THORChain Data Digest: Impermanent Loss Protection in Practice

Impermanent Loss (IL) is a worry for any participant in a Liquidity Pool (LP). Thorchain’s Impermanent Loss Protection (ILP) goes a long way to alleviate this fear, providing 1% IL protection per day in the pool, up to a maximum of 100% after 100 days. In this article I dive into the numbers, and see how ILP has played out in practice since the launch of MCCN. There are two key behaviours of LPers that stand out in the data. Firstly, LPers have a strong dislike for impermanent loss, and are actively timing their withdrawals to minimise it (and hence ILP). Secondly, in times of high greed in the crypto markets, ILP payments are minimal, but in times of fear, ILP payouts are massively accelerated.

Overview of ILP:

Table of ILP Paid (Rune) by pool, since launch of MCCN

A total of 11,515.72 Rune has been paid out in ILP since the launch of MCCN, which is equivalent to 85,671 USD at the time of writing. This has been paid out across 17 different pools, which are shown in the table above. In the table we can see the current size and APY of the pools, the total ILP Paid (Rune), as well as the number of withdrawals made from each pool, and the average ILP payment per withdrawal.

87% of all ILP paid has been through the native BTC and ETH pools. This is more than expected, given they account for just 45% of total liquidity. It is due to their relatively poor performance compared to Stable Coin pools, which have excelled in the recent market downturn. It is no secret that Stable Coin pools have had some of the best APYs lately, all currently ≥ 39%, and this had led to them having the lowest ILP payouts of all pools, relative to their size. BUSD has averaged 0.11 Rune paid in ILP per withdrawal, USDT 0.02 Rune, and USDC just 0.01 Rune.

The highest ILP payout relative to number of withdrawals of any pool is Litecoin. However, this is due to a single large withdrawal on the 23rd June, when Litecoin saw a huge price dive, and appears just to be an anomaly. The sample size of withdrawals for the Litecoin pool is small enough that single large withdrawals can skew the data, so I wouldn’t draw any conclusions on this data point just yet.

LPers are timing their withdrawals to avoid IL, and hence ILP:

ILP Paid (USD) and Number of Withdrawals, since launch of MCCN

Here we can see the daily withdrawal count (WC) from liquidity pools against the ILP Paid on MCCN since its launch. The day axis shows days since launch of MCCN. Basic intuition says that the higher the WC, the higher the ILP Paid should be, since more withdrawals means more opportunities for payouts. However, we see the exact opposite in the data. WC and ILP Paid are strongly inversely correlated. This gives us an important insight into the behaviour of LPers. When IL occurs, and ILP has to be paid on a withdrawl, investors heavily avoid withdrawing. Similarly, investors have a clear preference to withdraw when little to no IL will be suffered. Put simply, LPers strongly avoid IL, and hence ILP.

This is encouraging in two ways. Firstly, it is good news for the Thorchain treasury. In periods of high price volatility, where significant ILP might have to be paid, people are hesitant to withdraw, so liability is in practice minimal. Second, it shows the effectiveness of education in the Thorchain space, and hints at the success of platforms such as runeyield.com and LP University. Investors are clearly paying attention to this, and only withdrawing when IL is minimal.

An important question is will this behaviour continue in the long run for Thorchain? My feeling is this might be a phenomenon exclusive to the early days of MCCN. ILP builds up over time, providing 1% coverage per day in the pool. As people pool for longer, and achieve up to 100% ILP coverage, we may see a stronger willingness to withdraw regardless of IL, and hence a more consistent payout pattern of ILP over time.

Rune Price Crashes are the trigger for ILP spikes:

Rune Price (USD) with ILP Paid (USD), since launch of MCCN

If we plot the Rune price against the ILP payouts, we can clearly see the ILP spikes occur when Rune price crashes. There are three days of huge ILP payouts: day 18 after launch of MCCN, day 45 and day 74, which corresponds to the 1st May, 28th May and 26th June.

What is interesting to consider here is that impermanent loss responds equally to rises and falls in price. A doubling of price of one asset in a pool will cause the same IL as a halving in price of the asset. However, in periods of significant Rune growth, such as between days 20 and 40 on our graph, ILP Paid is minimal. Again, this gives an important clue into the behaviour of LPs. When times are good, and Rune is pumping, LPs will keep their funds in the pool, and very little IL is actually realised. However, when times are tough, and Rune price crashes, significantly more of this IL is realised. ILP payouts are not linked 1:1 with IL, but instead appear to accelerate in market crashes.

Fear accelerates ILP Payouts:

Drawing from this idea, I wanted to investigate more broadly whether fear in the market accelerates ILP Payouts. To measure fear I am using an aggregated measure of crypto market sentiment, which determines how greedy or fearful the market is. A full breakdown of the measure can be found here:

https://alternative.me/crypto/fear-and-greed-index/

Below is a table of ILP Paid (Rune) when the crypto market has a greedy, neutral and fearful sentiment:

ILP Paid (Rune) across different market sentiments

The first column gives the total ILP Paid out in periods of different market sentiment. The second column gives the average daily payouts in these periods. We can clearly see that payouts are much higher when the market is fearful, averaging 439 Rune paid per day in ILP, versus just 3.6 Rune in periods of greed.

One potential flaw in this data is that the periods of greed in the market were shortly after MCCN launched, while the periods of fear have been later on. Since ILP builds up 1% per day in a pool, the trend we are seeing here might simply be the build up of ILP over time, rather than any real impact of fear/greed on ILP. To account for this, I created a “Time Adjusted” measure, which scales the ILP payouts depending on how recent they were. This is what we are seeing in the “TA ILP Payout” and “TA Daily Average ILP Payout” columns.

Looking at our Time Adjusted Payouts, we see exactly the same trend. In periods of fear, the average Rune payout is 646, while in periods of greed it is just 43.9. Significantly more ILP is being paid out when the market is fearful.

Explanation of Time Adjusted Columns:

At the time of writing MCCN has been launched for 85 days. All ILP payments are multiplied by the scalar 85/(day_ILP_Paid).

For example, an ILP payment paid today would be multiplied by 85/85 = 1. An ILP payment paid on the second day of MCCN launch, when max ILP is 2%, would be multiplied by 85/2 = 42.5. This allows us to better capture the true effects of greed and fear on ILP Payments.

Conclusion

Impermanent Loss occurs both when prices rise and prices fall, but the trend of Impermanent Loss Protection payouts is clear. A greedy, booming market will choose not to realise their Impermanent Losses, and leave their funds in pools to accrue fees. However, when fear strikes the market, LPers are much more likely to bite the bullet and claim their ILP. There is also a clear pattern in the withdrawals of LPers, who actively avoid withdrawing in periods of high IL, and instead time their withdrawals to periods where IL, and hence ILP is minimised.

Msc Maths and Computer Science Student, Tsinghua University