Accuracy Law for the Future of Deep Time Series Forecasting

2025-10-03

AI Summary

This study aims to establish the theoretical performance upper bound for deep time series forecasting. Addressing limitations of conventional sequence-level predictability metrics, we propose the “Accuracy Law,” the first formulation revealing a significant exponential relationship between window-level pattern complexity and the minimum achievable prediction error. Methodologically, we conduct rigorous statistical validation across 2,800+ newly trained deep models, employ seq2seq architectures to capture dynamic window-level characteristics, and develop a framework for estimating fundamental error lower bounds. Our contributions are threefold: (1) We derive the first theoretically grounded performance limit specifically for deep forecasting models; (2) We identify saturated tasks—where further model improvements yield diminishing returns—across multiple mainstream benchmarks, empirically validating the universality of the Accuracy Law; and (3) Leveraging these insights, we propose an efficient large-model training strategy that substantially enhances research and development productivity.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Learning to Search
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Deep time series forecasting has emerged as a booming direction in recent years. Despite the exponential growth of community interests, researchers are sometimes confused about the direction of their efforts due to minor improvements on standard benchmarks. In this paper, we notice that, unlike image recognition, whose well-acknowledged and realizable goal is 100% accuracy, time series forecasting inherently faces a non-zero error lower bound due to its partially observable and uncertain nature. To pinpoint the research objective and release researchers from saturated tasks, this paper focuses on a fundamental question: how to estimate the performance upper bound of deep time series forecasting? Going beyond classical series-wise predictability metrics, e.g., ADF test, we realize that the forecasting performance is highly related to window-wise properties because of the sequence-to-sequence forecasting paradigm of deep time series models. Based on rigorous statistical tests of over 2,800 newly trained deep forecasters, we discover a significant exponential relationship between the minimum forecasting error of deep models and the complexity of window-wise series patterns, which is termed the accuracy law. The proposed accuracy law successfully guides us to identify saturated tasks from widely used benchmarks and derives an effective training strategy for large time series models, offering valuable insights for future research.

Problem

• Estimating performance upper bound for deep time series forecasting
• Identifying saturated tasks in standard forecasting benchmarks
• Discovering relationship between forecasting error and series pattern complexity

Innovation

• Estimates performance upper bound for forecasting
• Discovers exponential relationship between error and complexity
• Derives effective training strategy for large models