Quant Approach

A.I Powered Quantitative Hedge Fund

End-to-End Deep Reinforcement Learning

Recent advancements in Deep RL have enabled today’s computers to achieve groundbreaking success, such as beating the world’s best human at the board game “Go.” One of the amazingly key differences between Deep RL agents and traditional expert agents with handcrafted features is that Deep RL produces outputs from input data completely on its own, without any programming. Through SURPLUS ™ our proprietary algorithms have changed from “if A happens, then do B” to optimizing a total reward regardless of the action. SURPLUS ™ autonomously explores different trading strategies without any limits to find the best trades that humans can and cannot.

Machine Vision and Technical Analysis

Pouncil Capital uses the most up-to-date CNN to model Futures price data. We do this by embedding Deep Convolution Neural Network (CNN) models that have surpassed human performance in recent ImageNet visual recognition challenges into our SURPLUS ™ ecosystem. With millions of parameters in the model, the network is able to pick up features that human traders have ignored.

Systematic Approach Deprived of Human Emotion

It’s a well known fact that humans can be trained to be great traders in a simulation setting, but fail terribly when trading with real money. Emotions and doubts will lead one astray most time, but only those who can control their energy, emotions, patience, self control and disciple while executing risk management and greed management put themselves consistently in a high probability trading opportunity. Such traits are by design the most capable for computers. With computer programming and non-linearity provided by deep neural networks, the machines have the edge.

GPU Powered Trading Execution

SURPLUS ™ is trained, tested and run on the most cutting edge GPU hardware. Each agent is connected with one GPU dedicated to trading one security at one time frame.

Scalability and Flexibility in Model Design

Pouncil Capital’s models are designed from the ground up with open source Machine Learning libraries such as Tensorflow, OpenAI gym, and Keras. They are easily scalable for trading futures and other financial securities. Our model training platform can assign multiple trading time scales, input different formats of data, and optimize customized reward functions in regard to risk measurement metrics.