The Market Maker's Dilemma: Navigating the Fill Probability vs. Post-Fill Returns Trade-Off

Jakob Albers, Mihai Cucuringu, Sam Howison, Alexander Y. ShestopaloffarXiv 2025 · DOI · arXiv

Abstract

Using data from a live trading experiment on the Binance Bitcoin perpetual, we examine the effects of (i) basic order book mechanics and (ii) the persistence of price changes from immediate to short timescales, revealing the interplay between returns, queue sizes, and orders' queue positions. We document a fundamental trade-off: a negative correlation between maker fill likelihood and post-fill returns. This dictates that viable maker strategies often require a contrarian approach, counter-trading the prevailing order book imbalance. These dynamics render commonly-cited strategies highly unprofitable, leading us to model 'Reversals': situations where a contrarian maker strategy at the touch proves effective.

1. Introduction

In the Limit Order Book (LOB) environment, market participants face a complex trilemma: the simultaneous pursuit of high fill probabilities, optimal execution prices, and positive post-fill returns. Traditional academic literature often simplifies these dynamics, assuming stochastic fill rates that ignore the strategic nature of liquidity provision. In reality, market makers face a "fundamental trade-off" where the ease of order execution is frequently a precursor to adverse price movements.

This paper addresses the Market Maker’s Dilemma: providing liquidity essentially offers a free option to the market. When an order is filled, it is often due to adverse selection—the "taker" possesses information or momentum that the "maker" lacks. We quantify this trade-off using live experimental data to avoid the signal bias inherent in historical simulations.

2. Methodology and Data

2.1 Live Experimental Trading

To capture the true dynamics of order execution, we conducted a live-trading experiment on the Binance USDT-margined Bitcoin perpetual, the world's most liquid cryptocurrency market. Our algorithm submitted hundreds of thousands of minimum-sized maker orders, maintaining a continuous presence at the "touch" (the best bid and ask). This approach provides a signal-bias-free baseline, as orders were placed regardless of prevailing market signals.

2.2 Queue Position Analysis

Central to our analysis is the tracking of an order's relative position within the queue. Using Binance’s public trade feed, we calculate the exact volume ahead of and behind our order at the moment of execution.

Visualization of a maker order's position within the limit order book queue.

As illustrated in the figure above, the queue position ($QP_t(o)$) is defined by the liquidity ahead ($LA_t(o)$) and behind ($LB_t(o)$) the specific maker order $o$. This metric is a primary determinant of both the likelihood of a fill and the subsequent price drift.

2.3 Dataset Characteristics

The study utilizes two primary datasets:

  1. Primary Dataset: 232,897 maker orders collected over one week in February 2024.
  2. Validation Dataset: 172,800 orders collected in August 2024 to confirm the temporal stability of the observed dynamics.

3. The Fundamental Trade-Off

3.1 The Unprofitability Principle

We propose a heuristic for high-frequency trading: Ease of prediction × ease of exploitation < constant. If a market signal is easily identifiable, the associated profits are typically eroded by exchange fees or adverse selection.

3.2 Fill-Return Correlation

Our data reveals a stark negative correlation between fill likelihood and post-fill returns. Orders that are "easy" to fill—characterized by a small near-side queue and a large opposite-side queue—almost invariably result in negative price drift. Conversely, being at the front of a large queue yields the best average returns but carries the lowest fill probability.

3.3 The Imbalance Fallacy

While academic models often suggest "trading with the imbalance" (e.g., buying when buy-side pressure is high), we find this strategy is highly unprofitable for makers. Makers only receive fills in these scenarios when the imbalance fails to move the price, leading to systematic losses.

4. Empirical Results

  • Naive Strategy Performance: A baseline strategy of quoting at the touch without filtering for market signals proved disastrous, losing 60% of its value in 3.5 days with a Sharpe ratio of -109.0.
  • Predictability of Fills: Fill probability is highly sensitive to queue size, exceeding 90% when the near-side queue is minimal but dropping below 30% for deep queues.
  • Reversal Modeling: We framed the identification of profitable trades as a classification task, using Logistic Regression and Random Forests to predict "reversals." These are rare moments where contrarian liquidity provision (countering the imbalance) is profitable. By selectively trading these reversals, we achieved a Sharpe ratio of 11.97 at specific probability thresholds.
  • The Fee Barrier: Taker fees (approx. 1.5 bp) represent a significant hurdle; many strategies that appear profitable on paper become unviable once exchange costs are factored in.

5. Discussion and Limitations

While our findings provide a robust framework for understanding maker dynamics, several limitations remain. The experiment utilized minimum order sizes; scaling these strategies would introduce market impact, potentially degrading returns. Furthermore, the "race to the top" of the queue in live HFT involves hardware-level latency battles not fully captured here. Future research utilizing Level 3 (L3) data—tracking individual orders rather than aggregated price levels—could provide even more granular insights into queue dynamics and the mechanics of adverse selection.

Summary updated

Abstract

Using data from a live trading experiment on the Binance Bitcoin perpetual, we examine the effects of (i) basic order book mechanics and (ii) the persistence of price changes from immediate to short timescales, revealing the interplay between returns, queue sizes, and orders' queue positions. We document a fundamental trade-off: a negative correlation between maker fill likelihood and post-fill returns. This dictates that viable maker strategies often require a contrarian approach, counter-trading the prevailing order book imbalance. These dynamics render commonly-cited strategies highly unprofitable, leading us to model `Reversals': situations where a contrarian maker strategy at the touch proves effective.
BibTeX
@misc{albers2025market,
  author = {Jakob Albers and Mihai Cucuringu and Sam Howison and Alexander Y. Shestopaloff},
  title = {{The Market Maker's Dilemma: Navigating the Fill Probability vs. Post-Fill Returns Trade-Off}},
  year = {2025},
  eprint = {2502.18625},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2502.18625},
}
Figure