AI 交易沙盒挑战赛·美股站

主办方：Creek Labs by SAIFPartners
技术合作方：香港大学 HKUDS 实验室

赛事介绍

“如果让 AI 在真实市场结构下交易，它到底能走多远？”

Creek Labs 为开发者、投资爱好者、AI 创作者们搭建了一个开放的实验环境 C Field——这是一个专门为「AI × Trading」设计的线上沙盒。希望通过本次“AI 交易沙盒挑战赛·美股站”，让每个人都能在 C Field 上用自己的方式探索 AI 驱动交易的边界。

这里不是学院研究，也不是投研实训；更像是一次 AI 交易黑客松（Trading Hackathon） + 创意策略大赛。

赛事信息

• • 活动形式：
  线上初审 → 为期一个月的线上沙盒模拟盘挑战赛 → 实盘测试以及与优秀团队形成长效合作机制。其中线上挑战赛的实时表现将在 Creek Labs 官网通过 Live Trading Board 向全网实时更新。
  
• • 招募目标人群：
• • 有意向使用 AI 参与投资研究和决策的普通用户
• • 会 Prompt 的 AI 使用者
• • 有开发能力的自研开发者
• • 交易范围：美股股票、ETF和期权。
• • 参赛形式：个人或团队均可，团队 ≤ 3 人
• • 报名材料：交易策略或技术方案的简要展示材料
  （包含：逻辑框架、AI 应用方式、风控方法、调研依据）
• • 评审团：Creek Labs 投资专家、香港大学技术导师、私募基金 / AI 技术顾问
• • 沙盒平台：C Field 是 Creek Labs 为此次沙盒挑战赛开发的 AI 交易沙盒平台，提供从 Bot 对话 到 自定义 Agent 再到 自研框架接入 的全链路环境，其中模拟盘基于成熟的美股市场结构。
  

参赛分组

1. Bot 组：不写代码，也能参赛

• • 适合人群：投资爱好者、非技术背景参赛者
  通过 C Field 提供的对话式 AI 交易助手（Bot），提升交易逻辑与因子研究效率，叠加参赛者的个人判断，最终形成交易方案，在 C Field 模拟盘界面手动下单
  
• • 限制：不支持自动化执行，Token 限额（保证系统稳定性）
• • 核心：有想法，让 AI 帮你补足

2. Agent 组：Prompt Engineers 的主场

• • 适合人群：会 Prompt、懂投资逻辑、懂得调试基础 Agent 参数，能理解简单基础代码的参赛者
  在 Bot 基础上，通过 C Field 提供的自定义 Prompt 的 Agent 面板建立策略 Agent，进行深度交易逻辑与因子研究、形成交易方案，并可通过以下方式执行：
  
• • 全自动执行
• • 半自动执行（每次下单需确认）
• • 仅建议模式（后续需人工手动下单）
• • 限制：禁止 Agent 执行频率过高，Token 限额较 Bot 组宽松（保证系统稳定性）
• • 核心：让 Prompt 变成真正能跑的交易 Agent

3. 自研组：真正的 Builder

• • 适合人群：能写工程架构、能进行 AI 项目开发的开发型参赛者
  致力于以 AI 算法与模型开发为核心，探索利用 AI 发现交易信号和因子的可能；通过接入 C Field 开放接口，将自研项目接入 C Field 获取实时行情并完成交易
  
• • 核心：用技术把交易链路重构

AI Trading Sandbox Challenge · U.S. Market Edition

Presented by:
Creek Labs by SAIF Partners
Powered by:
Data Intelligence Lab@HKU

Event Introduction

“If we let AI in a real market structure — how far can it go?”

Creek Labs has built C Field, an open experimental environment for developers, investors, and AI creators —a sandbox designed specifically for “AI × Trading.”

Through this “AI Trading Sandbox Challenge · U.S. Market Edition,” we invite participants to explore the boundaries of AI-driven trading in their own way.

This is not academic research and not an investment training program —it’s more like a Trading Hackathon + Creative Strategy Competition.

Event Information

• • Format:
  Online screening → One-month sandbox trading challenge → Live trading tests and potential long-term collaboration with top teams. Real-time performance during the challenge will be displayed on the Creek Labs website via the Live Trading Board.
  
• • Who Should Join:
• • Anyone interested in using AI for investment research and decision-making
• • Skilled AI Prompt users
• • Independent developers with coding experience
• • Trading Scope:
  U.S. stocks, ETFs, Options.
• • Participation Type:
  Individual or team (up to 3 members)
• • Application Materials:
  A concise presentation of your trading strategy or technical proposal, including:
• • Logic framework
• • AI application method
• • Risk control approach
• • Research basis
• • Judging Panel:
  Creek Labs investment experts, HKU technical mentors, private fund managers, and AI technology advisors
• • Sandbox Platform:
  C Field is the AI trading sandbox platform developed by Creek Labs for this challenge.It provides a full-stack environment — from Bot-level interaction to custom Agents and independent framework integration, all simulated on the real U.S. market structure.
  
• • Key Evaluation Metrics:
  AI experimentation ability, strategic creativity, and technical execution

Competition Tracks

1. Bot Track – Trade Without Code

• • For:
  Investors and non-technical participants.
  Use C Field’s conversational AI trading assistant (Bot) to enhance your trading logic and factor research, combine it with your personal judgment, and manually execute trades on the C Field simulation interface.
  
• • Limitations:
  No automated execution; limited Tokens (to maintain system stability)
• • Core Idea:
  Bring your ideas — let AI fill in the gaps.

2. Agent Track – The Arena for Prompt Engineers

• • For:
  Participants familiar with Prompting, investment logic, and basic Agent tuning, as well as those who can understand simple coding structures. Build your own strategy Agent using C Field’s customizable Agent interface, conduct in-depth research, and execute strategies through one of three modes:
• • Fully automated execution
• • Semi-automated execution (manual confirmation required per trade)
• • Advisory mode (manual trading after recommendations)
• • Limitations:
  Restricted high-frequency Agent execution; higher Token limit than the Bot Track (for system stability)
• • Core Idea:
  Turn your Prompt into a truly functioning trading Agent.

3. Developer Track – The True Builders

• • For:
  Developers capable of designing architectures and building AI-based trading systems. Focus on using AI models and algorithms to detect trading signals and factors. Integrate your project with C Field’s open API, access real-time market data, and execute trades directly.
• • Core Idea:
  Rebuild the entire trading pipeline with technology.