Skip to main content
Featured

Stablepools Are Boring? You're Wrong.

Dr. C

Someone told us that stablecoin pools were boring 😬🤯

Here's why they're wrong! 👇

Intro

  • For starters, around 1B out of 3.7B TVL in Uni v3 is a stablecoin pool. That's around 27%!
  • Of that 1B, about half of it (0.45B) is concentrated in the DAI-USDC pools at 0.01% and 0.05%.
  • But it's not just about size. Let's dig deeper into some of these pools 👇

Experiments

We investigate the following UniV3 pools:

  • USDC-USDT (0.01%)
  • USDC-USDT (0.05%)
  • USDC-DAI (0.01%)
  • USDC-DAI (0.05%)
  • FRAX-USDC (0.05%)

Let's start by looking at prices. They should be super stable...right?

Mostly! Some stable pools are more stable than others 🐷🏠

img-1

Some pools exhibit short periods of large variability, likely due to whales + MEV + UniV3 liquidity math (e.g. USDC-USDT 0.05%)

Remark: The prices 👆 are ordered on a "per-tx" basis; thus, some of the larger spikes happen inside the same block as their neighboring transactions.

Some pool stats:

  • Most stable: DAI-USDC (0.01%)
  • Most popular: USDC-USDT (0.01%)
  • Largest volatility: USDC-USDT (0.05%)
  • Approximately log-normal distribution of txs (eyeballing it)

See histograms + summary statistics👇

img-2 img-3

What about time between txs?

  • USDC-USDT (0.01%) is the busiest pool (~1 tx/min on avg).

  • FRAX-USDC (0.05%) is the quietest pool.

  • Each distribution resembles a geometric (and hence exponential) distribution with a given parameter λ (but with fatter tails)

    See figs👇

img-4 img-5 img-6

Now let's examine price correlation b/t pools:

  • Since txs occur at irregular intervals, we take the avg price over 5 minute intervals
  • Not all are as strongly correlated as one would think 🤔
  • High correlation b/t USDC-{USDT,FRAX} pools
  • Low correlation for DAI pairs

Key insights:

  • Per tx price is less stable than one would think!
  • There is, on average, ~1 tx every 5 minutes across all pools
  • Tx amount ~log-normal; time between txs ~exponential
  • Price across some pools are weakly correlated, with Dai being the odd one out

Why is this important?

  • Understanding the behavior of stablecoin pools can help us model them (e.g. for forward testing); can't use GBM
  • Understanding correlations b/t pools can help us develop beta-trading strategies (e.g. stat-arb)

(👆topics of future #ResearchBites)

Why can't we use GBM?

Intuitively, stablecoins should:

  • Oscillate around 1 (e.g., USDC)
  • Have bounded variation (not stable otherwise)

Whereas for GBM:

  • Price either becomes unbounded (µ>0) or goes to 0 (µ<0) in expectation
  • Variance increases with time