The software has already paid for itself several times over

Posted by on Apr 24, 2013 in | No Comments

I read some white papers about Statistical Arbitrage with cointegration method, and was keen to include it into my currency trading. I found there were tutorials for it using Matlab and R. While I’m familiar with Matlab, I was looking for a software that is ready to use, turn-key solution.

A couple weeks later a new software company showed up at ForexFactory forum. My email record indicated that I have been using the Arb-Maker Software since August 2012. Back then the software allows me to search for cointegrated pairs, once found, I watched for its z-score (spread) to come near 2 standard deviation level, then I’d open the trades at MT4 platform manually, let it run overnight to several days, while occasionally checking the z-score charts. Today, the AM software has feature for Automatic trading and MT4 integration, a dream come true!

Before the MT4 integration, I have to use a trial version of IQfeed to get the data feed (extra fee), now I’m able to mine the tick data from MT4 directly. Back then, sometimes I have to wait for a week to find a good pair using the Fixed Beta method (Beta is the hedge ratio of weighting/lot size between the two legs), now I’m using the DESP Beta method which puts me in a new trade about every couple days. I found that this “market neutral” trading method is preferable over my other directional tech trades (S/R, MAs, and Fib).

And I guess this is also why they are not teaching technical analysis academically, when a 61.8 Fib level holds, it’s just because the crowd thinks so (self-fulfilling prophecy, if you will). With statistical arbitrage method, it gives me a feeling that there is academic reason for my trade entry/exit. For example, I can set the level of statistical confidence in my search criteria (99 or 95%) so that it will filter out those pairs with autocorrelation above/below these confidence levels as it indicates that the pairs are non-random. Random data is good because in probability theory it is approximately normally distributed. Normal distribution is said to be predictable (the bell-shaped curve). So when a pair has spread its legs for about 2 standard deviation, it’s time for me to enter. And it feels good, because I know over time, the spread 95% of a time will return to mean (according to normal bell-shaped curve theory).

This method makes me avoid overtrading. I now have a system and I trade the system with specific entry/exit criteria. I also noticed that when I trade a cointegrated pair, the Net P/L shown would only drift around +/- 10-100 pips, for days, ranging like sine-wave approximation. There are times when I got stopped out, and thanks to the automatic trading feature, I can set my exit strategy when the spread has drifted to more than 3 standard deviation. I don’t use regular stop loss, I have another EA that handle global emergency stop loss (10% of account), not having to worry about stop hunting, etc. Before taking actual live trade, I backtested the pair with my entry/exit strategy. Once I found a particular pair to be profitable in the backtest, I went ahead and switch on the automatic trading feature. Now the AM software is linked up to MT4 order execution, and places and manages my trade according to the backtested strategy.

So far I only trade 1:1 (no leverage) per leg and the software already paid for itself several times over. The software has developed quite a bit since its inception, I guess partly due to the openness of the developer to client suggestions/wish lists. It is a pleasure communicating with one of the partners, Rawdon, he’s very helpful, responsive, and open to suggestions. I look forward to future developments of this software.