Paper: arXiv 2510.16008

Authors: Rui Gonçalves, Vitor Miguel Ribeiro, Roman Chertovskih, António Pedro Aguiar

Abstract

This study presents the implementation of a short-term forecasting system for price movements in exchange markets, using market depth data and a systematic procedure to enable a fully automated trading system. The case study focuses on the UK to Win Horse Racing market during the pre-live stage on the world’s leading betting exchange, Betfair. Innovative convolutional attention mechanisms are introduced and applied to multiple recurrent neural networks and bi-dimensional convolutional recurrent neural network layers. Additionally, a novel padding method for convolutional layers is proposed, specifically designed for multivariate time series processing. These innovations are thoroughly detailed, along with their execution process. The proposed architectures follow a standard supervised learning approach, involving model training and subsequent testing on new data, which requires extensive pre-processing and data analysis. The study also presents a complete end-to-end framework for automated feature engineering and market interactions using the developed models in production. The key finding of this research is that all proposed innovations positively impact the performance metrics of the classification task under examination, thereby advancing the current state-of-the-art in convolutional attention mechanisms and padding methods applied to multivariate time series problems.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces novel deep learning architectures (convolutional attention, bi-dimensional CNN-RNN layers) with custom padding methods, indicating advanced mathematical modeling; it is grounded in a comprehensive, end-to-end framework using real-world exchange market data (Betfair) for trading simulation, emphasizing data pre-processing, feature engineering, and production-level deployment.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop an automated<br>trading system for price movement<br>forecasting in betting exchange markets"] --> B["Data Collection & Pre-processing<br>Market depth data from Betfair<br>(UK Horse Racing, pre-live stage)"]
  B --> C["Key Methodology Innovations<br>1. Novel Multivariate Time Series Padding<br>2. Convolutional Attention Mechanisms"]
  C --> D["Computational Models<br>RNN / Bi-directional Conv-GRU with<br>Convolutional Attention Layers"]
  D --> E["Training & Evaluation<br>Supervised learning on historical data<br>Tested on new unseen data"]
  E --> F["End-to-End Framework<br>Automated feature engineering &<br>Production trading execution"]
  F --> G["Key Findings & Outcomes<br>All innovations improved classification performance<br>Advanced SOTA for time series forecasting"]

Cited References (234)

  1. Deep reinforcement learning for the optimal placement of cryptocurrency limit orders (2022)
  2. Distributed mean reversion online portfolio strategy with stock network (2024)
  3. An adoption model of cryptocurrencies (2025)
  4. Optimizing credit limit adjustments under adversarial goals using reinforcement learning — → 2306.15585
  5. Credit default prediction from user-generated text in peer-to-peer lending using deep learning (2022)
  6. Deep learning (2016)
  7. Big data: New tricks for econometrics (2014)
  8. Implementing deep neural networks for financial market prediction on the Intel Xeon Phi (2015)
and 226 more
  1. Neural networks for time series processing (1996)
  2. Pairs selection and outranking: An application to the S&P 100 index (2009)
  3. Pairs trading and outranking: The multi-step-ahead forecasting case (2010)
  4. Long short-term memory (1997)
  5. Deep learning with long short-term memory networks for financial market predictions (2018)
  6. A Survey of Deep Learning: Platforms, Applications and Emerging Research Trends (2018)
  7. A Logical Calculus of Ideas Immanent in Nervous Activity (1943)
  8. Adam: A Method for Stochastic Optimization (2014) — link
  9. Learning Internal Representations by Error Propagation (1985) — link
  10. Weakly supervised hand pose recovery with domain adaptation by low-rank alignment (2016)
  11. Multi-view 3d object retrieval with deep embedding network (2016)
  12. A deep learning approach to block-based compressed sensing of images (2016) — link
  13. Early visual concept learning with unsupervised deep learning (2016) — link
  14. Deep residual learning for image recognition (2016)
  15. DeepCorrect: Correcting DNN models against image distortions (2017) — link
