Project Description and Overview
AlgoTrade is a fully automated stock market trading algorithm consisting of multiple diversified trading strategies, which makes up a whole strategy portfolio. The AlgoTrade Team is divided into three sub teams. They are the Algorithmic Team, the Quantitative Analysis Team, and the Financial Analysis Team. During the winter semester, members of AlgoTrade will be redistributed, giving them the chance to work on the specific algorithm that they are most passionate about. The goal is to ensure that each algorithm is brainstormed and developed by team leads and members from all three teams, resulting in a diversified team distribution.
Overview of Team Workflow
All team leads and members will be redistributed to contribute to one specific strategy based on their passion and background. In other words, each strategy-based team will contain talents from algorithmic team, quantitative analysis team, and financial analysis team. However, each individual’s task may still vary based on the team he or she was in during the acceptance stage. The following context explains how:
Responsible for the algorithm implementation, specifically using Python. The implementation will be based on the model provided by the Quantitative Analysis Team. Responsible for backtesting the algorithm and assisting quantitative analysis team with the analysis process.
Quantitative Analysis Team
Responsible for designing models and algorithms for a specific strategy. The model will be based on certain hypothesis proposed by the financial analysis team and the quantitative analysis team itself. Conduct mathematical and statistical analysis to optimize trading strategies.
Financial Analysis Team
Responsible for giving financial insights to the data that AlgoTrade collects and recommending the most suitable data resources for each specific strategy. Financial Analysis Team will be working very closely with the Quantitative Analysis Team on strategy implementation.
With the dynamic and volatile property of financial market, reinforcement learning can avoid mathematically modelling its behaviour. Instead, the algorithm will directly learn from its past behaviours and their corresponding consequences. In other words, the investment decisions being made is a Markov Decision Process (MDP), meaning the trading strategies obtain its learning resources from its very own interaction with the market. This could potentially eliminate the need to forecast and bluntly predict the market. Such algorithm will be consisted of multiple supplementary algorithms and methods. Such as Kalman Filter and Unsupervised Learning (applied to data analysis).
As the name indicates, factor investing is the method of identifying factors that may affect the stock returns. These factors might include value, size, momentum and volatility. We might also be considering psychological factors or anything that we believe will affect the stock price. Each factor will have different impacts on stock price, as well as different magnitude of impact. We will be able to quantify that using statistical and mathematical model.
Delta, theta, and vega hedging
Delta, theta, and vega are different ways to measure the sensitivity of an option's price to quantifiable factors. Strategies like delta neutral, theta neutral, and vega neutral are used to hedge against the risks of price sensitivity, time sensitivity, and implied volatility, naturally bringing down its exposure to volatility.