---
title: "When Does Order Flow Matter? State-Dependent L2 Liquidity-State Transitions in Crypto Futures"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.09230v1"
permalink: "https://feed.astralor.com/items/itm_c3e8841b1e173c38f883428b"
published: "2026-07-10T09:21:00.000Z"
score: 79
channel: "finance"
tags: ["Binance","BTCUSDT","ETHUSDT","L2 liquidity state","market microstructure","order flow"]
---

# When Does Order Flow Matter? State-Dependent L2 Liquidity-State Transitions in Crypto Futures

## AI commentary

> This paper pinpoints a common pitfall in event-driven trading: don't rush to throw order flow into your model—check the liquidity state first. For high-frequency quant teams and market makers, the evaluation protocol serves as a litmus test for whether more complex models (e.g., LLMs) truly add incremental value.

## At a glance

This study uses top-20 L2 order book data, trade-flow records, and macro-event windows from Binance BTCUSDT and ETHUSDT futures (2023-2026) to define a supervised discrete L2 liquidity-state transition task. The pre-event L2 liquidity state is the strongest predictor of post-event liquidity regimes; a shallow nonlinear model yields additional gains. Order flow adds value only on top of the L2 state model, not as a replacement, and the value is not robust across symbols. The paper proposes a state-first design principle and an evaluation protocol for microstructure models.

## Key points

- Pre-event L2 liquidity state is the strongest predictor of post-event liquidity regimes, with a strong coarse baseline.
- A shallow nonlinear L2 model yields consistent gains over the state baseline, outperforming interpretable linear logit models.
- Order flow adds predictive value only when layered on top of the L2 state model, and the benefit is inconsistent between BTC and ETH.
- The paper provides a liquidity-state transition baseline and evaluation protocol for validating incremental value of complex models (e.g., RL, LLM).

## Full text

Building 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 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 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 event timing but not the event's label content, so pre-event state competes against an uninformed within-window baseline, not against the event type. Order flow adds further value only when layered on top of the L2 state model, not as a replacement. This value is not 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 exceed before their added value is credited.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_c3e8841b1e173c38f883428b