Quantum Computing Predicts Stock Trends with Accuracy Exceeding Limits of Current Models

Researchers are increasingly exploring the potential of quantum computing to address complex challenges in financial modelling. Wendy Otieno, Alexandre Zagoskin, and Alexander G. Balanov from the Department of Physics, Loughborough University, alongside Juan Totero Gongora and Sergey E. Savel’ev, present a novel quantum reservoir computing (QRC) framework designed for forecasting nonlinear financial time-series data. This collaborative work demonstrates the ability of a small-scale system of up to six interacting qubits to predict daily stock trading volumes from 20 sectors between April 2020 and April 2025, and even minute-by-minute fluctuations, achieving stock trend classification accuracies exceeding 70%. Significantly, this platform-agnostic model, realisable with technologies such as superconducting circuits and trapped ions, highlights the expressive power of compact quantum reservoirs and offers a promising pathway towards practical, near-term quantum solutions for financial forecasting.

Researchers analysed future daily closing trading volumes of 20 quantum-sector publicly traded companies over the period from April 11, 2020, to April 11, 2025, as well as minute-by-minute trading volumes during out-of-market hours on July 7. The analysis identifies optimal reservoir parameters that yield stock trend (up/down) classification accuracies exceeding 86%.

Importantly, the QRC model is platform-agnostic and can be realised across diverse physical implementations of qubits, including superconducting circuits and trapped ions. These results demonstrate the expressive power and robustness of small-scale quantum reservoirs for modelling complex temporal correlations in financial data.

Quantum computing tackles multifaceted influences on stock market behaviour

Scientists have explored potential applicability to real-world forecasting tasks on near-term quantum hardware. The complex, non-linear and volatile nature of stock returns have made stock market prediction an arduous endeavor. Despite this challenge, various aspects of the market’s behaviour have been captured by mathematical models such as quantitative financial models (Black-Scholes option pricing and risk management), stochastic processes (Brownian motions, random walk models), agent-based modelling and decision-making strategies that involve the intersection of game theory and behavioural economics.

These models act as approximations where they are beneficial in forecasting controlled scenarios, but adjustments are needed to account for the evolution of real-world conditions. The Efficient Market Hypothesis (EMH) introduced by Eugene F. Fama defines a theoretical benchmark of perfect market efficiency, claiming that past prices have no predictive information, public information does not lead to consistent excess returns and public and private information cannot generate consistent excess returns making forecasting trends impossible.

In principle, perfect information market efficiency is an ideal state. In reality, real financial data can and do deviate from this perfect efficiency, exhibiting nonlinear and dynamic structures that contradict EMH, as evidenced by well-documented anomalies such as volatility clustering, volume-driven spikes, momentum, return autocorrelation and short term predictability which draw market microstructure effects, revealing clear temporal memory in price and volume formation.

These deviations are what predictive models such as Artificial Neural Networks (ANNs) and Quantum Machine Learning (QML) exploit. ANNs and QML models can detect nonlinear patterns in financial data and learn its hidden structure by operating in high dimensional state spaces where the interwoven interactions between price, volume, returns and order flow exists.

QML is advantageous in exploring extremely large, high-dimensional data spaces enabling them to learn complex correlations. Thus, if QRC can demonstrate significant predictive power for future daily closing volumes (DCV) and market trends (up/down), this is an implication of the presence of non-linear, exploitable structure in the DCV financial time series, contradicting EMH and therefore demonstrating deviations from perfect market efficiency.

To tackle all complexities associated with the inherent nature of the stock market, advanced techniques such as classical machine learning, and now quantum computing, have been explored to achieve better accuracy in stock market price prediction. This entails employing various types of specialised Artifical Neural Networks (ANNs) i.e. Long Short Term Memory (LSTM) networks, Recurrent Neural Networks (RNNs) and Self Organizing Fuzzy Neural Networks (SOFN), which have shown promise in capturing temporal dependencies.

