Skip to content
All library documents

Market-Making Models and Practical Quoting Decisions

Article Quant Q&A · Author: confucius_is_confused
Read the original

Summary

The document compares academic optimal-control models, especially Avellaneda–Stoikov, with the concerns a practitioner might face. The question highlights inventory-based quote skew, model complexity across many assets, and omitted inputs such as predictive signals and client toxicity. It proposes that real quoting decisions may combine interpretable components rather than rely on a single theoretical control framework.

The answer broadly agrees that such academic models are not necessarily used directly in production and describes practical market making as combining toxicity analysis, predictive analytics, and proprietary methods. This is a high-level characterization, not evidence from a named firm or a detailed alternative model. It also cautions that market-making methods vary by market segment, including equities and fixed income. The note motivates practitioner research but does not provide implementation guidance or compare approaches quantitatively.

Key ideas

  • Inventory-based optimal-control models provide a theoretical framework for quote placement.
  • Scaling such models across many assets can raise dimensionality challenges.
  • Practical quoting may incorporate client toxicity and predictive signals alongside inventory.
  • Market-making methods can differ across asset classes and trading segments.
  • The response is qualitative and does not document specific firm practices or model performance.

Tags

Market makingMarket microstructureExecutionEquitiesFixed income
Full text
# Market Making in practice


# Market Making in practice












I read some of the papers in the market making literature such as: Avellaneda - Stoikov market making model but was wondering if these types of models are actually use in pratice?

It seems that when the number of assets traded grow then applying that type of models is difficult because of the curse of dimensionality.

Also I feel that these models are missing a lot. They always assume we have no alpha (predictive power of the mid-price) and basically just put bid/ask depending on the inventory. I feel that prop shop for example would have a less 'statistical way' of treating market making by doing client toxicity analysis for example (and skew the quote accordingly), having a bit of alpha and so on. So that at the end you can decompose the skewness of your quote in different component that are highly interpretable.

Hence I believe that using stochastic optimal control the way it's done in papers like Olivier Geant papers or Stoikov paper is not what is used in pratice in big MM firms like Citadel Securities, HRT and so on...

So my question is am I correct? and do you know any papers that have a different approach (a more practitioner approach) to making market making rather than using optimal stochastic control?

## Answer by Sane (score 2)

https://quant.stackexchange.com/a/79240

While academic models such as the Avellaneda-Stoikov market making model provide a theoretical framework for market making, they may not necessarily be directly applicable in practice. In practice, market making firms often use a combination of techniques, including client toxicity analysis, predictive analytics, and other proprietary strategies to optimize their market making activities.

Also, MMing nuances are different in different segments, e.g., equity, fixed income.

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.