Quantitative Trading: The Predictability Of Stock Markets

Tomer Solel is a Financial Analyst at I Know First. He graduated from Cal Poly Pomona with a bachelor’s degree in applied mathematics.

Quantitative Trading

Summary

  • People have been trying to forecast the stock market for a very long time. Today, we have more advanced tools to calculate it, making markets very computerized.
  • The I Know First algorithm has 6 different time horizons, and each stock in each time horizon has a predictability between -1 and 1 representing a quality control indicator.
  • Some markets are more predictable than others, and we found the long term forecast to be more reliable than the short term one.
  • There are different reasons for which some markets are more predictable than others and those include time horizon, uncertainty, the human factor, and more.
  • Using computer models generates more objective predictions.

Quantitative Trading

Introduction

People have been trying to forecast the stock market for a very long time. A hundred years ago, the common tools to predict the stock market were hand drawn charts and lines, but with the introduction of computers, more sophisticated tools to predict the stock market were developed, leading to many indicators being available today. But the more popular these indicators are, the less useful they become. Thus, we need better tools to exploit the inefficiencies that the market possesses.

Nowadays, the markets are becoming, even more, computerized, making the response time going down to being measured in milliseconds. We ask ourselves the question of whether it is still possible to forecast the stock market in advance for different time horizons, or whether the markets have become totally efficient and impossible to predict.

Markets’ Predictability

From August 2011 to June 2012, we recorded the predictability of over a hundred equities by our machine learning system, a tool used to forecast the future movement curve of the market based on past history. Our algorithms constantly look for patterns in the markets, make and test conjectures, and provide a daily stock forecast for six different time horizons (3 days, 7 days, 14 days, 1 month, 3 months, and 1 year). This self-learning algorithm is evolving constantly, adding more data and creating a better machine model.

The measure of predictability, our quality control indicator, accompanies each market forecast. P is a correlation coefficient between the predicted move and the actual move. P ranges anywhere from -1 to 1 where -1 corresponds to an actual move that is opposite of the forecast, 0 corresponds to an actual move that has no relationship to the forecast, and 1 corresponds to an actual move that is identical to the forecast.

Results

To see how predictable, the markets are, we make observations from the records for the time period mentioned above. We see the following:

  1. Through the time period, the predictability of the top 100 most predictable markets in our system was a remarkable 0.53.
  2. Some markets were on average more predictable than others. For examples, our graphs show that the DAX index was more predictable than the Disney stock.
  3. Each market had a unique predictability curve, not necessarily synchronized with other markets.
  4. There were long periods of predictability interspersed with a few short unpredictability spikes.
  5. The long term forecast was more reliable than the short term forecast.
  6. We can see some different waves in predictability.

Quantitative Trading Quantitative Trading

 

Discussion and Conclusions

 

Looking at averages, some markets were more predictable than others, making them more efficient. But sometimes, shocking news can have a drastic effect regardless of its relevance to a specific market. A good example is what happened between 5/11/12 – 5/24/12 when the topics of European sovereign debt fears and the disappointing U.S. data were all over the news. This unpredictability spike affected both the German DAX index and the Disney stock which is a stock that should seemingly not be affected by this type of news.

Our long term forecast was more predictable than our short term forecast. That is because the short term forecast is more affected by market news created by the daily news. The long term forecast reflects deeper fundamental trends.

In spite of proliferation and computerization and algo-trading, the markets are still chaotic yet predictable.

 

This is because:

  • There are differences in evaluating the equity between different models and different markets, humans or machines. What seems undervalued in one place can seem overvalued in another place.
  • Time horizon factor: Short-term forecasts look at different things than long-term forecasts.
  • Uncertainty: It is very difficult to quantify the effect of every news event on the stock market.
  • Human factor: different people react to news differently. There is no such thing as an objective human mind, as we tend to react to the latest headline on the news, only until the next item catches our mind. Fear and greed can lead us to irrational decisions.

These inevitable factors listed above all lead to a shift in prices. Using computer models is a lot more objective. By monitoring predictability, we can get a head start in preparing for when the market paradigm change gets in progress.


premium