Paper: arXiv 2508.16589
Authors: Ziyi Wang, Carmine Ventre, Maria Polukarov
Abstract
We advance market-making strategies by integrating Adversarial Reinforcement Learning (ARL), Hawkes Processes, and variable volatility levels while also expanding the action space available to market makers (MMs). To enhance the adaptability and robustness of these strategies – which can quote always, quote only on one side of the market or not quote at all – we shift from the commonly used Poisson process to the Hawkes process, which better captures real market dynamics and self-exciting behaviors. We then train and evaluate strategies under volatility levels of 2 and 200. Our findings show that the 4-action MM trained in a low-volatility environment effectively adapts to high-volatility conditions, maintaining stable performance and providing two-sided quotes at least 92% of the time. This indicates that incorporating flexible quoting mechanisms and realistic market simulations significantly enhances the effectiveness of market-making strategies.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 6.0/10 (capped from 6.5: no out-of-sample test; see penalty flags)
- Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 6.5)
Why this score: The paper employs advanced stochastic processes (Hawkes) and adversarial reinforcement learning, requiring dense mathematical modeling. It reports specific performance metrics (e.g., 92% quoting time) under variable volatility, indicating robust simulation and data-driven evaluation.
Research Flowchart
flowchart TD A["Research Goal<br>Develop robust MM strategies<br>with expanded action space"] --> B["Methodology<br>ARL + Hawkes + Variable Volatility"] B --> C["Input Data<br>Simulated HFT Data<br>Low Vol = 2 | High Vol = 200"] C --> D["Computation<br>Train 4-Action Model<br>Test Cross-Volatility"] D --> E["Outcome 1<br>Low Vol: Optimal Performance"] D --> F["Outcome 2<br>High Vol: 92% Two-Sided Quotes"] D --> G["Outcome 3<br>Stable & Adaptive Strategy"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a market‑making strategy that combines Adversarial Reinforcement Learning, Hawkes process modeling, variable volatility levels, and an expanded multi‑action space.
Source sentence
We advance market-making strategies by integrating Adversarial Reinforcement Learning (ARL), Hawkes Processes, and variable volatility levels while also expanding the action space available to market makers (MMs).
Assumes 4
- Execution dynamics are modeled with a Hawkes process instead of a Poisson process. model
Source sentence
we transition from using a Poisson process to a Hawkes process for modeling execution dynamics.
- Two volatility regimes are considered: low (vol=2) and high (vol=200). market
Source sentence
we investigate the effects of volatility on market maker strategies by employing two distinct volatility levels: low and high.
- The adversary can be fixed, random, or strategic, with parameters drawn from predefined ranges. model
Source sentence
Strategic Adversary. In this setup, the adversary adjusts the market parameters 𝑏𝑛, 𝐴±𝑛, and 𝑘±𝑛 to maximize its own reward.
- Market makers may choose among four actions: two‑sided quote, bid‑only, ask‑only, or no quote. model
Source sentence
flexibility allows market makers to dynamically choose among providing two-sided quotes, offering only a bid price, providing only an ask price, or declining to quote.
Data
- Synthetic market simulation with Hawkes process synthetic unknown
Source sentence
we shift from the commonly used Poisson process to the Hawkes process, which better captures real market dynamics and self‑exciting behaviors.
Costs
Unclear
Finds 2
- A 4‑action market maker trained at low volatility adapts to high volatility, maintaining stable performance and quoting both sides at least 92% of the time. performance
Source sentence
Our findings show that the 4-action MM trained in a low-volatility environment effectively adapts to high-volatility conditions, maintaining stable performance and providing two-sided quotes at least 92% of the time.
- The 4‑action MM trained at high volatility performs poorly in high‑volatility test because it was trained with an always‑quoting strategy generated under low volatility. negative
Source sentence
the 4-Action MM (Train & Test @ vol=200) performs poorly in high-volatility environments, primarily because the always-quoting strategy used during training was generated under a volatility of 2, leading to a mismatch with the testing environment.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 12 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.