[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82263-en":3,"doc-seo-82263-105":28,"detail-sidebar-cat-0-en-105":94},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":13,"seo_description":14,"update_tm":26,"read_time":27},82263,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","When Does Order Flow Matter? State-Dependent L2 Liquidity-State Transitions in Crypto Futures","Building event-conditioned market models requires separating macro-event labels from persistent microstructure state. This study examines Binance BTCUSDT and ETHUSDT futures (2023–2026) using top-20 L2 order book data, trade-flow records, and macro-event windows. A supervised L2 liquidity-state transition task is evaluated with rolling monthly out-of-sample folds, event-cluster validation, and blocked permutation tests. The pre-event L2 liquidity state strongly predicts post-event regimes; logit models over continuous L2 features do not outperform it, while a shallow nonlinear L2 model adds comparable incremental gain. Event calendars only define windows and matched controls; order-flow overlays help only when layered on state, not replacing it. Effects are not cross-symbol robust: ETH gains across regimes, BTC gains only briefly, guiding state-first modeling principles.","When Does Order Flow Matter? State-Dependent L2 Liquidity-State Transitions in Crypto Futures  \nJoohyoung Jeon  \nKorea University  \nSeoul, South Korea  \n[joohyoung@korea.ac.kr](joohyoung@korea.ac.kr)  \narXiv :2607 .09230v1 [ q-fin .TR] 10 Jul 2026  \nAbstract  \nBuilding event-conditioned market models requires separating macro-event labels from persistent microstructure state. We study this distinction in Binance BTCUSDT and ETHUSDT futures from 2023–2026, combining top-20 L2 order book data, trade-flow records, and macro-event windows. We define a supervised discrete L2 liquidity-state transition task, distinct from latent-regime detection and from price-direction prediction, and evaluate models in rolling monthly out-of-sample folds with event-clustered validation and blocked permutation tests, admitting each feature layer only if it improves on the layer below it on the same panel. Within these event windows, the first-order predictive signal is the pre-event L2 liquidity state: a coarse pre-event state baseline strongly predicts post-event liquidity regimes, interpretable logit models over continuous L2 features fail to improve on it, and a shallow nonlinear L2 model adds a robust further predictive gain of comparable size to the state baseline’s own. The macro-event calendar enters only by locating the windows and supplying matched non-event controls; we use the event timing but not the event’s label content as a competing predictor, so the comparison is between pre-event state and an uninformed within-window baseline, not against the event type. Order flow provides further incremental value only when layered on top of the L2 state model, not as a replacement. This value isnot robustly cross-symbol: for ETH it is present across calm, mixed, and stressed regimes and largest under stressed pre-event liquidity, whereas BTC shows only isolated five-minute passes and no regime that clears at both horizons. These findings motivate a state-first design principle for market microstructure models. We provide a liquidity-state transition baseline and evaluation protocol that reinforcement-learning, execution-policy, or LLM-based context layers should be required to exceed before their added value is credited.  \nCCS Concepts  \n• Computing methodologies → Supervised learning by classification; • Applied computing → Economics; • Theory of computation → Sequential decision making.  \nKeywords  \nmarket microstructure, limit order book, liquidity state, order flow, crypto futures, out-of-sample evaluation  \nTable 1: Model comparison. The 1m and 5m columns are out-of-sample joint improvements (Section 4); a positive value means the model improves on its baseline. Upper panel: staged baselines on the held-out event windows, the first row coarse state vs. marginal, later rows vs. the coarse state. Lower panel: order-flow overlay (augmented vs. L2-only) against the flow-shuffle null.  \n\n| Comparison | 1m | 5m | Null 95th pct. / interval |\n| --- | --- | --- | --- |\n| Baselines vs. coarse pre-event | state |  |  |\n| Coarse state (vs. marginal) | +0 .034 | +0 .045 | — |\n| Multinomial logit | −0 .048 | −0 .034 | below zero |\n| Ordered logit | −0 .052 | −0 .047 | below zero |\n| Nonlinear L2-shape | +0 .044 | +0 .060 | [ .041, . 046]; [ .057, . 062] |\n| Order-flow overlay vs. flow-shuffle null |  |  |  |\n| Pooled\u003Cbr>BTC (not established) ETH (clears null) | +0 .010 +0 .010 +0 .004 / +0 .003 +0 .001 +0 .003 +0 .002 / +0 .002 +0 .020 +0 .016 +0 .006 / +0 .003 |  |  |\n\nThe two proper scores agree in sign in every cell; for the baseline panel the ΔNLL and ΔBrier components are coarse state +0 . 046/+0 .061 and +0 . 022/+0 . 028, multinomial logit −0 . 