  16. Deep convolutional triplet network for quantitative medical image analysis with comparative case study of gamma image classification (2017)
  17. Human action recognition in RGB-D videos using motion sequence information and deep learning (2017)
  18. Analysis of facial expressiveness captured in reaction to videos (2017)
  19. Object category learning and retrieval with weak supervision (2018) — link
  20. Deep fully-connected networks for video compressive sensing (2018)
  21. Direct speech reconstruction from articulatory sensor data by machine learning (2017)
  22. Statistical parametric speech synthesis incorporating generative adversarial networks (2018)
  23. Audio recapture detection using deep learning (2015)
  24. Noise power spectral density estimation for binaural noise reduction exploiting direction of arrival estimates (2017)
  25. Robust acoustic event classification using deep neural networks (2017)
  26. Natural language description of remote sensing images based on deep learning (2017)
  27. Multichannel signal processing with deep neural networks for automatic speech recognition (2017)
  28. Learning to rank short text pairs with convolutional deep neural networks (2015)
  29. Deep reinforcement learning for dialogue generation (2016) — link
  30. Enhancing deep learning sentiment analysis with ensemble techniques in social applications (2017)
  31. Hdltex: Hierarchical deep learning for text classification (2017)
  32. Real-time deep learning of robotic manipulator inverse dynamics (2015)
  33. Fast single shot detection and pose estimation (2016)
  34. Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates (2017)
  35. Supervised autonomy for online learning in human-robot interaction (2017)
  36. Repeatable folding task by humanoid robot worker using deep learning (2017)
  37. Dex-net 2.0: Deep learning to plan robust grasps with synthetic point clouds and analytic grasp metrics (2017) — link
  38. Eyeball gesture controlled automatic wheelchair using deep learning (2017)
  39. Self-supervised deep reinforcement learning with generalized computation graphs for robot navigation (2018)
  40. Unsupervised obstacle detection in driving environments using deep-learning-based stereovision (2018)
  41. Integrating mobile and cloud for PPG signal selection to monitor heart rate during intensive physical exercise (2016)
  42. Deep learning for identifying metastatic breast cancer (2016) — link
  43. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs (2016)
  44. A deep learning approach to on-node sensor data analytics for mobile or wearable devices (2016)
  45. Deep learning with convolutional neural networks for decoding and visualization of EEG pathology (2017)
  46. Multiscale rotation-invariant convolutional neural networks for lung texture classification (2018)
  47. Deep learning for ocean remote sensing: an application of convolutional neural networks for super-resolution on satellite-derived SST data (2016)
  48. Deep learning approach to the Higgs boson CP measurement in H$\rightarrow$ $\tau$ $\tau$ decay and associated systematics (2017)
  49. Deep learning in color: towards automated quark/gluon jet discrimination (2017)
  50. Matching matched filtering with deep networks for gravitational-wave astronomy (2018)
  51. Stacked sparse autoencoder modeling using the synergy of airborne LiDAR and satellite optical and sar data to map forest above-ground biomass (2017)
  52. Multimodal deep learning for solar radio burst classification (2017)
  53. Deep learning for credit card data analysis (2015)
  54. Deep learning for event-driven stock prediction. (2015)
  55. Deep features for person re-identification (2015)
  56. Numerical encoding to Tame SQL injection attacks (2016)
  57. Increasing web service availability by detecting application-layer DDoS attacks in encrypted traffic (2016)
  58. Long short term memory recurrent neural network classifier for intrusion detection (2016)
  59. A new deep-Q-learning-based transmission scheduling mechanism for the cognitive Internet of Things (2018)
  60. Deep learning in finance (2016) — link
  61. Deep learning for financial time series forecasting in A-Trader system (2017)
  62. Market efficiency in person-to-person betting (2006)
  63. The role of speculative trade in market efficiency: Evidence from a betting exchange (2017)
  64. Efficient backprop (2012)
  65. Understanding deep neural networks with rectified linear units (2016) — link
  66. Deep learning (2015)
  67. Learning long-term dependencies with gradient descent is difficult (1994)
  68. Long short-term memory based recurrent neural network architectures for large vocabulary speech recognition (2014) — link
