Paper: arXiv 2302.13850

Authors: Fazl Barez, Paul Bilokon, Arthur Gervais, Nikita Lisitsyn

Abstract

This paper explores the novel deep learning Transformers architectures for high-frequency Bitcoin-USDT log-return forecasting and compares them to the traditional Long Short-Term Memory models. A hybrid Transformer model, called \textbf{HFformer}, is then introduced for time series forecasting which incorporates a Transformer encoder, linear decoder, spiking activations, and quantile loss function, and does not use position encoding. Furthermore, possible high-frequency trading strategies for use with the HFformer model are discussed, including trade sizing, trading signal aggregation, and minimal trading threshold. Ultimately, the performance of the HFformer and Long Short-Term Memory models are assessed and results indicate that the HFformer achieves a higher cumulative PnL than the LSTM when trading with multiple signals during backtesting.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces a novel hybrid Transformer architecture (HFformer) specifically modified for high-frequency crypto trading without positional encodings and using quantile loss. Empirical results are supported by signal aggregation and execution backtests on Bitcoin order flow data, although full reproducibility details are moderate.

Research Flowchart

  flowchart TD
    A[Goal: Enhance High-Frequency Bitcoin Trading via Transformers] --> B[Data: High-Frequency Bitcoin-USDT Log-Returns]
    B --> C[Model Comparison: Standard LSTM vs. Transformer Architectures]
    C --> D[Develop HFformer: Encoder, Linear Decoder, Spiking Activation & Quantile Loss]
    D --> E[Formulate Trading Strategies: Signal Aggregation, Sizing & Thresholds]
    E --> F[Backtesting & Assessment]
    F --> G[Outcome: HFformer Achieves Higher Cumulative PnL than LSTM]