Using a Leading Asset to Adjust a Market-Making Mid Price

Summary

The document asks how a market maker can incorporate information from a correlated asset when estimating the mid price of the instrument being quoted. In delta-neutral market making, the mid estimate can respond to moves in a related asset that tends to lead the traded asset. Linear regression and vector autoregression are raised as possible starting points, but the answer gives a more constrained modeling suggestion.

When there is a prior belief that one asset’s returns lead the other, the proposed approach is to construct features that move in the same direction as the leading asset’s returns, then estimate their weights using nonnegative least squares. The nonnegative constraint encodes the assumed directional relationship. This is a compact modeling recommendation, not a documented empirical comparison or complete implementation guide; the source does not specify feature timing, validation, execution costs, or how to handle a changing lead-lag relationship.

Key ideas

• A related asset’s returns may provide information for updating a market-making mid estimate.
• A prior belief about which asset leads can guide the direction of predictive features.
• Nonnegative least squares can constrain feature weights to preserve the assumed direction of influence.
• The recommendation does not provide empirical validation or details for feature timing and model maintenance.

Tags

# Constructing a mid using signals from another asset


# Constructing a mid using signals from another asset












When delta-neutral market making it is important to construct a mid price. Often the mid price of the asset you are trading is influenced by another (correlated) asset. What methodologies would you recommend to combine the returns of the leading asset into the mid price for the asset I am trading? Linear regression (var) models spring to mind.

## Answer by databento (score 2, accepted)

https://quant.stackexchange.com/a/74240

If you have a prior that returns of one asset should lead the other, you could construct features with the same direction as the returns of the leading asset and constrain the weights of those features with nonnegative least squares.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.