{"paper_id":"2510.16008","title":"Convolutional Attention in Betting Exchange Markets","authors":["Rui Gonçalves","Vitor Miguel Ribeiro","Roman Chertovskih","António Pedro Aguiar"],"paper_date":"2025-10-14","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2510.16008v1","page_url":"https://thequant.space/flowcharts/convolutional-attention-in-betting-exchange-markets/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper introduces innovative convolutional attention mechanisms and a novel roll padding method for multivariate time series forecasting in exchange markets.","quote":"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.","summary":"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…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The market operates as a closed‑loop system where only internal market data influence price fluctuations.","type":"market","stated":"explicit","quote":"within this restricted time frame, the market under consideration operates as a closed-loop system, where only internal market data influence price fluctuations."},{"assumption":"Trading on the exchange incurs low transaction costs.","type":"execution","stated":"explicit","quote":"Modeling exchange markets requires considering a platform where individuals and entities trade fungible items of value with low transaction costs"},{"assumption":"A minimum of 1,200 examples is required to train a model from scratch.","type":"data","stated":"explicit","quote":"To train a model from scratch, we define the minimum number of examples required as 1200."},{"assumption":"The predictive system relies solely on market data without external factors.","type":"model","stated":"implied","quote":"Consequently, predictive accuracy depends exclusively on market data, so that this study is solely concerned with purel…"},{"assumption":"The state‑machine scalping strategy assumes that if the price has already moved when the order reaches the scalp module, the opportunity is lost.","type":"execution","stated":"explicit","quote":"If the price has already moved (i.e., [PBR̸ = PBN]) when the order reaches the scalp module (i.e., the starting state), it is assumed that the opportunity has been lost"},{"assumption":"The model’s profit‑and‑loss evaluation assumes that trades are executed without slippage beyond the described market dynamics.","type":"execution","stated":"implied","quote":"These inconsistencies reinforce the importance of real-time market dynamics, order execution efficiency, and slippage in trading performance."}]},"datasets":{"status":"available","items":[{"name":"Betfair UK to Win Horse Racing pre‑live market data","provider":"Betfair","kind":"order-book","access":"proprietary","period":null,"universe":"UK to Win Horse Racing market (pre‑live stage)","quote":"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."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"l2-book"},"transaction_costs":{"status":"not_extracted","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"none","rule":"R0","candidates":{"front_matter":"unclear","evidence_card":"unclear","evidence_card_quoted":true},"card_detail":null,"card_quote":"Modeling exchange markets requires considering a platform where individuals and entities trade fungible items of value with low transaction costs","meaning":"the classifier was not confident"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"The best DL model (LSTM with multi‑head Conv1D attention) achieves an accuracy of 30.92%, only 11 percentage points above the baseline.","kind":"predictive-accuracy","quote":"Table 8 presents the best DL model, which achieves an accuracy of 30.92%."},{"finding":"LSTM‑based models outperformed all other alternatives, indicating LSTM architecture is well‑suited to this data.","kind":"performance","quote":"LSTM models outperformed the other alternatives presented."},{"finding":"Inclusion of roll padding improves the accuracy of both simple CNN and WaveNet models compared to their base versions.","kind":"performance","quote":"it is evident that the inclusion of roll padding improves the accuracy of both simple CNN and WaveNet models compared to their base versions."},{"finding":"Using a £3 stake yields a higher return on investment than using a £100 stake in the simulated 30‑day test.","kind":"performance","quote":"By executing the same predicted trades, it is evident that using a stake of £3.00 generates a higher return on investment (ROI) compared to using £100.00."},{"finding":"The model predicts 173 trades with expected positive PL and 98 with expected negative PL, resulting in 75 net positive‑PL trades in validation.","kind":"performance","quote":"the model predicts 173 trades with expected positive PL and 98 trades with expected negative PL, resulting in a total of 75 predicted trades with expected positive PL."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":48,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:24:36Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}