{"paper_id":"2312.16637","title":"Price predictability at ultra-high frequency: Entropy-based randomness test","authors":["Andrey Shternshis","Stefano Marmi"],"paper_date":"2023-12-27","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2312.16637v3","page_url":"https://thequant.space/flowcharts/price-predictability-at-ultra-high-frequency-entropy-based/","scores":{"empirical_rigor":4,"math_complexity":7.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a statistical test for randomness of a symbolic sequence that can be applied to ultra‑high‑frequency price data.","quote":"First, we propose the test for randomness of a symbolic sequence. It can be applied with varying numbers of distinct symbols, presuming the sequence length is sufficiently large.","summary":"We use the statistical properties of Shannon entropy estimator and Kullback-Leibler divergence to study the predictability of ultra-high frequency financial data. We develop a statistical test for the predictability of a sequence based on empirical frequencies. We show that the degree of randomness…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The price direction series is a realization of a stationary random process with symbols from a finite alphabet.","type":"statistical","stated":"explicit","quote":"We introduce a statistical test designed to evaluate the predictability of a sequence. The input for a statistical test is a realization X of a stationary random process X with symbols from a finite alphabet A = {0, 1, · · · , s −1}."},{"assumption":"Under the null hypothesis, symbols occur independently of the past history.","type":"model","stated":"explicit","quote":"H0: The occurrence of a new symbol in sequence X is independent of the sequence’s prior history."},{"assumption":"The sequence length is sufficiently large for the asymptotic chi‑square distribution to hold.","type":"statistical","stated":"explicit","quote":"It can be applied with varying numbers of distinct symbols, presuming the sequence length is sufficiently large."},{"assumption":"Probabilities of symbols need not be equal; the test remains valid.","type":"model","stated":"explicit","quote":"We show that the proposed test statistics is valid even if the probabilities of appearing symbols are not equal."},{"assumption":"Transactions occurring at the same nanosecond are aggregated to a single price to avoid spurious predictability.","type":"execution","stated":"explicit","quote":"To mitigate the impact of such events on the analysis, we aggregate volumes and consider the final available price for each nanosecond showcasing trading activity."},{"assumption":"Predictability does not generate arbitrage because it is offset by transaction‑cost fluctuations.","type":"market","stated":"explicit","quote":"According to [16, 17], such type of predictability does not lead to arbitrage opportunities because it is compensated by fluctuations in transaction costs and liquidity"}]},"datasets":{"status":"available","items":[{"name":"LOBSTER limit order book data","provider":"LOBSTER (www.lobsterdata.com)","kind":"order-book","access":"public","period":"01.08.2022 to 21.11.2022","universe":"Apple, Microsoft, Tesla, Intel, Eli Lilly, Snap, Ford, Carnival, SPY (80 trading days)","quote":"... The timeframe under consideration spans from 01.08.2022 to 21.11.2022, encompassing a total of 80 trading days. Table 1 presents the tickers","known_id":"lobster"}],"known":[{"id":"lobster","name":"LOBSTER","provider":"LOBSTER","kind":"order-book","access":"commercial","quote":"We explore limit order book data [35] downloaded from LOBSTER (www.lobsterdata.com).","mentions":1,"used_in":1}],"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":"partial","evidence_card_quoted":true},"card_detail":"Predictability is said to be offset by fluctuations in transaction costs and liquidity, but no explicit cost model is incorporated into the headline results.","card_quote":"According to [16, 17], such type of predictability does not lead to arbitrage opportunities because it is compensated by fluctuations in transaction costs and liquidity","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":"The degree of randomness of price direction increases as the aggregation level in transaction time grows.","kind":"statistical","quote":"We show that the degree of randomness grows with the increase of aggregation level in transaction time."},{"finding":"Predictable days are associated with unusually high trading volumes and a larger number of price changes.","kind":"statistical","quote":"We also find that predictable days are usually characterized by high trading activity, i.e., days with unusually high trading volumes and the number of price changes."},{"finding":"Excluding simultaneous nanosecond‑level transactions reduces the fraction of predictable days.","kind":"statistical","quote":"The exclusion of simultaneous transactions leads to a reduction in the predictability of price return signs."},{"finding":"For TSLA, predictability decreases sharply with higher aggregation; only a few days remain predictable at aggregation a = 65.","kind":"predictive-accuracy","quote":"Increasing the aggregation level by 5, we observe that the probability of repeating a symbol is significantly high even when a = 60. However, by further increasing the aggregation level to a = 65, only 4 predictable days remained"},{"finding":"Predictable intervals generally do not occur consecutively, except for SNAP where several predictable intervals appear in a row.","kind":"statistical","quote":"However, considering all transactions of the SNAP stock (Snap Inc.), we are able to detect several predictable time intervals going in a row."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[{"paper":{"paper_id":"ssrn-2440866","title":"Market Intraday Momentum","paper_date":"2014-01-01","authors":["Lei Gao","Yufeng Han","Sophia Zhengzi Li","Guofu Zhou"],"empirical_rigor":8,"page_url":"https://thequant.space/flowcharts/market-intraday-momentum/"},"question":"Does price predictability increase on days with high trading volume and volatility?","this_paper_finding":"Predictable days are associated with unusually high trading volumes and a larger number of price changes.","other_paper_finding":"Predictability is stronger on more volatile days, higher volume days, recession days, and major macro‑economic news release days.","other_paper_quote":"The predictability, both statistically and economically signiﬁcant, is stronger on more volatile days, on higher volume days, on recession days, and on major macroeconomic news release days.","note":"Both findings report higher predictability on unusually high‑volume days; B adds that volatility and macro news also strengthen the effect, confirming A's volume link.","confidence":0.795}],"contradicts":[],"qualifies":[],"compared_with":33,"citations":{"cited_by_in_corpus":1,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"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":"2026-11-14T18:17:18Z"},"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/"}