant mean and standard deviation, and it is predictable. If we ever want to predict the future difference between the two, there are a lot of model options to choose from for a stationary time-series.

The ADF test results for total volume for the entire 6 months period show the difference in volume between v2 and v3 is not stationary, hence the trend has not stabilised.

> ADF Test for total volume (past 6 months): >- Dickey-Fuller Stats = -2.914 >- Lag order = 5 >- p-value = 0.194 -> fail to reject null hypothesis at 5% significance level > - The difference is not stationary and hasn't stabilised.

Using the same ADF to test if the difference between v2 and v3 for WETH-USDT LP pool has stabilised, the results show that the total volume has stabilised in the past 6 months. > ADF Test for total volume (past 6 months): >- Dickey-Fuller Stats = -3.4785 >- Lag order = 5 >- p-value = 0.047 -> reject null hypothesis at 5% significance level >- The difference is stationary and has stabilised.

The ADF stationary test results show that the total volume hasn't stabilised in the past 6 months. > ADF Test for total volume (past 6 months): >- Dickey-Fuller Stats = -2.9117 >- Lag order = 5 >- p-value = 0.197 -> reject null hypothesis at 5% significance level >- The difference is not stationary and has not stabilised.