This post will recap the last month of progress in the Shell Protocol and chart our future course.
Shell Protocol liquidity providers
As the new year was upon us, we migrated liquidity from an old set of pools to a new set of pools. Thank you to everyone who endured the high gas costs!
Our AMM (automated market maker) algorithm, the core logic that powers the pool, is designed to be flexible, to allow the system to evolve. We can adjust it by setting different parameters. When we deployed our maiden pools back in October, we (in retrospect) selected sub-optimal parameters. Basically, most of the pool’s liquidity was allocated to 1:1 trades. When stablecoins trade at exactly $1.00, our old pools had deep liquidity. But when the stablecoins deviated even slightly from their peg, liquidity would be relatively thin.
These sub-optimal parameters were hurting the performance of the pool by greatly reducing trade volume. We deployed new pools with updated parameters. These new pools spread liquidity more evenly across a price range from $0.95 to $1.05. These changes made the new pools much more competitive.
We were confident that the new pools would outperform the old pools; however, we were surprised at how much trade volume they were getting. In one day, our stablecoin pool received $30 million in swaps while having only $8 million in TVL (total value locked), peaking at a 24 hour utilization ratio (volume divided by TVL) of 385%. On that day, the pool was only 3% the size of Curve’s 3-pool, yet it fielded 25% of the trade volume!
At first, we were incredulous, fearing that this performance was a one-off. Thus we waited to see if the trend continued before writing a blog post. But day after day, our pool kept outperforming Curve. After two weeks of success, we were confident enough to draft a new strategy.
Then on January 20, we noticed trade volume had abruptly dropped. We dug a little deeper and learned that previously one of the aggregators had been underestimating Shell’s gas cost, resulting in approximately a ten to twenty dollar discount. The aggregator corrected the gas estimation on January 20, exactly when trade volume decreased. You can see it clearly in the data.
That is why writing this blog post took so long. First, we had to recalibrate our whole strategy when we realized that our new pool was able to handily surpass Curve. Then, right when we were about to publish the blog post, we had to do another about-face once it became clear that the initial success was only temporary. Our apologies for the delay!
Although temporary, the boosted trade volume was a great stress test for Shell. You never know for sure how your protocol will respond to hundreds of millions in trade volume until it happens. The pool passed with flying colors without a hiccup.
This episode also demonstrates that stablecoin liquidity pools are a game of basis points and the difference between dominance and second place is only ten dollars. Slight improvements in efficiency can result in a substantial increase in trade volume. In particular, success revolves around two things:
gas efficiency
capital efficiency
The difference before and after January 20 shows just how decisive gas costs can be. The migration to L2 may attenuate the primacy of gas to an extent. Even then, gas optimization will be a big part of any liquidity pool.
Capital efficiency matters as much or more than gas efficiency. Despite being less gas efficient, Curve nonetheless outperforms Uniswap in stablecoin markets because it allocates a majority of its liquidity to a narrow price range around $1.00. In contrast, Uniswap allocates liquidity evenly across all possible price ranges. Because stablecoins primarily trade 1:1, Curve has much lower slippage than Uniswap per unit capital, i.e. Curve is more capital efficient.
After the gas cost correction, the new pools still outperformed the old pools. The old Shell pools were less capital efficient than the new Shell pools because they over-allocated liquidity to exactly 1:1 trades while under-allocating liquidity to trades that are close to but not exactly 1:1.
A skeleton key has the power to open any lock
A skeleton key has the power to open any lock