CSTrader: A Testbed for Language-Grounded Trading in a Community-Driven Virtual Asset Market

2026-06-30

AI Summary

This work proposes CSTrader, a multi-agent framework designed to address the challenges of trading niche, highly volatile virtual assets—such as CS2 skins—that are driven by community sentiment and poorly served by traditional quantitative models. CSTrader uniquely integrates large language models (LLMs) within a multi-agent architecture, where specialized agents process heterogeneous signals including technical indicators, liquidity metrics, event data, and contrarian sentiment to generate trading decisions. The framework further incorporates transaction cost modeling and portfolio risk management to enable an end-to-end mapping from natural language inputs to executable trades. Evaluated on real-world CS2 skin market data, CSTrader achieves a cumulative return of 7.58%, substantially outperforming both the market benchmark (−15.62%) and a single-prompt LLM baseline, thereby demonstrating the efficacy and feasibility of structured language-based agents in high-noise trading environments.

Abstract

Niche asset markets, such as Counter-Strike 2 (CS2) weapon skins, are small, volatile, and heavily driven by community discussions and platform rules. These properties make them hard for traditional quantitative models, but provide an ideal testbed for studying how large language models (LLMs) turn unstructured text into trading actions. We present CSTrader, a multi-agent framework for language-grounded trading in the CS2 skin market. The system first integrates heterogeneous signals from various sources, then uses specialized agents for technical analysis, liquidity, events, and (reversed) sentiment, and finally applies risk control, transaction friction, and portfolio management agents to produce buy, sell, or hold decisions under realistic trading frictions. We build a live-like evaluation environment with real CS2 data from a highly volatile period and evaluate several recent LLM backbones. Across models, CSTrader consistently outperforms both a falling market index (-15.62%) and simple single-prompt LLM baselines, achieving up to a 7.58% cumulative return with controlled risk. Ablation studies show that liquidity, reversed sentiment, and transaction friction agents are crucial for turning noisy language signals into stable profits, suggesting that niche, language-driven markets are a useful benchmark for future language-to-action research. Code is available at: https://github.com/IatomicreactorI/CSGOTrading?tab=readme-ov-file#quick-start

Problem

• language-grounded trading
• niche asset markets
• large language models
• unstructured text
• virtual asset market

Innovation

• language-grounded trading
• multi-agent framework
• large language models
• virtual asset market
• trading friction