  69. Visualizing and understanding convolutional networks (2014)
  70. Power, not fear: a collusion-based account of betting market inefficiency (2014)
  71. Betting and equilibrium (1986)
  72. Why is there a favourite-longshot bias in British racetrack betting markets (1997)
  73. The utility of wealth (1952)
  74. A more general non-expected utility model as an explanation of gambling outcomes for individuals and markets (2009)
  75. Stability of choices among uncertain alternatives (2008)
  76. Biased evaluations of randomly determined gambling outcomes (1986)
  77. The degree of inefficiency in the football betting market (1991)
  78. Gluck’s second law: an empirical investigation of horserace betting in early and late races (1993)
  79. An empirical study of the impact of complexity on participation in horserace betting (1997)
  80. High level architecture for trading agents in betting exchange markets (2013)
  81. Ensembling neural networks: many could be better than all (2002)
  82. The distribution of information in speculative markets: a natural experiment (2016)
  83. Survey on multiclass classification methods (2005)
  84. A Survey on Methodologies for Handling Imbalance Problem in Multiclass Classification
  85. A Survey on Methodologies for Handling Imbalance Problem in Multiclass Classification (2020)
  86. Going Deeper with Convolutions (2014) — link
  87. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling, 2018 (2018) — link
  88. Recursive distributed representations (1990)
  89. XCM: An Explainable Convolutional Neural Network for Multivariate Time Series Classification (2020) — link
  90. Guidelines for developing an air quality (Ozone and PM2. 5) forecasting program (2003)
  91. Variable Split Convolutional Attention: A novel Deep Learning model applied to the household electric power consumption (2023)
  92. Deep learning in exchange markets (2019)
  93. Encog: Library of Interchangeable Machine Learning Models for Java and C# (2015) — link
  94. Towards Efficient Electricity Forecasting in Residential and Commercial Buildings: A Novel Hybrid CNN with a LSTM-AE based Framework (2020)
  95. Reference evapotranspiration time series forecasting with ensemble of convolutional neural networks (2020) — link
  96. Learning hidden patterns from patient multivariate time series data using convolutional neural networks: A case study of healthcare cost prediction (2020) — link
  97. Time Series Multiple Channel Convolutional Neural Network with Attention-Based Long Short-Term Memory for Predicting Bearing Remaining Useful Life (2020)
  98. 2D Convolutional Neural Markov Models for Spatiotemporal Sequence Forecasting (2020)
  99. Multiple time-series convolutional neural network for fault detection and diagnosis and empirical study in semiconductor manufacturing (2020)
  100. Individual household electric power consumption data set: UCI Machine Learning Repository, School of Information and Computer Science; 2012. (2012)
  101. Predicting residential energy consumption using CNN-LSTM neural networks (2019)
  102. Regression analysis for prediction of residential energy consumption (2015)
  103. Time Series Analysis and Prediction of Electricity Consumption of Health Care Institution Using ARIMA Model (2017)
  104. Data normalization before statistical analysis: keeping the horse before the cart (1988)
  105. Electricity consumption forecasting models for administration buildings of the UK higher education sector (2015)
  106. Short-term electrical load forecasting using the Support Vector Regression model to calculate the demand response baseline for office buildings (2017)
  107. Energy consumption prediction using people dynamics derived from cellular network data (2016)
  108. Data-driven heating and cooling load predictions for non-residential buildings based on support vector machine regression and NARX Recurrent Neural Network: A comparative study on district scale (2018)
  109. Normalization in econometrics (2007)
  110. Imaging and fusing time series for wearable sensor-based human activity recognition (2020) — link
  111. Survey on Human Activity Recognition based on Acceleration Data (2019) — link
  112. Deep Learning Models for Real-time Human Activity Recognition with Smartphones (2020) — link
  113. A prospective study (SCOPE) comparing the cardiometabolic and respiratory effects of air pollution exposure on healthy and pre-diabetic individuals (2018)
  114. Long-term exposure to traffic-related air pollution and cancer among survivors of myocardial infarction: A 20-year follow-up study (2017)
