{"paper_id":"2503.01629","title":"A New Traders' Game? -- Empirical Analysis of Response Functions in a Historical Perspective","authors":["Cedric Schuhmann","Benjamin Köhler","Anton J. Heckens","Thomas Guhr"],"paper_date":"2025-03-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2503.01629v2","page_url":"https://thequant.space/flowcharts/a-new-traders-game----empirical-analysis-of-response-funct/","scores":{"empirical_rigor":7,"math_complexity":6.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper empirically extends the analysis of self‑ and cross‑response functions and trade‑sign correlators from the crisis year 2008 to the years 2007, 2014 and 2021, providing averaged response functions and investigating their non‑stationarity over time.","quote":"Further details have been explored, such as the market collective response [42] and the role of asymmetric information in market impacts [43] Here, we extend the analysis of Refs. [30, 31, 32] for the year 2008 by studying the years 2007, 2014 and 2021. In doing so, we rely on the well–established and theoretically explored framework of self– and cross–response functions in returns and trade signs [30, 31, 32, 33,","summary":"Traders on financial markets generate non-Markovian effects in various ways, particularly through their competition with one another which can be interpreted as a game between different (types of) traders. To quantify the market mechanisms, we empirically analyze self-response functions for pairs…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The finest available time resolution for 2007 and 2008 is one second, and non‑Markovian effects are considered relevant on time scales much larger than one second.","type":"data","stated":"explicit","quote":"For 2007 and 2008, one second is the finest available time grid resolution possible [30, 31, 32]. For the years 2014 and 2021 considerably higher temporal resolutions up to nanoseconds for 2021 are available, but anticipating later findings, we point out that the non–Markovian effects are relevant on time scales much larger than one second.","support":"full"},{"assumption":"Trade sign zeroes are excluded from some analyses, which affects the sign of the trade‑sign series and the resulting correlations.","type":"statistical","stated":"explicit","quote":"Excluding the trade sign zero makes the analysis more similar to studying markets on a trading time scale. Only when zero trade signs are not included, Eq. (6) is a time–lagged correlation as the time series of trade signs then have unit variance.","support":"full"},{"assumption":"Midpoint prices are used as a proxy for order‑book dynamics instead of trade prices.","type":"market","stated":"explicit","quote":"We use the midpoint price mi(t) = ai(t) + bi(t) 2 as average of the best as ai(t) and best bid bi(t) for stock i in the order book where t runs in one–second steps. Here, we use the midpoint prices instead of the trade prices as we are also interested in the dynamics of the order book near the midpoint between consecutive trades.","support":"full"},{"assumption":"Averaging across all stocks and stock pairs yields representative market‑wide response functions.","type":"statistical","stated":"explicit","quote":"4.1 and 4.3, we work out the responses and trade sign correlations for the whole market by averaging across all individual stocks and stock pairs, respectively. Market Response Functions We obtain a better statistical significance for the estimation of market self– and cross– impact by averaging across all self–responses R s(τ) = ⟨⟨Rii(τ)⟩i⟩i (7) and across all cross-responses R c(τ) = ⟨⟨Rij(τ)⟩j⟩i , for i̸ = j (8)","support":"full"},{"assumption":"The set of 99 stocks selected for 2008 can be used as a common universe for the other years after replacing delisted stocks with high‑liquidity equivalents.","type":"data","stated":"explicit","quote":"To compare different years, our focus is on selecting the same 99 stocks as in 2008 for all years. We replace the stocks delisted in a respective economic sector by high–liquid stocks from the same economic sector, i.e by stocks with a high trading value.","support":"full"},{"assumption":"The trade sign self‑correlator and cross‑correlator can be approximated by power‑law functions over the examined time lags.","type":"model","stated":"explicit","quote":"The power law trend is fitted very well by the function Θ κ(τ) = ϑ κ (1 + (τ/τ κ)2)γ κ/2 , κ = s, c . (11) We list the fit parameters ϑ κ, τ κ and γ κ, κ = s, c, in Tab. 2.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Daily TAQ (Trade and Quote) of the New York Stock Exchange","provider":"NYSE","kind":"order-book","access":"public","period":"2007, 2008, 2014, 2021","universe":"most US stocks; 99 stocks selected for consistency across years","quote":"Data set We use the Daily TAQ (Trade and Quote) of the New York stock exchange (NYSE) [44] for the years 2007, 2008, 2014 and 2021. This data set records the trading activity of most stocks on the US stock markets, i.e. the prices are listed in the trade file and best bid and ask prices in the quote file.","support":"full","known_id":"taq"},{"name":"NASDAQ high‑liquidity stock sample (10 stocks per sector)","provider":"NASDAQ","kind":"market-prices","access":"public","period":"2008","universe":"10 stocks with high liquidity from each of the ten GICS sectors","quote":"We also analyze the stocks as in Refs. [30, 31, 32] from the NASDAQ stock exchange for 2008, i.e. ten stocks with high liquidity from each of the ten economic sectors according to the Global Industry Classification Standard (GICS) see Tab. 1. In the case of the telecommu-nications sector, the number of available stocks is limited to nine.","support":"full"}],"known":[{"id":"taq","name":"NYSE TAQ","provider":"NYSE","kind":"market-prices","access":"commercial","quote":"Data set We use the Daily TAQ (Trade and Quote) of the New York stock exchange (NYSE) [44] for the years 2007, 2008, 2014 and 2021.","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"tick"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"Significant variations in self‑ and cross‑response functions over time reveal changes in the traders’ game, with larger responses during the 2008 crisis.","kind":"statistical","quote":"We find significant variations over time revealing changes in the traders' game. First, the responses for the crisis year of 2008 are larger than for the other years.","support":"full"},{"finding":"Self‑responses for 2014 and 2021 increase for larger time lags and remain at higher levels than in 2007, indicating a lasting effect of the crisis.","kind":"statistical","quote":"The self–responses of 2014 and 2021 indicate that self–responses increasing for larger time lags are quite stable over time. The higher values in the self–responses for 2014 and 2021 compared to 2007 indicate that the crisis has had a lasting influence on the trader’s actions.","support":"full"},{"finding":"Both market self‑responses and cross‑responses become less concave in 2014 and 2021, suggesting a shift possibly linked to algorithmic trading.","kind":"statistical","quote":"Second, both market self–responses and cross–responses are strikingly less concave. The shape of the self– and cross–responses is no longer concave for 2014 and 2021 hinting at different causes, i.e. the algorithmic trading might have changed the responses.","support":"full"},{"finding":"Trade‑sign self‑correlator exponents γ_s are below 0.8 in 2007 (long‑term memory) and above one in 2021 (short‑term memory), with a distinct regime change around 100–1000 s.","kind":"statistical","quote":"The results for γ s and γ c are interesting as values above one indicate short–term memory and values smaller than one long–term memory [46]. For 2007, we notice values smaller than 0.8, for the other years they are larger except for the year 2014 when excluding trade sign zero.","support":"partial"},{"finding":"Passive and active cross‑response functions show different time‑scale behavior in 2007 and 2008, but not in 2014 and 2021.","kind":"statistical","quote":"The time scales of the passive and active response functions are different for 2007 and 2008, while we do not observe such behavior for 2014 and 2021.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":21,"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":2,"extracted_at":"2026-11-14T19:52:11Z"},"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/"}