Reflecting on the Trading Competition
Wow! It might not have been the most realistic options trading competition, but it certainly wasn't boring. This post will look at some high level trends we observed in the competition, how we dealt with a bot attack, and what we learned from the process.
Each participant was given $100,000 in sUSD, along with the ability to swap the sUSD into the base assets of sETH/sBTC/sLINK/sAAVE/sUNI in order to sell options. The top 25 traders were to receive payouts, with the top 3 receiving Bored Apes plus a substantial bump in the cash prize. 3 days into the competition, the prize pool was extended to pay the 26th-50th placed traders $150, and the 51st-100th placed traders $100.
What Would Happen in the Real World
The below graph (sourced from Deribit) shows the realized volatility of ETH over the trading comp period. Whilst this wouldn't exactly equate to where IV should trade (difference between implied and realized volatility), the magnitude of the Lyra IV swings would roughly reflect the movements shown here.
The below graphs show plots of the IV for one option on ETH/AAVE over the lifespan of the option (they expired on July 20th).
As the graphs show, because of:
- The capital inefficiencies surrounding selling options
- The relative ease of buying options versus selling options (selling in the testnet comp required you to go to the portfolio page, where there was a swapping interface to acquire base assets)
- The fact that a fair number of traders didn’t realize that you could sell options
- The competition incentives to maximize variance (by buying options) to achieve a top 3 rank
- The lack of a true arbitrage incentive
demand for options immediately blew any supply out of the water, driving IV (and option premiums) through the roof. This meant that people who managed to buy options right after the open saw the value of their buys 3-4x in no time, allowing them to take a swift early lead in the competition.
We figured that once people realized that the ETH options were trading at a level which implied a 50% average daily move, they would begin to close quickly (or open a short options position), resulting in a swift decline in the IV. But this didn’t happen, presumably because some traders:
1. Entered the comp
2. Bought a few options for fun
3. Forgot about the comp
Because there were no ‘real’ penalties - besides a dubious position on the Lyra Leaderboard - for booking enormous losses by holding onto overpriced options, the IVs never really returned to a reasonable level.
The graphs do show a notable drop in IV prior to the expiration trading cutoff, presumably from traders who were trying to bank their gains before their jacked up options were forced to reconcile with the real world on expiry.
We purposefully designed the competition onboarding to avoid doxxing Lyra comp participants. Because of this, we threw up multiple hurdles to prevent users from botting. Unsurprisingly, some participants found a way around the blockades, and were active manipulating certain strikes, pumping vol in some accounts to dump it in other accounts. This activity was concentrated across certain strikes in AAVE/LINK/UNI options. See if you can spot the difference between the 'normal' graphs above, and the manipulated graph below:
We were able to identify large groups of bots through the following checks:
- Joined the discord in waves, primarily on July 10th and July 19th/20th
- Shared a discord username pattern ('fName/lName', 5 random characters, etc)
- Were particularly active trading the suspect strikes/expiries
These were the accounts that were banned from the competition, and the reason we had to shut off the faucet/stop new entries to the comp with 5 days to go.
How did the AMM perform?
The AMM performed very well, although this really isn’t a reflection of mainnet performance. Because of the buy/sell imbalance in the market, the AMM was left short a decent number of options at vols 10-20x higher than where they should trade. This meant that the AMM sold options for 1000% IV on assets which realized ~90%. This would never happen on mainnet, as the presence of arbitrageurs would quickly dampen unreasonable spikes in IV. Stay tuned in the coming weeks for a full analysis of the AMM's performance in the testnet comp (sometime after the August options expire).
Turns out it's difficult to simulate a real world trading environment with a trading competition powered by an AMM. The competition did provide a fun way to see the AMM in action whilst giving traders an incentive to play around with the Lyra system and learn about options (it certainly succeeded in teaching a few users the pain of theta decay). In retrospect, a better design for a (relatively) sybil-resistant AMM testnet trading comp would involve:
- Relaxing the collateralization constraint for sellers, making it equally efficient to buy and sell
- Building an arbitrage bot with ‘infinite’ collateral to keep prices relatively in line with CEXs/other exchanges
- Having a more 'manual' element to the onboarding - i.e. messaging one of the mods in discord
Despite some speedbumps, we've really enjoyed the last couple of weeks, and hope you did too! The feedback and response from the Lyra community has been fantastic, driving improvements on both the protocol and the dApp.
Join the Community
Stay tuned for more important updates, key date announcements, and exciting opportunities by following us on Twitter.
Join the Lyra community on Discord to get involved; provide early feedback, be the first to learn about new opportunities with Lyra, and be a part of building a new DeFi community.