Composition of the Balancer PoolsDesign Oversights
This article intends to address the cause of losses to the Balancer LP tokens for Tracer's Perpetual Pools 3p* markets. First, let's understand how we structured the Balancer pools and the assumption on which their composition was based.
*3p is a quoting convention for the market's power leverage multiplier. It is used in place of 3x, though is approximately equivalent in most circumstances.
When Perpetual Pools are skewed, meaning that one side holds more USDC than the other, the leverage on gains is different for long and short but remains the same for losses. We call this effect polarised leverage, and it's displayed on the UI like this:
We can exploit this polarised leverage to earn excess risk-adjusted returns, using a strategy that either a) holds an equal notional value of short pool tokens and the underlying asset when skew > 1, or b) holds an equal notional value of long pool tokens and a short position on the underlying asset when skew < 1. Note that we calculate skew as long TVL / short TVL.
This strategy is called skew farming. It's similar to basis farming on products like perpetual swaps. Like funding rates for perpetual swaps, the polarised leverage for gains favours the side of pool that is less demanded by traders, which we assume is the short side most always. This is the case historically if you observe perpetual swap funding rates.
Skew is the indicator of long/short demand for Perpetual Pools. As previously stated, we can hold a combination of the underlying asset and short Pool tokens while skew is > 1 to earn what is essentially interest from the long side of the Pool. Our assumption that skew would generally be > 1 led us to choose the following composition for the 3p Balancer AMM pools:
Balancer pool composition
3-ETH/USD Weighted Pool: 50% wETH, 33.33% 3Short, 16.67% 3Long
3-BTC/USD Weighted Pool: 50% wBTC, 33.33% 3Short, 16.67% 3Long
Let’s dissect these weightings. The greater proportion of 3-Short tokens leaves the Balancer pool with a short exposure of 3 x (33.33 -16.67) = 50% which hedges the 50% long exposure presented by the underlying assets in the pool. This is what leads us to describe these pools as delta-neutral and skew-farming.
The delta-neutral description holds when the Perpetual Pools have equal TVL in each side or, in other words, the skew = 1. When the skew is > 1, the short pool token losses hedge the wETH or wBTC gains (depending on the balancer pool), but actually have gains that exceed the losses experienced by the underlying assets. That’s why the Balancer pools are described as skew-farming.
This is good for LPs. In upwards trending markets, value accrues to their position automatically every time the short tokens receive the value transfer. But what happens if skew is less than one (1 <)?
Herein lies the oversight. While skew is < 1, the Balancer pools are long exposed. During a market downturn (or a crash), traders shift their positions in the Perpetual Pools from the long side to the short, causing the leverage multiple for short positions to drop below 3. In some cases, we observed the leverage multiple for the short tokens drop as low as 1.8 during the 4th Dec. market downturn. When this happened, the gains made by the short tokens were unable to fully hedge the losses experienced by the underlying assets in the Balancer pools. This design flaw contributed to LP losses over that period.
Balancer pool LPs may experience losses for a number of other reasons, among them impermanent loss (IL) and the volatility decay of their leveraged tokens (though this effect is far less pronounced).
Usually, AMMs like the Balancer pools reach an equilibrium where the capital supplied to the pools meets the natural demand for the assets (trading activity). This is the point where the swap fees generated (minus IL) returns a certain yield for LPs. However, the yield LPs have been earning to date – in the form of TCR rewards – is artificial. This has encouraged more supply than is demanded, which exacerbates the aforementioned effects. Now that the emissions of TCR, and the TCR price itself, is in decline, LPs are barely breaking-even.
If there existed natural trading activity, the Balancer pools may be able to generate sufficient swap fees to compensate LPs at the current levels of capital. We have observed that most of the trading activity, however, is (necessary) arbitrage rather than organic demand.
We have recently made changes to the pools.tracer.finance UI to direct trading activity primarily to the Balancer pools, which we believe is the best market for retail investors to buy and sell Tracer’s leveraged tokens. This change will hopefully stimulate trading on these secondary markets and boost the swap fees for LPs.
We are still exploring a number of options to mitigate LP losses during market pullbacks, including tweaks to the current composition of the pools to reduce exposure to the underlying during such periods. It’s worth saying that the Balancer team has been extremely helpful while we investigate the capabilities of their platform, and we hope to present a solution very soon.