Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Finance-Grounded Optimization For Algorithmic Trading

Kasymkhan Khubiev, Mikhail Semenov, Irina Podlipnova, Dinara Khubieva

TL;DR

This work tackles the misalignment between standard deep learning objectives and financial performance in algorithmic trading by proposing finance-grounded loss functions derived from $Sharpe$, $PnL$, and $MDD$, coupled with turnover regularization to constrain trading activity. It demonstrates that optimizing these economics-oriented objectives on Binance crypto data (61 assets across multi-frequency horizons) yields superior risk-adjusted returns compared with traditional regression losses, with LogMDDLoss and ModSharpeLoss often providing the strongest performance and stability. The findings underscore the value of directly optimizing economic performance metrics for trading and portfolio management, offering a path toward more robust, interpretable AI-driven strategies in finance.

Abstract

Deep Learning is evolving fast and integrates into various domains. Finance is a challenging field for deep learning, especially in the case of interpretable artificial intelligence (AI). Although classical approaches perform very well with natural language processing, computer vision, and forecasting, they are not perfect for the financial world, in which specialists use different metrics to evaluate model performance. We first introduce financially grounded loss functions derived from key quantitative finance metrics, including the Sharpe ratio, Profit-and-Loss (PnL), and Maximum Draw down. Additionally, we propose turnover regularization, a method that inherently constrains the turnover of generated positions within predefined limits. Our findings demonstrate that the proposed loss functions, in conjunction with turnover regularization, outperform the traditional mean squared error loss for return prediction tasks when evaluated using algorithmic trading metrics. The study shows that financially grounded metrics enhance predictive performance in trading strategies and portfolio optimization.

Finance-Grounded Optimization For Algorithmic Trading

TL;DR

This work tackles the misalignment between standard deep learning objectives and financial performance in algorithmic trading by proposing finance-grounded loss functions derived from , , and , coupled with turnover regularization to constrain trading activity. It demonstrates that optimizing these economics-oriented objectives on Binance crypto data (61 assets across multi-frequency horizons) yields superior risk-adjusted returns compared with traditional regression losses, with LogMDDLoss and ModSharpeLoss often providing the strongest performance and stability. The findings underscore the value of directly optimizing economic performance metrics for trading and portfolio management, offering a path toward more robust, interpretable AI-driven strategies in finance.

Abstract

Deep Learning is evolving fast and integrates into various domains. Finance is a challenging field for deep learning, especially in the case of interpretable artificial intelligence (AI). Although classical approaches perform very well with natural language processing, computer vision, and forecasting, they are not perfect for the financial world, in which specialists use different metrics to evaluate model performance. We first introduce financially grounded loss functions derived from key quantitative finance metrics, including the Sharpe ratio, Profit-and-Loss (PnL), and Maximum Draw down. Additionally, we propose turnover regularization, a method that inherently constrains the turnover of generated positions within predefined limits. Our findings demonstrate that the proposed loss functions, in conjunction with turnover regularization, outperform the traditional mean squared error loss for return prediction tasks when evaluated using algorithmic trading metrics. The study shows that financially grounded metrics enhance predictive performance in trading strategies and portfolio optimization.

Paper Structure

This paper contains 19 sections, 72 equations, 6 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Methodology
  4. Problem Statement
  5. Evaluation Metrics
  6. Custom Loss Functions
  7. Turnover Regularization
  8. Experiment
  9. Baselines and predictive models.
  10. Prediction protocol and portfolio construction.
  11. Optimization details.
  12. Portfolio optimization experiments.
  13. Implementation.
  14. Results and Discussion
  15. Loss Functions
  16. ...and 4 more sections

Figures (6)

  • Figure 1: Return distributions and QQ plots for BTCUSDT, ETHUSDT, and SOLUSDT. Top: histograms of returns with density estimates; bottom: QQ plots against the normal distribution, highlighting heavy-tailed characteristics.
  • Figure 2: The dependence of the SharpeLoss and ModSharpeLoss values on the magnitude of the generated positions.
  • Figure 3: Cumulative profit-and-loss (PnL) trajectories of the top 15 alphas ranked by Sharpe ratio over the test interval, together with the Buy&Hold benchmark. Finance-grounded loss functions produce consistently smoother and higher-performing strategies compared to the passive baseline.
  • Figure 4: Alphas performance.
  • Figure 5: Alphas correlation heatmap.
  • ...and 1 more figures