Trade in Minutes! Rationality-Driven Agentic System for Quantitative Financial Trading

TL;DR

The paper addresses the challenge of autonomous quantitative trading by reducing emotional biases and deployment latency through a rationality-driven agentic system. TiMi introduces a three-stage policy-optimization-deployment chain that decouples offline strategy development from minute-level live execution, combining semantic analysis, code programming, and mathematical reasoning across four specialized agents. It formalizes the market environment as $(\mathcal{M}, \mathcal{W}, \mathcal{S}, \mathcal{F}, \mathcal{J})$ and aims to maximize $\mathcal{J}(\pi_\Theta)$ via a hierarchical optimization that yields a pair-specific, programmatic trading bot with layers for strategy, function, and parameter. Empirically, TiMi demonstrates stable profitability and efficient deployment across 200+ trading pairs in stock and crypto markets, outperforming baselines in ARR, risk-adjusted metrics, and latency.

Abstract

Recent advancements in large language models (LLMs) and agentic systems have shown exceptional decision-making capabilities, revealing significant potential for autonomic finance. Current financial trading agents predominantly simulate anthropomorphic roles that inadvertently introduce emotional biases and rely on peripheral information, while being constrained by the necessity for continuous inference during deployment. In this paper, we pioneer the harmonization of strategic depth in agents with the mechanical rationality essential for quantitative trading. Consequently, we present TiMi (Trade in Minutes), a rationality-driven multi-agent system that architecturally decouples strategy development from minute-level deployment. TiMi leverages specialized LLM capabilities of semantic analysis, code programming, and mathematical reasoning within a comprehensive policy-optimization-deployment chain. Specifically, we propose a two-tier analytical paradigm from macro patterns to micro customization, layered programming design for trading bot implementation, and closed-loop optimization driven by mathematical reflection. Extensive evaluations across 200+ trading pairs in stock and cryptocurrency markets empirically validate the efficacy of TiMi in stable profitability, action efficiency, and risk control under volatile market dynamics.

Figures (6)

• Figure 1: Architecture of the proposed TiMi system comprising three stages: policy, optimization, and deployment.TiMi implements a decoupling mechanism where the initial two stages develop and optimize prototype trading bots through offline simulations by leveraging specialized LLM capabilities, while the deployment stage executes thoroughly refined bots with tuned parameters in live trading. This paradigm separates complex reasoning from time-sensitive execution, enabling both comprehensive strategy development and quantitative-level efficiency across market dynamics.
• Figure 2: TiMi Implementation for $\mathcal{B}^*$
• Figure 3: Comparison of action latency (left) and capital utilization (right) for representative methods.
• Figure 4: Comparative performance (ARR) distributions of different methods across trading pairs.
• Figure 5: Comparison of trading bot variants: $\mathcal{B}$, $\mathcal{B}^*$, and their intermediate versions (1/3-cycle optimization), simulated in 2024 cryptocurrency markets.