Quantative Trade Strategies

Algorithmic Trading is a technique of deploying algorithms that automatically buy and sell stocks in response to market data.

The alpha idea

At fallbrook we work the full lifecycle srom the first step in developing a trading algorithm is coming up with an “alpha idea” to iterating its implementation.

What is alpha?

Basically, the return of a portfolio can be written as:

r = beta * rm + alpha

Where beta measures the correlation of the portfolio return to the overall market return (r_ _m), and alpha is the excess return.

1. For instance, an index fund has alpha=0, beta=1. But, what I am really after is a positive non-zero alpha, which would indicate that my algorithm is doing better than the market, and at the same time beta=0, i.e. I want my algorithm to be independent form the market, “market-neutral”.

To get started, I want to try the following alpha idea: Companies with a high revenue growth are doing well, and I want to invest in them. Companies with low revenue growth are doing poorly, and I want to short them.

• Backtest on a larger timeframe. • Try adding more alpha factors. • Analyze alpha factor correlations. • Weight distribution: how to optimally weigh the factors? (this is currently one of the fields of active research, and a field where Machine Learning could help.) • Optimize trading behavior, e.g. the trading quantiles. • Analyze sector exposure. • etc.

Summary

To summarize, the algorithmic trading workflow is this: 1. come up with an “alpha idea” 2. estimate alpha using alphalens 3. run full backtest on past market data 4. analyze the backtest using pyfolio