Paper: arXiv 2502.13722
Authors: Remi Genet
Abstract
Volume-Weighted Average Price (VWAP) is arguably the most prevalent benchmark for trade execution as it provides an unbiased standard for comparing performance across market participants. However, achieving VWAP is inherently challenging due to its dependence on two dynamic factors, volumes and prices. Traditional approaches typically focus on forecasting the market’s volume curve, an assumption that may hold true under steady conditions but becomes suboptimal in more volatile environments or markets such as cryptocurrency where prediction error margins are higher. In this study, I propose a deep learning framework that directly optimizes the VWAP execution objective by bypassing the intermediate step of volume curve prediction. Leveraging automatic differentiation and custom loss functions, my method calibrates order allocation to minimize VWAP slippage, thereby fully addressing the complexities of the execution problem. My results demonstrate that this direct optimization approach consistently achieves lower VWAP slippage compared to conventional methods, even when utilizing a naive linear model presented in arXiv:2410.21448. They validate the observation that strategies optimized for VWAP performance tend to diverge from accurate volume curve predictions and thus underscore the advantage of directly modeling the execution objective. This research contributes a more efficient and robust framework for VWAP execution in volatile markets, illustrating the potential of deep learning in complex financial systems where direct objective optimization is crucial. Although my empirical analysis focuses on cryptocurrency markets, the underlying principles of the framework are readily applicable to other asset classes such as equities.
Complexity vs Empirical Score
- Math Complexity: 6.0/10
- Empirical Rigor: 4.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper employs advanced mathematical modeling with deep learning and custom loss functions, indicating high complexity, but the empirical analysis is based on cryptocurrency markets and lacks the detailed backtest-ready data, statistical metrics, or code implementation typically required for high empirical rigor.
Research Flowchart
flowchart TD A["Research Goal<br>Optimize VWAP Execution<br>in Crypto Markets"] --> B["Core Methodology<br>Deep Learning Framework"] B --> C["Data & Inputs<br>Market Data &<br>Trade Parameters"] C --> D["Computational Process<br>Direct Objective Optimization<br>via Automatic Differentiation"] D --> E["Key Finding 1<br>Lower VWAP Slippage<br>vs. Conventional Methods"] D --> F["Key Finding 2<br>Strategies Diverge from<br>Volume Curve Prediction"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a deep learning framework that directly optimizes the VWAP execution objective, bypassing volume curve prediction.
Source sentence
I propose a deep learning framework that directly optimizes the VWAP execution objective by bypassing the intermediate step of volume curve prediction.
Assumes 2
- Market impact is negligible. market
Source sentence
One key assumption in the current framework is that market impact is negligible.
- The model operates in a static framework without dynamic updating. model
Source sentence
Another limitation is that the present model operates in a static framework without dynamic updating in real time.
Data
- Binance perpetual futures contracts market-prices public
Source sentence
Historical data from Binance perpetual futures contracts for Bitcoin (BTC), Ethereum (ETH), and Cardano (ADA) covering the period from contract inception until July 1, 2024
- Hourly data from five major cryptocurrencies market-prices public
Source sentence
The strategies are evaluated using hourly data from five major cryptocurrencies (BTC, ETH, BNB, ADA, and XRP) traded on Binance perpetual contracts covering the period from January 1, 2020, to July 1, 2024.
Costs
Not modeled
Source sentence
One key assumption in the current framework is that market impact is negligible.
Finds 5
- Directly optimizing the VWAP execution objective consistently achieves lower VWAP slippage than conventional volume‑curve methods. performance
Source sentence
My results demonstrate that this direct optimization approach consistently achieves lower VWAP slippage compared to conventional methods
- The performance gap between direct‑VWAP methods and volume‑prediction methods persists or widens with longer horizons. performance
Source sentence
the performance gap between methods directly optimized for VWAP execution and those focused on volume prediction tends to persist or even widen with longer horizons
- The proposed method shows lower slippage during periods of high volatility. performance
Source sentence
the proposed method generally exhibits lower slippage during periods of high volatility
- End‑loaded allocation strategies derived from direct VWAP optimization mitigate extreme slippage events. performance
Source sentence
the prevalence of negative values during volatile periods provides strong visual evidence that the end-loaded allocation strategy is effective in mitigating execution slippage
- Direct VWAP optimization yields a strategy that deviates from simply mimicking the historical volume curve. performance
Source sentence
These differences underscore that directly optimizing the VWAP objective (via absolute or quadratic loss) yields a strategy that deviates from simply mimicking the historical volume curve
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 33 related papers). Coverage is partial, so this is neither replication nor refutation.
3 papers in the archive
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (30)
- High frequency trading under scrutiny (2009)
- The Total Cost of Transactions on the NYSE (1988)
- The implementation shortfall: Paper versus reality (1988)
- VWAP strategies, Transaction Performance: The Changing Face of Trading Investment Guides Series (2002)
- Optimal execution of portfolio transactions (2000)
- Optimal slice of a VWAP trade (2002) — link
- Optimal VWAP Trading Strategy and Relative Volume (2007) — link
- Mean Variance Optimal VWAP Trading (2012) — link
and 22 more
- Price, trade size, and information in securities markets (1987) — link
- A Theory of the Interday Variations in Volume, Variance, and Trading Costs in Securities Markets (1990) — link
- The Price Variability-Volume Relationship on Speculative Markets (1983) — link
- The Relation between Price Changes and Trading Volume: A Survey (1987)
- Intra-day market activity (1999)
- Optimal VWAP trading under noisy conditions (2011)
- Optimal execution of a vwap order: A stochastic control approach (2013) — link
- Dynamic Programming and Trade Execution (2013)
- VWAP Execution and Guaranteed VWAP (2014)
- Are the Trading Volume and the Number of Trades Distributions Universal (2011)
- Permanent Market Impact Can Be Nonlinear (2013)
- Hierarchical Deep Reinforcement Learning for VWAP Strategy Optimization — → 2212.14670
- An Adaptive Dual-level Reinforcement Learning Approach for Optimal Trade Execution — → 2307.10649
- An Optimal Control Strategy for Execution of Large Stock Orders Using LSTMs — → 2301.09705
- TKAN: Temporal Kolmogorov-Arnold Networks (2024) — link
- SigKAN: Signature-Weighted Kolmogorov-Arnold Networks for Time Series (2024) — link
- A Temporal Kolmogorov-Arnold Transformer for Time Series Forecasting (2024) — link
- A Gated Residual Kolmogorov-Arnold Networks for Mixtures of Experts (2024)
- A Temporal Linear Network for Time Series Forecasting (2024)
- SigGate: Enhancing Recurrent Neural Networks with Signature-Based Gating Mechanisms (2025) — link
- Are Transformers Effective for Time Series Forecasting? (2022) — link
- TSMixer: An All-MLP Architecture for Time Series Forecasting (2023) — link
3 of 30 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.