📶Signal-Driven LP

Signal-Driven Rebalancing

BBOX incorporates other assets' price movement into dynamically managing the liquidity provision. Users are encouraged to express their views regarding the co-movement between the Perp asset they're market-making and the signal asset. The signal-driven approach is inspired by statistical arbitrage techniques, which trade against the spread between the asset pair.

Motivation: Passive v.s. Proactive Rebalancing

Many existing vault strategies on the market fall into passive rebalancing. i.e., They tend to concentrate the liquidity to a new range only after the price gets updated. While proactive and passive rebalancing generate similar swap fees (differ by the epoch in which the rebalancing occurs), passive rebalancing generates a great degree of capital inefficiency due to price $p$ evolving outside the market-making range and being adversely selected.

[Fig 1: Impermanent Loss via Passive Rebalancing. Backtesting results were generated in the ETH-USDT Pool between 2023-01 and 2023-02 with approximately 2% - 5% LP Range. The black line is the portfolio value of holding the initial asset passively, while the blue line is the portfolio value of LPing. The gap between the two captures impermanent loss. ]

A Signal-Driven Approach to Reduce IL

We incorporate multi-asset signals in the AMM design, in which LPs can correlate price ranges $[p_a, p_b]$ with other assets or signals. The core idea is to help LPs emulate proactive market makers' strategies while preserving the convenience of easy liquidity provision.

Notice that if the signal provides robust predictive content over the (near) future mark price, the LP can greatly reduce impermanent loss through rebalancing before the mark price $p$ goes out of the LP range. We take the following hypothetical example:

• For example, suppose that the current mark price $p$ is at 1850 and Alice supplies (1000 V-USDC, 0.5262 V-ETH) into the price range $[p_a, p_b] = [1800, 1900]$, summing up to a position value of approximately 1973.508 V-USDC. Suppose that Alice’s correlated price signal indicates that the mark price has a high chance of depreciating to 1700. Then by taking action and rebalancing in advance, Alice will:

  • Withdraw liquidity at the current price & sell the 0.5262 V-ETH in the pool, resulting in approximately 1973.5076 V-USDC.

  • Rebalance the price range around [1600, 1800], which in this case is equivalent to supplying an LP with position size $L$ around 813.346.
  • Suppose that now price depreciates to 1,700 and Alice’s updated range becomes active. Then Alice’s position now includes 0.5557 V-ETH and 1001.274 V-USDC, which sums up to approximately 1946.12 V-USDC.

• Comparatively, suppose that Alice balances passively to $[1600, 1800]$ after the price depreciates to 1700. Then Alice will:

  • Force to convert her position to longing V-ETH at 1800 during rebalancing, with a position of (1.0742 V-ETH, 0 V-USDC) and a net position value at 1.0742 V-ETH * $p$ = 1.0742 V-ETH * 1700 = 1826.17 V-USDC.

  Before Price Update & Rebalancing	Position Value at 1973.508 V-USDC	Position Value at 1973.508 V-USDC
  After Price Update & Rebalancing	Position Value at 1946.12 V-USDC	Position Value at 1826.17 V-USDC
  Notes	Sell assets in advance and avoid adverse selection. Partial buy-in at the new LP range.	Forced to buy in assets at the lower bound and experience adverse selection.

  Despite earning similar swap fees given the same range width, proactively rebalancing through signals increases capital efficiency and reduces IL compared with passive rebalancing.