  115. Real-time air quality forecasting, part I: History, techniques, and current status (2012) — link
  116. A hybrid deep learning model with multi-source data for PM 2.5 concentration forecast (2020)
  117. ARIMA forecasting of ambient air pollutants (O 3, NO, NO 2 and CO) (2010)
  118. Ambient air quality classification by grey wolf optimizer based support vector machine (2017)
  119. Short-term prediction of air pollution in Macau using support vector machines (2012)
  120. Neural network forecast of daily pollution concentration using optimal meteorological data at synoptic and local scales (2015) — link
  121. Urban Air Quality Forecasting: A Regression and a Classification Approach (2017)
  122. Air Quality Forecasting in Madrid Using Long Short-Term Memory Networks (2017)
  123. Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation (2017) — link
  124. Reusable Patterns for Agent Coordination (2001)
  125. Sports API Reference Guide - v1.101
  126. Using GLS to generate forecasts in regression models with auto-correlated disturbances with simulation and Palestinian market index data (2014)
  127. Sports Data Mining (2010)
  128. Understanding the difficulty of training deep feedforward neural networks (2010) — link
  129. Formal development of reactive agent-based systems (2005)
  130. Trading on the Edge: Neural, Genetic, and Fuzzy Systems for Chaotic Financial Markets (1994)
  131. Adaptation in Natural and Artificial Systems (1975)
  132. Genetic Algorithms in Search, Optimization, and Machine Learning (1989)
  133. The Design of Innovation: Lessons from and for Competent Genetic Algorithms (2002)
  134. Analysis of the Publications on the Applications of Particle Swarm Optimisation (2008) — link
  135. ANN hybrid model versus ARIMA and ARIMAX models of runoff coefficient (2013) — link
  136. Word-level training of a handwritten word recognizer based on convolutional neural networks (1994) — link
  137. Learning Methods for Generic Object Recognition with Invariance to Pose and Lighting (2004) — link
  138. Show, attend and tell: Neural image caption generation with visual attention (2015)
  139. Assessing Beijing’s PM2. 5 pollution: severity, weather impact, APEC and winter heating (2015)
  140. Gradient-based learning applied to document recognition (1998) — link
  141. Convolutional networks for images, speech, and time series (1995)
  142. Deep Residual Learning for Image Recognition (2015) — link
  143. Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting (2015) — link
  144. WaveNet: A Generative Model for Raw Audio (2016) — link
  145. Attention Is All You Need (2017) — link
  146. Particle swarm optimization (1995)
  147. Convolutional networks and applications in vision (2010) — link
  148. A Functional Approach to Border Handling in Image Processing (2015) — link
  149. WWW. BETFAIR. COM: WORLD WIDE WAGERING (2005)
  150. Pricing combinatorial markets for tournaments (2008) — link
  151. Mastering the Trade: Proven Techniques for Profiting from Intraday and Swing Trading Setups (2007)
  152. High Level Architecture for Trading Agents in Betting Exchange Markets (2013) — link
  153. Roll Padding and WaveNet for Multivariate Time Series in Human Activity Recognition (2021)
  154. A method of solving a convex programming problem with convergence rate O (1/k2) (1983)
  155. On the Importance of Initialization and Momentum in Deep Learning (2013) — link
  156. What data scientists should know about deep learning (2015)
  157. Understanding Deep Neural Networks with Rectified Linear Units (2016) — link
  158. On Information and Sufficiency (1951)
  159. Estimation of Citywide Air Pollution in Beijing (2013) — link
  160. Population Exposure to PM2.5 in the Urban Area of Beijing (2013) — link
  161. Visualizing and Understanding Convolutional Networks (2014)
  162. Efficient BackProp (1998) — link
  163. Autoencoders, unsupervised learning, and deep architectures (2012)
  164. Recent advances in convolutional neural networks (2018)
  165. Anomaly Detection Based on Convolutional Recurrent Autoencoder for IoT Time Series (2020)
  166. An Optimal Stacked Ensemble Deep Learning Model for Predicting Time-Series Data Using a Genetic Algorithm—An Application for Aerosol Particle Number Concentrations (2020)
  167. On the automated, evolutionary design of neural networks: past, present, and future (2020)
  168. Cascaded Denoising Convolutional Auto-Encoders for Automatic Recovery of Missing Time Series Data (2020)
  169. A constructive development environment for parallel and distributed programs (1994) — link
