Pit.ai: Reinforcement Learning Hedge Fund Aims to Outperform Traditional Markets

Most hedge funds struggle to deliver consistent returns, but a new wave of startups is leveraging artificial intelligence and machine learning to gain an edge. Pit.ai, a startup adopted into the YC W17 class, is one such venture, aiming to revolutionize the hedge fund industry by employing a unique variant of reinforcement learning.

The Challenge with Traditional Hedge Funds

Hedge funds typically seek "alpha," or above-market returns, by employing aggressive trading strategies. In recent years, many firms have focused on acquiring vast amounts of data for information arbitrage. Companies specializing in geospatial analytics, for instance, use satellite imagery and computer vision to analyze data like car counts in retail parking lots to project earnings before official reports.

Pit.ai's Novel Approach: Reinforcement Learning

Unlike many of its peers, Pit.ai's founder, Yves-Laurent Kom Samo, explained to TechCrunch that the company deviates from the paradigm of using information not yet factored into stock prices. Instead, Pit.ai utilizes a specialized form of reinforcement learning to evaluate trading strategies directly. This approach contrasts with traditional reinforcement learning methods that rely on value functions, which can become overly complex and pseudo-scientific in the dynamic financial markets.

Traditional Reinforcement Learning:

In a typical reinforcement learning scenario, an agent learns to make optimal decisions by estimating value functions. For example, in a video game, an agent might gain points for increasing speed and lose points for driving off the road. Over many iterations, the algorithm is fine-tuned to maximize its score.

Pit.ai's Financial Application:

Yves-Laurent Kom Samo points out that applying this directly to finance is challenging because it requires modeling returns for every decision in every market state. Financial markets are incredibly complex systems. Pit.ai's strategy bypasses this by evaluating the trading strategies themselves, taking into account key financial metrics such as:

Sharpe Ratios: A measure of risk-adjusted return.
Maximum Drawdown (MDD): The peak-to-trough decline in an investment during a specific period.

Disrupting the Fee Structure

By focusing on a more efficient AI-driven strategy, Pit.ai aims not only to deliver above-average returns but also to break the traditional "two and twenty" fee structure common in the hedge fund industry. Without the need for large teams of analysts searching for macro-economic trends, Pit.ai can remain lean. This allows them to potentially eliminate management fees altogether, opting instead to collect only "carry" (a share of the profits) from their limited partners.

Company Vision and Future Plans

While Pit.ai has not yet raised a formal fund from limited partners, it is actively seeking venture capital to support its team of machine learning experts. Yves-Laurent Kom Samo, who holds a PhD in machine learning from Oxford and was a Google Fellow, plans to leverage his network for recruitment. Currently, the models are being run without real money, but the initial signs are promising. Kom Samo hopes to have a fund established and formal trading initiated within a year.

Key Takeaways:

AI in Finance: Pit.ai exemplifies the growing trend of AI and machine learning transforming the financial sector.
Reinforcement Learning: The company's unique application of RL to evaluate trading strategies offers a novel approach.
Efficiency and Fees: By optimizing operations through AI, Pit.ai aims to offer a more competitive fee structure.
Future Outlook: The startup is poised to make significant strides in the quantitative finance space.

Pit.ai: Reinforcement Learning Hedge Fund Aims to Outperform Traditional Markets