Painting the picture: Just as an artist needs inspiration about what they want to paint, a Quant needs to have an idea about the effect they want to their algorithm to capture. Before an artist chooses which colors to use or what size of canvas to pick they need to have an idea of what they want to paint - the sky, a lake or a cloud of colored dots. Their idea and their capabilities determine which methods to use to realize their vision.
Finding a new vaccine: A painter always ends up with a result. The result may be poor or excellent, there will be a result you can hang on the wall. A Quant doesn’t always have something to show in the end. You can start with the best idea from the scientific literature, do everything 100% right when implementing it, and it still might not succeed on the markets. This comparable to the work of a scientist trying to find a new vaccine: They have an idea about what causes the disease, and an idea about the methods they could use to cure it. But even if they did everything 100% right they might end up without a remedy. If that happens, the best thing to do is to restart with a fresh perspective. This highlights the importance for Quants to have fun experimenting and trying new techniques.
Here are three practical tips for generating ideas
A Quant is an artist. Like every artist, Quants have their own sources of inspiration. This can be the most fun part of the process. You come up with an idea, a hypothesis. You try to find patterns in seemingly random data. Visualization can help a lot to here. Let’s take a look at this chart from the S&P 500 Index Future, taken from the Quantiacs markets page:
The first thing that comes to mind is cycles. There seems to be repetitive patterns of ups and downs in this time series: There are two 6 year cycles from 97-03 and 03-09 and smaller 6 and 12 month cycles of ups and downs on top of them.
Another noticeable effect is the different size and speed of the up and down moves. The market appears to be more volatile when it is falling and less volatile when it is rising.
But a Quant is also a scientist. So far these observations are just a hypothesis. Now it’s time to put them into a quantitative model and test them on the market data. The whole process follows the scientific method:
Create testable prediction
Design experiment to test prediction
Support or falsify hypothesis
Ask new questions --> (1.)
Your hypothesis won’t always be correct. That can be frustrating. The good news is it’s time to start over with the creative part.
2. Apply methods you already know
Another good way to start is to simply apply methods that you’ve already mastered on market data: Physicists might use particle tracking algorithms to predict market data, economists use modern portfolio theory to compute efficient portfolios, computer scientists use genetic algorithms and neural networks, and mathematicians might use Fourier transforms to detect repetitive cycles in market data. No matter what your background is, if you have manipulated data once try to apply the same method again on the market time series. Don’t be intimidated by the problem. Financial data are just time series. You don’t need to know everything about finance to succeed as a Quant.
3. Start simple
It doesn’t always need to be a very complex method. A model based on the trivial insight of what goes up must come down can already get you somewhere. The success of a trading strategy is not proportional to its complexity.
To sum it up: Find what’s inspiring you. Don’t be intimidated by the problem. Start simple. Apply the scientific method. If your idea doesn’t work, be creative and try a new one.