How AI Trading Systems Perform in Live Markets: Inside the WEEX AI Hackathon Beta Test

Launched by WEEX Labs, Alpha Awakens: the WEEX AI Hackathon is a truly global trading hackathon that pushes AI directly into real market conditions, with a total prize pool of up to $1.88 million. So far, 788 participants from around the world have gathered, pitting their AI strategies head-to-head in live market action — no “paper trading,” only real trades and results proven by the market. The preliminary round is now in full swing, watch it live here: https://www.weex.com/events/ai-trading

Before the official battle began, WEEX launched a Beta Test to provide participants with a proving ground that closely mirrors real trading conditions, allowing strategy stability, execution capability, and risk control to be tested by the market in advance, while refining an AI trading infrastructure that can truly be deployed in live markets. During this Beta Test, a group of standout participants emerged — they are not just “people who write code,” but AI architects actively competing with the market. Next, we take you into the real-world thinking and strategy landscape of pioneer participant Kivi.

From single-factor quant strategies to Neuro-Symbolic AI: building a fallback-ready, constraint-aware, multi-factor intelligent trading system based on the WEEX API

Kivi, a liquidity derivatives intern at WEEX and a seasoned quant trading enthusiast, started from single-factor quant strategies and progressively introduced risk control and capital management, integrating large language model capabilities to build a “constraint-aware, fallback-ready” Neuro-Symbolic AI hybrid trading system. The foundation relies on Python for stable execution, while AI operates at a higher level to handle alpha discovery and parameter optimization. During the WEEX Beta Test, he refined high-concurrency stability through the API and applied a multi-dimensional scoring mechanism based on WAD and volatility to filter false breakouts, allowing AI to dynamically decide whether to enter trades. The core value of the system lies not in greater aggressiveness, but in greater robustness — enabling AI to automatically switch risk control modes based on market conditions, transforming traders from direct market participants into architects of an AI-driven trading legion.

Q1: Can you briefly introduce your background and the type of trading or technical direction you are currently focused on?

A1: I am a passionate quant trading enthusiast, currently an undergraduate finance student at Curtin University Singapore, as well as an intern in the Liquidity Derivatives Department. My current focus is on Neuro-Symbolic AI. While today’s AI models have strong reasoning capabilities, allowing them to directly control an entire quantitative system carries significant risk and effectively turns it into a black box. To address this, I built a hybrid system: the lower layer uses Python to ensure a highly robust execution framework, while core components such as alpha discovery, parameter optimization, and risk-control decisions are fully handled in real time by large language models like DeepSeek and GPT.

Q2: When did you start participating in the WEEX Beta Test / Hackathon, and why did you decide to join?

A2: I began participating in the WEEX Beta Testing competition on December 31, 2025. I decided to join primarily to put my unconventional system architecture to the test in a real trading environment.

Q3: How did you initially start designing this strategy or system, and what was the overall approach?

A3: Before successfully implementing the system, my initial design focused on upgrading a traditional single-factor model. Over the following months, new ideas kept emerging. The first step was adding capital management and risk control modules to support the single-factor model. Later, I began to consider whether introducing AI assistance could better adapt the system to market uncertainty. Based on the existing core modules, I gradually integrated AI so that before each trading cycle, the AI could adjust parameters for different trading pairs according to current market conditions. Eventually, I introduced additional AI components such as an Executor Agent and an Alpha Factory, enabling the system to evolve from a single-factor model into a multi-factor quantitative trading system.

Q4: How did you integrate and use the WEEX API in this project, and what key problems did it help you solve?

A4: I connected to the WEEX API using the Python Software Development Kit (SDK), primarily to address stability under high-concurrency conditions, ensuring the system can operate smoothly and continuously.

Q5: What specific debugging work did you focus on during the tuning process?

A5: The most memorable adjustment was modifying the TWAP (Time-Weighted Average Price) algorithm. TWAP works well for large orders, but when applied to smaller positions, the split orders can become too small to open a position successfully. To address this, I changed the logic so that if a TWAP entry fails, the system automatically switches to a market order, avoiding missed opportunities due to failed entries.

Q6: Which core signals does your strategy primarily rely on, such as trend, volatility, or sentiment, and why did you choose them?

