Paper: arXiv 2505.05595

Authors: Wenhao Guo, Yuda Wang, Zeqiao Huang, Changjiang Zhang, Shumin ma

Abstract

In the complex landscape of traditional futures trading, where vast data and variables like real-time Limit Order Books (LOB) complicate price predictions, we introduce the FutureQuant Transformer model, leveraging attention mechanisms to navigate these challenges. Unlike conventional models focused on point predictions, the FutureQuant model excels in forecasting the range and volatility of future prices, thus offering richer insights for trading strategies. Its ability to parse and learn from intricate market patterns allows for enhanced decision-making, significantly improving risk management and achieving a notable average gain of 0.1193% per 30-minute trade over state-of-the-art models with a simple algorithm using factors such as RSI, ATR, and Bollinger Bands. This innovation marks a substantial leap forward in predictive analytics within the volatile domain of futures trading.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 5.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces advanced deep learning models like Transformers and quantile regression with mathematical formulations, but also reports specific backtested results (0.1193% gain) on futures data using real factors, indicating moderate empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal: Predict<br>Futures Price Range & Volatility"] --> B["Key Inputs: LOB Data &<br>Technical Indicators RSI/ATR/Bollinger"]
  B --> C["Methodology: FutureQuant Transformer<br>using Attention Mechanisms"]
  C --> D["Computational Process:<br>Learn intricate market patterns"]
  D --> E["Key Outcomes: Range Forecasting<br>Improved Risk Management"]
  E --> F["Final Result: 0.1193% Avg Gain<br>per 30-min trade vs SOTA"]

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 the FutureQuant Transformer model for futures price quantile forecasting and a trading strategy that uses indicators such as RSI, Bollinger Bands, and ATR.

Source sentence
we introduce the FutureQuant Transformer model, leveraging attention mechanisms to navigate these challenges.

Assumes 2

  • Features are normalized using min‑max scaling before model training. data
    Source sentence
    This paper employs min‑max normalization to rescale the data features into a smaller, uniform range.
  • The trading strategy operates on 30‑minute intervals using predefined indicator thresholds. execution
    Source sentence
    The essence of this trading strategy is to utilize a threshold and a series of indicators such as Bollinger Band, RSI, and ATR.

Data

Backtest sample: futures · intraday

Costs

Not modeled A trading or execution result is reported without costs.

Finds 3

Reported performance (FutureQuant Transformer): annual return 264.6% · total return 264.6% · max drawdown 36.0%

  • The FutureQuant Transformer achieves an average gain of 0.1193% per 30‑minute trade over state‑of‑the‑art models. performance
    Source sentence
    significantly improving risk management and achieving a notable average gain of 0.1193% per 30‑minute trade over state‑of‑the‑art models
  • The model attains a Prediction Interval Coverage Probability (PICP) of 0.962 and a Coverage Width‑based Criterion (CWC) of 3.175. predictive-accuracy
    Source sentence
    FutureQuant Transformer model exhibits outstanding performance with a Prediction Interval Coverage Probability (PICP) of 0.962 and a Coverage Width‑based Criterion (CWC) of 3.175.
  • Backtesting shows a 30‑minute cumulative return of 0.175%, a 1‑month return of 21%, and a 1‑year return of 264.6% with volatility of 1.653 and maximum drawdown of -35.997%. performance
    Source sentence
    Cumulative Return (30mins) 0.175% Cumulative Return (1-month) 21% Cumulative Return (1-year) 264.6% Volatility 1.653 Maximum Drawdown -35.997%

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 71 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (10)

  1. Robust Testing for Bollinger Band, Moving Average and Relative Strength Index (2022) — link
  2. A hybrid method of exponential smoothing and recurrent neural networks for time series forecasting (2020) — link
  3. The limits of distribution-free conditional predictive inference (2019) — link
  4. Interval prediction method based on Long-Short Term Memory networks for system integrated of hydro, wind and solar power (2019) — link
  5. Wind Power Forecasting Using Multi-Objective Evolutionary Algorithms for Wavelet Neural Network-Optimized Prediction Intervals (2018) — link
  6. Forecasting stock market volatility: Do realized skewness and kurtosis help? (2017) — link
  7. Attention is All you Need (2017) — link
  8. A Quantile Regression Neural Network Approach to Estimating the Conditional Density of Multiperiod Returns (2000) — link
and 2 more
  1. Skewness Persistence in Common Stock Returns (1986) — link
  2. Regression Quantiles (2007) — link

0 of 10 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.