063/−0 .044 and −0 . 033/−0 . 024, ordered logit −0 . 069/−0 .067 and −0 . 034/−0 . 026, and nonlinear L2-shape +0 . 061/+0 .083 and +0 . 028/+0 .037 at the two horizons. Both logit intervals lie entirely below zero. Throughout, point estimates and ΔNLL/ΔBrier components are on the held-out event windows w","cbCaibkkqvydaHKj","https://ap.wps.com/l/cbCaibkkqvydaHKj","pdf",471590,1,"English","en",105,"# Abstract\n# 1 Introduction\n## Event-conditioned modeling motivation\n## Prediction target vs price forecasting","[{\"question\":\"What is the document’s main prediction task?\",\"answer\":\"It formulates a supervised discrete L2 liquidity-state transition task that predicts the post-event liquidity regime (calm, mixed, or stressed) rather than price movement.\"},{\"question\":\"How does the study treat event information?\",\"answer\":\"Event timing defines and matches event windows, while the event label content is not used as a competing predictor; non-event controls are matched within the window framework.\"},{\"question\":\"What determines whether order flow adds value?\",\"answer\":\"Order flow provides incremental predictive value only when added on top of the L2 liquidity-state model, not as a replacement for it.\"},{\"question\":\"Why are results described as not robust across symbols?\",\"answer\":\"For ETH, the order-flow increment appears across multiple regimes and is largest under stressed pre-event liquidity, while BTC shows only isolated short-horizon effects and lacks a regime that clears at both horizons.\"}]",1784179236,20,{"code":4,"msg":29,"data":30},"ok",{"site_id":23,"language":22,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":89,"head_meta":91,"extra_data":93,"updated_unix":26},"when-does-order-flow-matter-state-dependent-l2-liquidity-state-transitions-in-crypto-futures","",{"@graph":34,"@context":88},[35,52,67],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/research-report/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/when-does-order-flow-matter-state-dependent-l2-liquidity-state-transitions-in-crypto-futures/82263/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":22,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80,84],{"name":71,"@type":72,"acceptedAnswer":73},"What is the document’s main prediction task?","Question",{"text":74,"@type":75},"It formulates a supervised discrete L2 liquidity-state transition task that predicts the post-event liquidity regime (calm, mixed, or stressed) rather than price movement.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does the study treat event information?",{"text":79,"@type":75},"Event timing defines and matches event windows, while the event label content is not used as a competing predictor; non-event controls are matched within the window framework.",{"name":81,"@type":72,"acceptedAnswer":82},"What determines whether order flow adds value?",{"text":83,"@type":75},"Order flow provides incremental predictive value only when added on top of the L2 liquidity-state model, not as a replacement for it.",{"name":85,"@type":72,"acceptedAnswer":86},"Why are results described as not robust across symbols?",{"text":87,"@type":75},"For ETH, the order-flow increment appears across multiple regimes and is largest under stressed pre-event liquidity, while BTC shows only isolated short-horizon effects and lacks a regime that clears at both horizons.","https://schema.org",{"og:url":50,"og:type":90,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":92,"canonical":50},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":95},[96,100,104,108,113,118,123,126,130,133,137],{"id":20,"doc_module":4,"doc_module_name":44,"category_name":97,"show_sort_weight":98,"slug":99},"Story & Novel",90,"story-novel",{"id":45,"doc_module":4,"doc_module_name":44,"category_name":101,"show_sort_weight":102,"slug":103},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":44,"category_name":105,"show_sort_weight":106,"slug":107},"Exam",70,"exam",{"id":109,"doc_module":4,"doc_module_name":44,"category_name":110,"show_sort_weight":111,"slug":112},5,"Comic",60,"comic",{"id":114,"doc_module":4,"doc_module_name":44,"category_name":115,"show_sort_weight":116,"slug":117},6,"Technology",50,"technology",{"id":119,"doc_module":4,"doc_module_name":44,"category_name":120,"show_sort_weight":121,"slug":122},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":44,"category_name":12,"show_sort_weight":124,"slug":125},30,"research-report",{"id":127,"doc_module":4,"doc_module_name":44,"category_name":128,"show_sort_weight":27,"slug":129},9,"Religion & Spirituality","religion-spirituality",{"id":27,"doc_module":4,"doc_module_name":44,"category_name":131,"show_sort_weight":27,"slug":132},"World Cup","world-cup",{"id":134,"doc_module":4,"doc_module_name":44,"category_name":135,"show_sort_weight":134,"slug":136},10,"Lifestyle","lifestyle",{"id":138,"doc_module":4,"doc_module_name":44,"category_name":139,"show_sort_weight":109,"slug":140},19,"General","general"]