AMM Performance Update

AMM Performance Update

Now that nearly 9 rounds of trading are in the books, it's time to examine how the Lyra AMM has performed. We’ll first dive into the mechanics of the AMM and then dig into the numbers to see how the AMM has fared, and how it might perform going forward.

AMM Refresher

Lyra’s AMM accepts sUSD deposits, which are then used to quote options that traders can buy/sell. It prices these options by shifting an implied volatility surface up and down according to supply and demand: the more options traders sell, the cheaper options get, and vice versa.

In order to ensure that the call options the AMM sells are fully collateralized, when a trader purchases, say, an ETH call, the AMM purchases the base asset (sETH) from the Synthetix debt pool. This means that LPs typically accrue long exposure to the underlying asset. The below graph shows the net balance of sETH collateral for the AMM across each of the first 8 rounds, reaching as high as 4000 in round 8. At an ETH price of $2000, this represents ~$8,000,000 of exposure.

AMM ETH collateral held across each round 


To date, delta hedging has not been enabled for the ETH and BTC markets. This has meant that the PNL of those pools has been dominated by the swings in collateral. A 10% fall in the ETH price during the peak in round 8 would cost LPs $800,000 in losses, a significant amount! Periodic delta hedging would significantly reduce this risk and ensure that returns are more heavily influenced by options spread collection, as opposed to delta risk-taking.

Performance So Far

The below table shows the round performance of the ETH market pool, each round is typically 3-4 weeks, with Round 1 beginning on November 5th, 2021, and Round 8 ending on May 27th, 2022:

Round

PNL

1

-2.07%

2

0.00% 

3

-0.63%

4

-0.70%

5

+0.36%

6

+2.55%

7

-1.89%

8

-7.74%

On November 5th 2021 ETH was trading around $4,550, and at the end of round 8 (May 27 2022)  it was trading around $1800, a roughly 60% decline (with a 38% decline in round 8 alone).

As mentioned in the introduction, in the absence of delta hedging, the AMM is almost always long the underlying asset, and this has led to drawbacks in the value of the pool. In round 8 alone, the pool lost $1.58m in value, suffering $1.8m from the declines in the value of its held collateral. With delta hedging switched on, the PNL of the pool would have ranged between +$100,000 (+0.5%) and -$170,000 (-0.82%), depending on the assumed hedging frequency. This is a substantial improvement and a relatively healthy performance for the pool given disadvantageous conditions for the AMM (high volatility of implied volatility).

Delta hedging will be live for all pools during the Avalon release, so we will now fully examine how the ETH pool would have fared if delta hedging was switched on, isolating the options market-making component of the returns.

Options Trading Performance

Next, we examine AMM performance assuming it hedged its net delta across different time intervals in order to smooth any randomness associated with lucky or unlucky hedge trade timing. Running these simulations with more granular hedging and averaging out the results would smooth these discrepancies even further, but the disparities can largely be treated as random. The following graph illustrates the results.

AMM $ PNL per round with varying hedge frequency

One takeaway from this graph is that the AMM has largely been profitable across rounds assuming that its deltas were hedged (!!). Hedging frequency is a balancing act between reducing risk (by hedging often) and minimizing hedging fees paid (by hedging less). Despite this, even during extremely volatile rounds (like round 8) the underlying mechanics of the AMM have yielded a fairly stable return profile assuming hedging occurs between every 3-48 hours. As the AMM is iterated upon with more sophisticated volatility surface modeling and capital efficiency, it’s very possible that these returns could markedly improve. Regardless, this is a strong foundation with which to proceed.

Conclusion

  • The AMM has been live for 8 rounds (~7 months) and has largely suffered from declines in the prices of the underlying assets it’s traded, due to its underlying long delta bias
  • The addition of delta hedging dramatically smooths the returns of the AMM
  • If delta hedging had been live for the first 8 rounds, the AMM would have delivered positive returns for LPs, due to its position as the spread-taker in the market
  • More data and rounds are needed to fully verify these claims, but version 1 of the protocol has yielded a strong start for the AMM

View the full source code and report here! And if anyone has questions please hit us up in the #mechanism channel of the Lyra Discord.

Notes on Methodology

  • Synthetix spot hedging fees were assumed to be 10 basis points, in line with the current rate that LPs are paying for hedging and collateralization trades
  • Due to data limitations, hedging is assumed to have occurred at the next trade after the new hedging time. For instance, if the first trade with the AMM occurs at 10am Monday and the hedging delay was 24 hours, then the next hedge would occur at the first trade after 10am Tuesday. If there were no trades until, say, 2pm Tuesday, this is when the hedging would occur. For an active market like ETH this is not much of an issue (trades are quite frequent), but it is a small error that could affect these results.