Stock pretraining improves cross-market trading using diverse technical indicators
DIVINE: Simple Cross-Market Stock Pretraining via Diverse Indicator Reconstruction
Machine LearningComputational Engineering, Finance, and Science
Summary
Predicting stock prices is difficult because the rules change across markets and over time. The authors propose a new method called DIVINE that learns useful patterns by teaching a computer to reconstruct many common financial indicators from raw price and volume data. This approach uses data from multiple markets together and creates a compact model that performs better than bigger models on ranking stocks to buy or sell. Their findings show that having a range of different indicators and data from various markets helps the model make better decisions and transfer knowledge across markets.
What this means in practice
- •For quantitative traders: Use a pretrained lightweight model to improve stock ranking and portfolio construction across different equity markets.
- •For financial software developers: Build stock analysis tools that integrate diverse technical indicators reconstructed from raw market data for better prediction and robustness.$Commercial implications: Enables development of advanced trading software products with improved predictive features that appeal to investment firms.
Authors
Kuan-Yu Chen, Shu-Cheng Zheng, Yu-Chen Den, Wei-Cheng Liao, Tien-Hao Chang
Abstract
Financial time-series pretraining typically learns from masked observations, contrastive relations, or future outcomes---yet existing objectives struggle to simultaneously avoid future-supervision uncertainty and maintain return-prediction alignment. We propose DIVINE (DIVerse INdicator rEconstruction), a simple cross-market pretraining framework that reconstructs technical indicators from raw OHLCV history. Computed from observed price-volume history, technical indicators provide consistently defined supervision across markets while summarizing diverse market dynamics with established relevance to return prediction. Pretrained jointly on six-equity market datasets, DIVINE reconstructs 77 targets derived from 16 standard indicators and transfers only the learned encoder to downstream stock ranking. Across all six markets, DIVINE achieves the strongest average portfolio performance with a lightweight 0.05M-parameter encoder, outperforming pretraining baselines and matching or exceeding substantially larger financial foundation models, while remaining robust and data-efficient. Systematic analyses show that indicator diversity and market diversity provide complementary gains in transfer. Together, these results suggest that supervision design and cross-market diversity---rather than model scale---are the key drivers of strong, transferable financial representations.