A6: The entire system is built around a multi-dimensional scoring framework to determine whether to enter a trade. The two signals I value most are Williams Accumulation Distribution (WAD) and volatility. I chose these because large language models excel at handling nonlinear relationships. Relying on a single indicator can easily lead to losses due to false breakouts, but by combining RANK-WAD with volatility, the AI can more accurately distinguish between genuine and false breakouts. Only when the multi-dimensional scoring criteria are met does the AI open the “gate” for the multi-factor model to enter a trade.

Q7: During the strategy design process, which WEEX rules or mechanisms had a direct impact on your approach?

A7: During the beta testing phase, there were no specific rules or mechanisms that directly affected the design of my system.

Q8: Was there a series of decisions that made you clearly feel the stability or consistency of the AI for the first time?

A8: The introduction of AI enabled the system to dynamically switch risk-control modes based on market conditions, proactively reducing exposure in unfavorable environments instead of passively enduring consecutive losses.

Q9: Did your mindset change during the competition, and did the presence of AI affect your emotional involvement in decision-making?

A9: My mindset did not change significantly during the competition. To some extent, the presence of AI reduced my emotional involvement in the trading process. Since key strategy adjustments and risk-control mode switches are driven by clearly defined rules and state-based decisions, I no longer intervene frequently due to short-term gains or losses, and instead focus on ensuring that the system operates within its predefined risk boundaries.

Q10: Looking back on this Hackathon / Beta Test, what was your biggest takeaway or shift in understanding?

A10: This experience reinforced my belief that for an AI trading system to be truly deployable in real markets, the core lies not in the strategy model itself, but in positioning AI within a fallback-ready and constraint-aware framework—so it serves as an enhancer of system stability rather than an amplifier of risk.

Q11: Throughout the process, in what specific ways did WEEX genuinely support you?

A11: WEEX provided an environment that was close enough to real-market conditions while still allowing room for trial and error, enabling me to validate many of my system design assumptions under actual trading rules.

In Kivi’s view, the Beta Test was not about competing for profits ahead of time, but about placing AI and trading systems into real market conditions to see whether the system could run stably and whether the AI was properly constrained. In practice, this round of testing exposed a number of issues at both the system and execution levels, while also helping clear obstacles around API stability, rule alignment, and overall workflow—laying the groundwork for a smoother progression of the competition.

See how AI trading works in real markets during the WEEX AI Trading Hackathon beta with live Crypto trading, AI trading strategies, and real competition results.

The purpose of this pre-competition Beta Test was to move participants directly from the “idea stage” into real trading environments, allowing systems, strategies, and AI to be tested in live-market conditions and laying a solid foundation for the main event. The core rule of the competition is clear: AI must be involved—whether in decision-making, risk control, execution, or auxiliary analysis. The focus is not on how high the returns are, but on how AI is genuinely integrated into a trading system. The competition has now officially begun, with the event fully entering the live-market battle phase. This marks a key window in which strategies across AI Wars are being actively deployed and revealed. If you want a systematic view of how AI trading ideas operate in real markets, now is the best time to watch.

Related Reads

AI Crypto Trading in 2026: How AI Assistants Are Reshaping Trading Platforms and Strategies

AI Trading in 2026: Bitcoin, Ethereum, and the Shift Toward Changing Crypto Trading

About WEEX

Founded in 2018, WEEX has developed into a global crypto exchange with over 6.2 million users across more than 150 countries. The platform emphasizes security, liquidity, and usability, providing over 1,200 spot trading pairs and offering up to 400x leverage in crypto futures trading. In addition to traditional spot and derivatives markets, WEEX is expanding rapidly in the AI era — delivering real-time AI news, empowering users with AI trading tools, and exploring innovative trade-to-earn models that make intelligent trading more accessible to everyone. Its 1,000 BTC Protection Fund further strengthens asset safety and transparency, while features such as copy trading and advanced trading tools allow users to follow professional traders and experience a more efficient, intelligent trading journey.

Follow WEEX on social media

X: @WEEX_Official

Instagram: @WEEX Exchange

TikTok: @weex_global

YouTube: @WEEX_Global

Discord: WEEX Community

Telegram: WeexGlobal Group