Paper: arXiv 2508.04344
Authors: Charalampos Kleitsikas, Stefanos Leonardos, Carmine Ventre
Abstract
Financial models do not merely analyse markets, but actively shape them. This effect, known as performativity, describes how financial theories and the subsequent actions based on them influence market processes, by creating self-fulfilling prophecies. Although discussed in the literature on economic sociology, this deeply rooted phenomenon lacks mathematical formulation in financial markets. Our paper closes this gap by breaking down the canonical separation of diffusion processes between the description of the market environment and the financial model. We do that by embedding the model in the process itself, creating a closed feedback loop, and demonstrate how prices change towards greater conformity to the prevailing financial model used in the market. We further show, with closed-form solutions and machine learning, how a performative market maker can reverse engineer the current dominant strategies in the market and effectively arbitrage them while maintaining competitive quotes and superior P&L.
Complexity vs Empirical Score
- Math Complexity: 9.0/10
- Empirical Rigor: 3.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper presents a sophisticated mathematical framework with stochastic calculus, closed-form solutions, and HJB equations, yet relies on theoretical derivations and simulations rather than real-market backtests or live data.
Research Flowchart
flowchart TD A["Research Goal:<br>Formulate Market Performativity<br>in Financial Models"] --> B["Core Methodology:<br>Embed Model in Market Process<br>Create Feedback Loop"] B --> C["Data/Inputs:<br>Market Price Dynamics &<br>Prevailing Algorithmic Strategies"] C --> D["Computational Process:<br>1. Closed-Form Solution<br>2. Machine Learning Implementation"] D --> E["Outcome 1:<br>Prices Converge<br>toward Model Conformity"] D --> F["Outcome 2:<br>Performative Market Maker<br>Reverse-Engineers Strategies"] D --> G["Outcome 3:<br>Effective Arbitrage<br>maintaining Competitive P&L"]
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 formal mathematical framework that embeds model‑driven dynamics into price diffusion, creating a closed feedback loop, and introduces a performative market‑making strategy that can reverse‑engineer dominant strategies and arbitrage them while maintaining competitive quotes.
Source sentence
Our paper closes this gap by breaking down the canonical separation of diffusion processes between the description of the market environment and the financial model. We do that by embedding the model in the process itself, creating a closed feedback loop, and demonstrate how prices change towards greater conformity to the prevailing financial model used in the market.
Assumes 3
- Exponential utility is assumed for the market maker. model
Source sentence
Assume exponential utility and price dynamics of the form (2), and suppose that the conditional expectation E𝑡,𝑠[𝑆𝑇] is affine in 𝑠.
- The price dynamics follow the form (2) and the conditional expectation of the terminal price is affine in the current price. model
Source sentence
Assume exponential utility and price dynamics of the form (2), and suppose that the conditional expectation E𝑡,𝑠[𝑆𝑇] is affine in 𝑠. Then, under Assumption 1, the market-making optimisation problem admits a suboptimal solution for its value function 𝐻with a reservation price 𝑟∗ PS which is performatively stable if and only if it satisfies 𝑟∗ PS(𝑡,𝑠) = 𝑠−𝑞𝛾𝜎2 (𝑒𝜖(𝑇−𝑡) −1) 𝜖 under the performative price dynamics 𝑑𝑠𝑢=
- Linear approximation of the trading intensities and Assumption 1 are used to derive the performatively optimal reservation price. statistical
Source sentence
Therefore assuming exponential utility, the linear approximation of the trading intensities and Assumption 1, we can use the OptMM(𝑟∗) operator to derive the performatively optimal reservation price 𝑟∗ 𝑃𝑂.
Data
Synthetic data only.
Costs
Unclear - sources disagree
Finds 4
- Performative strategies consistently outperform the A&S baseline and the symmetric benchmark, achieving higher P&L across all risk‑aversion levels. performance
Source sentence
Performative strategies consistently outperform the A&S strategy with higher PnL across all 𝛾’s and the symmetric strategy after 𝜖surpasses a break-off point.
- The performative strategy exhibits superior P&L profiles compared to benchmarks in all simulated settings. performance
Source sentence
We notice a similar behaviour; in all cases, the performative strategy exhibits superior P&L profiles compared to the benchmarks.
- The performative advantage becomes noticeable for ε larger than 4, where the signal‑to‑noise ratio of the performative impact is strong enough to be exploited. performance
Source sentence
With respect to our A&S baseline, we notice that this happens for values bigger than 𝜖= 4.
- The P&L of the performative‑aware agent increases (almost) linearly with ε, unlike the deteriorating performance of the A&S model and the constant performance of the symmetric strategy. performance
Source sentence
Lastly, the fact that the P&L of 𝑟∗ 𝑃𝑂increases (almost) linearly with 𝜖in contrast to the deteriorating performance of A&S and the constant performance of a symmetric shows that the model does not just passively accumulate inventory but acts strategically just like we theoretically concluded in Section 5.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (32)
- High-frequency market-making with inventory constraints and directional bets (2012)
- High Frequency Trading in a Limit Order Book (2008)
- Optimal dealer pricing under transactions and return uncertainty (1981)
- Dealing with the inventory risk: a solution to the market making problem (2012)
- Pricing Interest-Rate-Derivative Securities (2015)
- Order-Flow and Liquidity Provision (2015)
- Robust Market Making via Adversarial Reinforcement Learning (2020)
- An Engine, Not a Camera: How Financial Models Shape Markets (2006)
and 24 more
- The Self-Fulfilling Prophecy (1948)
- The Predictability of Social Events (1954)
- Social Life as Bootstrapped Induction (1983)
- Multi-agent Performative Prediction: From Global Stability and Optimality to Chaos (2022)
- Constructing a Market, Performing Theory: The Historical Sociology of a Financial Derivatives Exchange (2003)
- The modes of performativity: A meta-theoretical review (2025)
- Performative Prediction (2021)
- Stochastic Optimization for Performative Prediction (2020)
- Outside the Echo Chamber: Optimizing the Performative Risk (2021)
- Performative Reinforcement Learning (2023)
- Performative Reinforcement Learning in Gradually Shifting Environments (2024) — link
- Performative Power (2022) — link
- Performative Prediction in a Stateful World (2022)
- Anticipating Performativity by Predicting from Predictions (2022)
- Buy Low, Sell High: A High Frequency Trading Perspective (2014)
- Algorithmic Trading with Model Uncertainty (2017)
- Model-based gym environments for limit order book trading (2022) — link
- Reinforcement Learning for High-Frequency Market Making (2018)
- Mbt-gym: Reinforcement learning for model-based limit order book trading (2023) — link
- Robust Market Making: To Quote, or not To Quote (2023) — link
- Multi-Agent Reinforcement Learning for Market Making: Competition without Collusion (2025) — link
- Algorithmic and high-frequency trading (2015)
- The Financial Mathematics of Market Liquidity: From Optimal Execution to Market Making (2016) — link
- Stochastic Differential Equations: An Introduction with Applications (2003) — link
0 of 32 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.