Thread by @AstroTechLabs

1/ #Terra Classic : A ponzi or a mechanism design failure?
Terra Classic was a use-case-specific Cosmos blockchain designed to create a market between #LUNA and #Terra stables, each coin assumed to be at price pegged to the denominated currency, eq, US dollars in case of #UST

2/ #Terra's market module was designed to do this via a constant product market-making algorithm, where,

base pool = pool_size * 2
terra pool = base_pool + terra_delta

where terra delta tracked the total net #terra coins (denominated in SDR) swapped for #Luna

3/ Since continuous one-directional swaps will lead to high terra delta resulting in pool imbalance and high spreads, the delta was replenished at the end of every block by subtracting the regression amount (terra_delta / pool_recovery_period) from it,

4/ having a static (as % of terra delta) regression amount implied that the pool couldn't sustain continuous high sell pressure on the #terra coins side as that will lead to high spreads resulting in poor arbitrage execution and #Luna inflation, initiating the death loop,

5/ this is evident from our analysis of the gathered #terra crash data set. Our analysis shows how sudden high sell (burn) pressure on #ust led to a quick increase in delta resulting in significantly high spreads per trade, (~38% max spread during 1st half of crash)

6/ with such an unhealthy on-chain swap execution, it became almost impossible for arbitrageurs to bring #ust back to peg, as the on-chain execution price (~$0.54 for some blocks) was much lower than the targetted $1 as visible in the chart here,

7/ The above chart also showcases the expected profits (in USD) made by arbitrageurs, averaging more than $100M per hour during the peak of the first half of the crash.

8/ The current design allowed traders to extract est. $1B+ worth of value via MEV during the initial half of the crash, which involved burning around 3.57 B UST worth ~$2.5B and minting 1.2 B LUNA worth ~3.5B calculated via oracle prices provided by the terra's oracle module

9/ #Terra classic's current virtual pool design was not mature enough to handle billions worth of trading volumes without any strong organic demand, an important factor which most of the people who worked on terra were probably well aware of.

10/ We believe that there are ways to improve upon the existing market module design that will allow it to dynamically adjust its per block regression amount so it can viably live through market turmoils and still support efficient arbitrage with minimal spreads.

11/ We will soon share more updates with the #LUNC community as we make progress with our terra classic research.

Here's the repo with the associated data set and the analysis logic : github.com/AstroTech-Labs…