The research begins by theoretically examining established game structures and their associated payoff mechanisms through simplified toy market models without validating these models against real-world market dynamics. Market-derived nonlocal games are introduced where historical order-book data and backtests of combined actions define both the inputs and payoffs, diverging from reliance on predetermined idealized models. The approach initiates with conventional trading strategies grounded in market analysis, employing quantum correlations solely to enable coordination among agents’ choices.

Exact classical optima and quantifiable quantum upper bounds for the payoff are then computed. For those games exhibiting potential quantum advantages, we developed an optimisation algorithm to identify concrete entangled strategies and conclude the investigation.

Quantum advantage emerges from market-based nonlocal game design

A 3.7% average quantum-classical gap was achieved, exceeding previously attainable levels limited to theoretical models lacking real-world validation. This represents the first demonstration of potential advantage derived directly from historical market data and backtesting rather than idealized simulations. We introduced Market-derived nonlocal games; these constructed game inputs and payoffs using actual order book information and joint action tests, a departure from reliance on preselected idealized scenarios used in prior work.

Further analysis revealed this average gap wasn’t constant but emerged under specific conditions. Examining controlled variations in liquidity competition and cross-stock repricing showed how the approach identifies regimes where quantum coordination could offer an advantage. However, testing across observed market data did not reveal a resolved quantum-classical gap at measured interaction scales.

Experiments utilised three highly liquid stocks, AMD, INTC, and MU, traded across NYSE Arca, Cboe BZX, and Nasdaq venues during twenty high-volatility days between August 2024 to March 2026. The NautilusTrader v1.224.0 backtesting engine processed data from Databento Level-1 quotes with window widths ranging from one to one hundred milliseconds; yet any observed benefit remains sensitive to precise parameter settings and doesn’t account for all real-world complexities like unpredictable order book events or sophisticated algorithmic counterstrategies.

Simulated trade suppression reveals challenges for quantum high-frequency trading backtests

College Park researchers have demonstrated a methodology translating complex market dynamics into quantifiable game theory problems, potentially unlocking new avenues for quantum coordination in high-frequency trading. This success hinges upon accurately recreating realistic conditions within backtests using historical data. The approach relies on “injecting” simulated opportunities, artificial trades designed to test strategies, but introduces an unexpected complication: a ‘guard’ mechanism where the system suppresses these injected orders if they conflict with existing quotes from earlier in the simulation. Identifying and quantifying such interactions is important when translating theoretical advantages into practical trading strategies. While initial tests revealed no immediately measurable advantage over classical strategies within current measurement scales, their work successfully identified conditions where quantum coordination could outperform conventional methods through controlled variations in liquidity and price adjustments across different exchanges.

Researchers developed a method for creating game theory models of financial markets based on historical order-book data from three stocks, AMD, INTC, and MU, traded across multiple venues between August 2024 and March 2026. This allows them to test whether quantum entanglement can improve coordinated decision-making by traders operating with communication constraints; the team used Databento Level-1 quotes processed via the NautilusTrader v1.224.0 backtesting engine at millisecond scales.

Their analysis revealed that while no advantage was found in initial tests using observed market data, they could identify specific conditions where quantum strategies may outperform classical ones through controlled simulations, though any benefit remains sensitive to parameter settings. The work highlights challenges related to accurately simulating real-world trading environments within these backtests.

👉 More information
🗞 Testing Quantum Advantage in High-Frequency Trading with Market-Derived Nonlocal Games
✍️ Daiwei Zhu, Brendan Pankovich, Alireza Marefat Khah, Haakam Aujla, Ramis Movassagh, Martin Roetteler and Johannes Borregaard (College Park)
🧠 ArXiv: https://arxiv.org/abs/2611.16113

More like this