  170. An architecture-independent software development approach for parallel processing systems (1995)
  171. Event based distributed real-time communication architecture for multi-agent systems (2010) — link
  172. Stochastic Gradient Descent Tricks (2012) — link
  173. Predicting the Household Power Consumption Using CNN-LSTM Hybrid Networks (2018)
  174. Predicting Consumer Loads for Improved Power Scheduling in Smart Homes (2016)
  175. Neural Networks: Tricks of the Trade - Second Edition (2012) — link
  176. RPROP - A Fast Adaptive Learning Algorithm (1992)
  177. Adaptive Subgradient Methods for Online Learning and Stochastic Optimization (2011) — link
  178. ADADELTA: An Adaptive Learning Rate Method (2012) — link
  179. Finding structure in time (1990)
  180. Long Short-Term Memory Based Recurrent Neural Network Architectures for Large Vocabulary Speech Recognition (2014) — link
  181. Deep Learning of Representations for Unsupervised and Transfer Learning (2012) — link
  182. Learning Long-term Dependencies with Gradient Descent is Difficult (1994) — link
  183. Acceleration techniques for the backpropagation algorithm (1990)
  184. Untersuchungen zu dynamischen neuronalen Netzen (1991)
  185. Striving for Simplicity: The All Convolutional Net (2014) — link
  186. Neural Turing Machines (2014) — link
  187. Deep Feature Consistent Variational Autoencoder (2016) — link
  188. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks (2015) — link
  189. Why Does Unsupervised Pre-training Help Deep Learning? (2010) — link
  190. Autoencoder for Words (2014) — link
  191. Serial Order: A Parallel, Distributed Processing Approach (1990)
  192. Backpropagation (1995) — link
  193. The Utility Driven Dynamic Error Propagation Network (1987)
  194. Generalization of backpropagation with application to a recurrent gas market model (1988) — link
  195. Approximation by superpositions of a sigmoidal function (1989) — link
  196. Approximation Capabilities of Multilayer Feedforward Networks (1991) — link
  197. On Computable Numbers, with an Application to the E ntscheidungsproblem (1936) — link
  198. Neural Networks: A Review from a Statistical Perspective (1994) — link
  199. Some methods of speeding up the convergence of iteration methods (1964) — link
  200. A Neural Network Ensemble Method with Jittered Training Data for Time Series Forecasting (2007) — link
  201. Time series forecasting using a hybrid ARIMA and neural network model. (2003) — link
  202. Least Squares Support Vector Machine Classifiers (1999) — link
  203. Large scale fitting of regression models with ARIMA errors (2000)
  204. Support vector machine with adaptive parameters in financial time series forecasting (2003) — link
  205. Chaotic time series prediction using knowledge based Green’s Kernel and least-squares support vector machines (2007) — link
  206. Application of critical support vector machine to time series prediction (2003) — link
  207. Limitations with activity recognition methodology & data sets (2014)
  208. A public domain dataset for human activity recognition using smartphones. (2013)
  209. Human activity recognition with smartphone sensors using deep learning neural networks (2016)
  210. High Accuracy Human Activity Monitoring Using Neural Network (2008) — link
  211. Real-time human activity recognition from accelerometer data using Convolutional Neural Networks (2018)
  212. Human Activity Recognition Using Wearable Sensors by Deep Convolutional Neural Networks (2015) — link
  213. A One-Vs-One Classifier Ensemble With Majority Voting for Activity Recognition. (2013)
  214. A sparse kernelized matrix learning vector quantization model for human activity recognition (2013)
  215. Time series analysis of household electric consumption with ARIMA and ARMA models (2013)
  216. The organization of behavior: A neuropsychological theory (1949)
  217. Threshold models in non-linear time series analysis (1983) — link
  218. Introduction to Modern Time Series Analysis (2007) — link
  219. Time series modelling of water resources and environmental systems / Keith W. Hipel, A. Ian McLeod ( 1994 )
  220. Time Series Analysis, Forecasting and Control (1990)
  221. Principles of Neurodynamics: Perceptrons and the Theory of Brain Mechanisms (1962)
  222. Perceptrons: An Introduction to Computational Geometry (1969)
  223. Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences (1975) — link
  224. Neural networks for self-learning control systems (1990) — link
  225. Deep Learning (2016)
  226. Learning Deep Architectures for AI (2009) — link

1 of 234 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.