Where We’ve Been
Since the 1950’s, the academic and professional finance industries have been developing models to more accurately price assets. It is clear that overall market forces impact an individual stock’s performance, but what other factors are there?
Fama and French – The OGs
In 1992, Eugene Fama and Kenneth French developed their revolutionary three factor model. It incorporated market forces, size premium, and value premium factors to determine the expected return, or estimated risk of a stock. The size premium refers to the market cap of the stock and whether it has a large or small market cap. The value premium refers to the book to market value of the stock. A high book value / market value would imply a relatively inexpensive stock.
In 2014, Fama and French published their five factor model which included momentum, quality, and volatility. The results were better than the three factor model, but the imfamous ϵ – or error/noise factor always plagued the model. It was clear that even five factors was not able to explain all of the price variances we see in the market
The Hunt is On
With this new mental framework, the asset pricing research exploded with ideas. A variety of papers came out testing a whole host of different factors. The idea of factor averaging (combining multiple metrics or metric rankings into one averaged metric) was born and implemented in a similar fashion to Fama and French’s model. The issue was how do we determine which model is the best?
Machine Learning to the Rescue
Over the past few years, the research has shifted from inventors creating factors, testing them out, and reporting their accuracy, to massive comparisons of a variety of models to determine their efficacy.
A few methods have been implemented. In 2018, the Basillas and Sha…. method was born that leverages complex mathmatics and bayesian techniques to compare a models ability to predict stock price movements.
Leveraging the model, the researchers can run simulations to see how accurate the models are and identify the best one.
The CZZ Models Changed the Game
This year, the team of …. revolutionized the model comparison game. They took apart the Basillias and Shen… model and identified mathematical points of failure. They developed a new method with additional assumptions that lead to a development of a much more accurate and valid model. Using their model, they tested 128 different asset pricing models using 17 metrics? which cover metrics discussed in the academic papers seen over the past 30 years to identify the best model.
What is the Best Model?
The CZZ method identified 3 models with significantly more predictive power than the rest. They ran millions of simulations and compared the model’s expectations to reality. A few major factors were identified as the significant predictors of asset pricing. The paper is not yet published, but once it is, we will all be able to see the results.
In the meantime, we have developed an algorithm to analyze stocks using the CZZ model in realtime. Knowing that the CZZ model has supreme predictive power, we know our real time model is predicting investment opportunities more accurate than any other model.
You can get access to our model by signing up to the website and going to the stock analyzer section