LSTM networks, for example, achieved over 86% accuracy in classifying stock trends, hinting at the potential for genuinely novel investment strategies and a deeper, more nuanced understanding of market dynamics. However, it is important to acknowledge the limitations, as the model’s success relies on carefully optimised parameters and its performance in live, real-time trading environments remains untested.

In this paper, we introduce a scalable and minimalistic QRC framework which is employed to predict the future daily closing volumes of 20 quantum invested companies. This minimalistic system consists of a quantum reservoir comprised of up to six transmon superconducting qubits. The QR is scalable as newly added qubits easily couple with existing qubits via mutual connection with the optical waveguide.

We implement non-autonomous prediction on the encoded input datasets, which requires good non-linearity and trace of memory for better predicted accuracy, and evaluate the performance of the QR by measuring the MSE, NMSE and RMSE. We predict future daily closing volume for each company using data that span 5 years, April 11 2020 -2025, and minute-by-minute volumes of one day trading for July 7.

We also evaluate the accuracy of the predicted stock price movement. We conclude that our reservoir can predict stock prices and trends with high accuracy, demonstrating how quantum reservoirs can improve accuracy and robustness in stock price predictive models, thus overcoming complexities and volatility in complex stock price data.

Quantum reservoir computing predicts stock trends with high accuracy and minimal qubits

Achieving a stock trend (up/down) classification accuracy exceeding 86%, this work demonstrates the potential of quantum reservoir computing for financial forecasting. This level of accuracy was obtained when applying the model to predict daily closing trading volumes of 20 sector-specific publicly traded companies over a five-year period, from April 11, 2020, to April 11, 2025.

Furthermore, the model successfully predicted minute-by-minute trading volumes during out-of-market hours on July 7, 2025, reinforcing its capacity to capture short-term market dynamics. The research utilised a quantum reservoir computing (QRC) model constructed with a remarkably small system of six interacting qubits. This compact configuration effectively modelled complex temporal correlations within the financial data, highlighting the expressive power achievable with limited quantum resources.

Optimal reservoir parameters were identified through analysis, directly contributing to the observed 86% accuracy in stock trend classification. Notably, the QRC model’s platform-agnostic design allows for implementation across various physical qubit systems, including superconducting circuits and trapped ions. This flexibility broadens the potential for practical application and scalability of the model on near-term quantum hardware.

The study’s findings suggest the presence of exploitable, nonlinear structures within daily closing volume financial time series, challenging the notion of perfect market efficiency as defined by the Efficient Market Hypothesis. These results indicate a significant deviation from random market behaviour, as the model consistently identified patterns beyond those explainable by chance. The ability to accurately forecast both daily and intraday trading volumes underscores the model’s robustness and its potential for real-world forecasting tasks.

The Bigger Picture

Scientists have demonstrated a significant advance in applying quantum computing to the notoriously difficult task of financial forecasting. For years, accurately predicting stock market movements has remained elusive, hampered by the sheer complexity of interacting factors and the inherent noise within market data. Traditional models, even those leveraging vast datasets and sophisticated algorithms, often struggle to capture the subtle correlations that drive price fluctuations.

This new work sidesteps the need for large, error-corrected quantum computers by utilising a quantum reservoir computing (QRC) model built with a remarkably small number of qubits, just six. The implications of this achievement extend beyond simply improving prediction accuracy. It suggests that useful quantum computations can be performed with near-term quantum hardware, even before fully fault-tolerant machines become a reality.

This is crucial because the development of such machines is proving to be a formidable engineering challenge. A model achieving over 86% accuracy in classifying stock trends is not merely a statistical curiosity; it hints at the potential for genuinely novel investment strategies and a deeper, more nuanced understanding of market dynamics. However, it is important to acknowledge the limitations.

While the model performs well on historical data, its performance in live, real-time trading environments remains untested. Furthermore, the model’s success relies on carefully optimised parameters, and it is unclear how robust these parameters are to changing market conditions. The next step will likely involve scaling up the number of qubits, exploring different reservoir designs, and integrating the model with broader economic indicators. Ultimately, the goal is not just to predict stock prices, but to build a more comprehensive quantum-enhanced toolkit for financial analysis and risk management.