This summer I worked in the research wing of a hedge funds firm. The job description was simple: find new formulae to predict the movements of stocks. Sounds interesting? Well it is, until you run out of ideas.
Before the intern, my idea was that we would be taught methods of predicting movements of prices of various stocks in this intern and that people do this all the time and earn a lot of money. Well, that’s not how it works in that one cannot actually predict the stock price, but one can earn a lot of money here and that takes skill. What actually is done is you design some indicators that give you a “hint” as to which way the prices are going to move. Some simple examples are intuitive enough to understand like if the price steadily increased for many days, chances are high that it will go down the next day. Things like that and more. So you try to use the data that you already have and try to find such directions in which the market is headed and use it to your advantage. Only difference being that instead of people interpreting these indicators, it was the computer that was supposed to take decisions of buying and selling, of course based on what you have told it to do, but on thousands of stocks. So here’s what my intern looked like.
All of this starts with getting familiar with the terms used. There aren’t too many terms and most of them are very logical so there is no problem in grasping them. Others are statistical tools that gauge your performance. The most important one that I used for was the Sharpe ratio and the information ratio. After that came the indicators. The first week generally goes into understanding the basics and all. It’s not that it takes that long to understand it but it does take a while to actively apply them. What we basically did was get familiar to the most popular of the indicators and how they are used to judge the prices.
The second week went in understanding the basics behind the indicators. It’s easy to make one but it is essential to understand the working especially if you want to improve the indicator, which is what we are looking for. Any one indicator cannot be trusted all the time, especially in the markets today that tend to go in one particular direction…down. Indicators are generally best in a steady or “sideways” market, where there is not clear trend. If there is any direction, it’s a different story.
Next comes using these indicators to actually buy stocks of appropriate companies. What I did was deal with 3000 stocks in the US and occasionally in other countries like Japan and European markets and came up with what is called an alpha for each of the stocks. Alpha in simple terms is relative probability of each stock to go up (or down). Relative to what?? Well… each other. It’s easy to talk in terms of 2 stocks but when 3000 stocks are to be considered, things are not as simple. You give each a number, indicative of how much probability that stock will have of going up. Someone else may be assigned a higher or lower number, which means that between the two, the higher numbered has a higher probability of going up. It looks simple enough but coming up with a simple formula can be a tough job, and boring too sometimes.
In the next week, we graduated to looking at blogs and articles on such things. By the end of 2nd week, I had taken to looking at chart patterns which I continued in the third week too. Chart patterns are basically formed by price vs. time graphs of a stock that are a direct effect of the psychology of humans. The “drive” that people get of buying and selling when someone like Harshad Mehta deals in a stock is a very good example of such a tendency. These are the factors that have nothing to do with the company or its performance or its expectation of doing good or bad. It’s just the herd mentality that we essentially have as traders. And analysing such patterns can be real fun.
Analyzing such patterns, I tried to make some of my own indicators, but landed up remaking a lesser known indicator. But it was great kick finding out that I was on the right track.
This went on for another week and by the end of it, I was begging my mentor for a change. So next week we took a break from this routine to do a small programming project. It involved making the best strategy for a famous casino game. Turns out even the simple games where you think casino can lose are made so that they never lose. The rules of the game ensure that. And doesn’t matter if the player wins or loses, at the end of the day the casino always wins. Unfortunately it took me a long time to realise that there was nothing wrong with the results I was getting in the simulations. I was meant to lose; I just had to reduce the loss. Once we were done with this in a couple of weeks, we got back to our previous task.
The next week I tried to combine more than one indicator, something that a trader usually does. Mostly, one would look at as many indicators as possible before one takes a decision. But doing this for thousands of stocks would take an insane amount of time or traders. However, if one can instruct a computer how to handle them correctly, one can trade at a higher frequency and for many stocks. The best multiple indicators for combining would be those based on completely different principles - things that do not have a common source of change. And if both of their predictions agree, then we would have a much stronger indication of an apparent trend. That being said, I really began to appreciate what our eyes can analyze in one glance because programming for just two indicators together was a pain. Only in the last week was I successfully able to combine two to make a useful formula.
In the end, I think I met most of the goals of the internship. I needed to make atleast 4 good formulae which I did, found out for myself that Bangalore IS expensive and got to know fellows from iit-d. It was a great learning experience.
)
Recent Comments