Finally I have made up my mind to become a quant trader. Becoming a quant trader looks like a daunting task at the moment. But I am confident if I work with determination, I will become a good quant trader. You might be wondering what the heck you were a naked trader now you want to become a quant trader is. Watch the video below that explains what a quant trader is!
I have been trading naked for a while now. But naked trading is just naked trading. I need to convert my naked trading strategies into quantitative algorithms. I am working on using a support vector machine to give me buy/sell signals. I will train the support vector machine to use my naked trading strategies. I have also developed a few neural networks that are giving good price predictions. So I think this quant trading stuff does work and if I can augment quantitative trading with naked trading, I can drastically reduce the risk per trade and increase the winrate substantially.
The journey started when I started developing an indicator for trading binary options. I learned MQL4. I have some backgroung in programming. In my college days I had learned FORTRAN. So learning MQL4 was not difficult. Once I learned MQL4 I realized it is only good for opening and closing a trade. It cannot predict the price. For that you need to use statistical algorithms that can take the data and then make forecasts.
Learning R Language
I came across a thread on Forex Factory when I person had developed an R indicator. R is a powerful statistical software that is open source and is being widely used for data analysis now a days. That R indicator was just using auto correlation function to build a model that it was then using to predict the closing price for the next 20 candles. After using that R indicator I realized it was painting and it was no good.
Making that R Indicator work forced me to learn R language. The dll and mqh files provided were not working on the latest MT4 Build 600. I had to work for a few weeks then I found one nice day on the MQL4 forum that a person had solved that problem. He had posted an update version of the dll and the mqh files that worked on MT4 Build 600.
Then I found out that the auto correlation function that R indicator was using was no good. I had to dig deep into the new machine learning algorithms. I then started learning how to model time series with ARCH and GARCH processes but again found that this was no good. Ultimately I found that neural networks are being using in lots of things now a days that includes image recognition and speech recognition. Neural Networks are good at classification. There are many types of neural networks. Now a days people are talking about Deep learning Networks. Then I found that Support Vector Machines are also used for regression and classification. So it was the start of the journey.
Learning C++ and Python
I learned C++ language. I have now a working knowledge of C++ language. Then I found out the R is not an efficient language. Python is rapidly replacing R and becoming the language of choice for the quants. So I have started learning Python also. I am now a days learning Python and hope that in the next few days I will start developing my own code in Python.
This is what I plan to do. First I test my idea in R. Use R language to test my hypothesis and develop the neural network or the Support Vector Machine. Once I have developed that and I find that I am getting good results. I develop the same hypothesis in Python. Python is 15 times faster than R. When it comes to algorithmic trading, time is of essence. Training a neural network in R is taking around 30-60 minutes which is too much if you want to trade on the intraday timeframes. It doesn’t matter if the neural network training taking 60 minutes when you are trying to make the predictions on the weekly or the daily timeframe as you have ample of time. But if you want to make the predictions on H4, H1 and M30 then you need calculations to finish within 5-10 minutes. For that I plan to use C++ as it is the fast language that exits as of now.
So this is how is going to work. First test the idea in R. If the idea succeeds and gives you good results, switch to Python and see if you get fast execution. If Python helps in making the execution fast good. If not then I switch to C++. Right now I am working on developing a indicator that works on Weekly, Daily, H4, H1 and M30 and makes predictions for the next 5 candles. As I said I got good results with the neural network and I am working on it. I will keep you posted on my progress so stay tuned.
Applying Machine Learning To Trading
You must have figured out by now if you have to be a very good coder if you want to succeed as a quant trader. This is true. Machine learning is a new subject that is being rapidly developed. Machine learning requires knowledge of mathematics, statistics, artificial intelligence and information technology. You need to be good at all these subjects if you really want to become good at machine learning. I see myself working hard for the next few months. I have learned C++ and R. Now I am working on Python and applying that to machine learning.