{"paper_id":"2504.15908","title":"Learning the Spoofability of Limit Order Books With Interpretable Probabilistic Neural Networks","authors":["Timothée Fabre","Damien Challet"],"paper_date":"2025-04-22","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2504.15908v1","page_url":"https://thequant.space/flowcharts/learning-the-spoofability-of-limit-order-books-with-interpre/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a spoofing detection framework based on the probabilistic market manipulation gain of a spoofing agent, using a neural network to compute expected gain in real time.","quote":"Next, we propose a spoofing detection framework based on the probabilistic market manipulation gain of a spoofing agent and use the previously trained neural network to compute the expected gain.","summary":"This paper investigates real-time detection of spoofing activity in limit order books, focusing on cryptocurrency centralized exchanges. We first introduce novel order flow variables based on multi-scale Hawkes processes that account both for the size and placement distance from current best prices…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"No partial fill occurs; orders are either completely filled or not filled at all.","type":"execution","stated":"explicit","quote":"Assumption 1 No partial fill occurs. At the trading horizon, the orders are either completely filled or not filled at all."},{"assumption":"Execution of an order is triggered if the fair price crosses the price of the order.","type":"execution","stated":"explicit","quote":"Assumption 2 The execution of an order is triggered if the fair price crosses the price of this order."},{"assumption":"The agent must execute the target quantity within a preset trading time frame, otherwise a marketable order is sent incurring extra transaction costs.","type":"execution","stated":"explicit","quote":"Assumption 3 The agent has to execute the target quantity q within a preset trading time frame. Hence, if the bona fide order is not filled at the horizon, a marketable order is sent on the opposite side, crossing the spread, and incurring extra transaction costs."},{"assumption":"Orders are kept only if posted within ±20% of the mid price and size ≥50 USD.","type":"data","stated":"explicit","quote":"We set the distance threshold at ±20% of the mid price. Moreover, we remove orders whose size is smaller than 50 USD."},{"assumption":"The bona fide order is quoted at touch (δa = 0 bp) and the taker fee is 5 bps with no maker fee.","type":"model","stated":"explicit","quote":"The fees are set according to Coinbase fee policy and we select the highest level, i.e. a taker fee of ε−= 5 bps, and no maker fee, i.e. ε+ = 0 bp."},{"assumption":"The model uses Level‑3 market‑by‑event data from Coinbase spot pairs BTC‑USD and ETH‑USD for a single week.","type":"data","stated":"explicit","quote":"We focus on the two most traded spot pairs on Coinbase, BTC-USD and ETH-USD, and we select an arbitrary week, from December 1st to December 7th, 2022."}]},"datasets":{"status":"available","items":[{"name":"Level‑3 market‑by‑event data","provider":"SUN ZU Lab’s feed handlers","kind":"order-book","access":"proprietary","period":"2022-12-01 to 2022-12-07","universe":"BTC-USD and ETH-USD spot pairs on Coinbase","quote":"we focus on the two most traded spot pairs on Coinbase, BTC-USD and ETH-USD, and we select an arbitrary week, from December 1st to December 7th, 2022."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"l2-book"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R4a","candidates":{"front_matter":"n/a","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Taker fee of 5 bps, no maker fee, applied to the spoofing cost calculations.","card_quote":"The fees are set according to Coinbase fee policy and we select the highest level, i.e. a taker fee of ε−= 5 bps, and no maker fee, i.e. ε+ = 0 bp.","meaning":"the paper reports no trading or execution result; the cost the extractor found is a model parameter"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"31% of large orders could spoof the market during the test period.","kind":"statistical","quote":"Running this algorithm on all submitted limit orders in the period 2024-12-04 to 2024-12-07, we find that 31% of large orders could spoof the market."},{"finding":"Larger order size increases impact on the Sharpe ratio, while larger posting distance reduces impact.","kind":"statistical","quote":"We observe in Figs 6 and 7 that the larger the posted size, the greater the impact over the Sharpe ratio. The distance of placement seems to have an inverse role over the Sharpe ratio, as expected: the smaller the distance, the larger the impact."},{"finding":"Spoofing detection models that ignore posting distance are inadequate to describe the data.","kind":"theoretical","quote":"Our empirical analysis highlights the critical role of the posting distance of limit orders in the price formation process, showing that spoofing detection models that do not take the posting distance into account are inadequate to describe the data."},{"finding":"Orders flagged as suspicious (≈31% of large orders) have distinct behavioral patterns and cause larger price moves than normal orders.","kind":"statistical","quote":"Orders flagged as suspicious exhibit distinctive behavioral patterns and have a demonstrably different impact on price formation."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":19,"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":"2026-11-14T17:40:15Z"},"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/"}