Analyzing Layer 2 Alpha Utilizing SushiSwap LP Tokens

5 min readOct 6, 2021


After recently expanding liquidity provisioning in Mosaic to include SushiSwap LP tokens, the team at Composable Labs has identified a new way to leverage these assets to earn liquidity providers more yield.

With Mosaic’s Proof of Concept (PoC) well underway, the team at Composable Labs is ramping up its research on how liquidity provisioning (LPing) can be optimized through our layer 2-layer 2 digital asset transferral system.

Given the recent announcement of adding SushiSwap LP (SLP) tokens to Mosaic, we are eager to provide liquidity providers (LPs) with information and tools that will help them earn the highest yield from their LPing.

By diving deeper into the on-chain data from a recently enhanced version of our in-house Research Development Kit (RDK), we show how the Composable Software Development Kit (SDK) can be used to generate a higher yield than initially theorized without increasing risk.

The Research Behind Composable’s Layer 2 SLP Alpha Strategy

To discover this new alpha strategy, we performed a comprehensive analysis of on-chain data across the SushiSwap pools USDC-WETH and USDT-WETH deployed on the mainnet (Layer 1, or L1), Arbitrum (L2), and Polygon (L2). In addition, we expanded our LP token transfer capability by opening up another set of cross-layer pools. Through this work, we were able to identify another alpha source powered by our cross-layer infrastructure technology.

In a previous announcement, we presented a similar cross-layer strategy for Curve pools and offered an initial bot based fund-manager template based on crossing moving averages.

For this research concentrating on SLP tokens, we obtained the data of all swaps occurring in all USCD-WETH and USDT-WETH pools across Mainnet, Polygon, and Arbitrum. We pulled raw transaction data from The Graph and organized by year and month per network into separate CSV files with all relevant swap information. From there, we created a global time grid with five second resolution, snapped all data onto this grid, and summarized the data in each chunk of time.

The annual percentage yield (APY) was computed in a 1-hour rolling window over the data. The swap volume (trades happening in the pool) was added up and converted to the USDX-equivalent (e.g. USDC for the USDC-WETH pool). A 0.25% fee was extracted from that. Then we scaled the data by ownership of the pool, which is the amount of dollars contributed to the pool divided by the total dollar value of the pool. Finally, we scaled the return computed in the window to be APY-equivalent.

This data allowed us to compare APYs across pools to see where LPs could earn the most yield in any given hour from September 1 to October 1, 2021. This timeframe was chosen since that was the interval in which all pools had swap data (in other words, while the mainnet and polygon pools had been active for longer, Arbitrum was added September 1st).

Data From Our Analysis

Through our analysis, we were able to plot and visually compare SLP pools in Arbitrum, Polygon, and the mainnet. Again, using a 1-hour rolling window, we were able to compute APY from LPing in each pool.

In the graph below, each bar represents an hour of fees collected giving rise to a particular APY. Some pools have higher APYs than other pools at different parts of time — and that is the point. This first graph has all the pools and networks plotted together over the one month period of time. We first show the USDC-WETH pool (we cap the APY at 100% for visual reasons):

The equivalent plot for the USDT-WETH pool is:

We see that, as expected, all pools experienced activity in this timeframe. It looks like the USDC pool was more active for Arbitrum than the USDT pool. Mainnet in both cases have a consistent underlying activity.

The following second kind of graph demonstrates regions in time when a specific pool is more lucrative to LP to when compared to the other pools. The x-axis is time and the y-axis separates the pools vertically. When a bar is drawn, it means that during that hour it was most lucrative to LP in the pool on that network.

Every time a vertical line is drawn, in the plots above, in its respective pool it means that pool is providing the best APY. For the USDC pool L2 shows up as the winner more than on USDT. For USDC Arbitrum had a burst of activity mid-month for USDC.

The Results

From our analysis, it is clear that having the ability to move SLP tokens across layers provides users with a more optimal alpha strategy than when staying single-network. This cross-layer mobility is made possible through Composable’s SDK technology.

This alpha strategy can be represented by moving tokens from Arbitrum or Polygon when a one-hour moving average of their yield difference crosses zero as presented in our previous publication linked above. Similar to our cross-layer strategy for Curve pools, an initial bot based fund-manager could be effective in managing assets to increase yield.

We should note that augmenting with any additional or alternative data and indicators (whether from traditional finance or on-chain-specific) can of course be used. It is also important to consider that building in the fees to move between pools is key to understanding if the turnover (the switching of pools) is always feasible. We refer to the alpha strategy presented in our previous work and in more detailed form here generally as “fee arbitraging.”

In addition to what we have shown before for Curve TriCrypto pool LPing, these findings present strong evidence that this new and optimized form of arbitrage can be facilitated successfully through Composable Finance’s cross-layer capable technologies.

To keep up-to-date with Composable Finance and its various interoperability and L2 infrastructure projects, make sure to follow our socials:

Twitter | Telegram | Discord | Website | GitHub | LinkedIn




Composable Finance Founder & CEO. I write about R&D at Composable Finance. Physicist by training