Motivation

Surfing for information on algorithmic trading, I discovered the strange words of Foreign eXchange (FX) Expert Advisors (EA), usually built on a Metatrader platform. One that picked my attention was Robominer:

• It was one of the very few EAs that seemed to work using forward testing at the PipCop website, including on a live account (that is an account with real money, after graduating from a demo account). PipCop was in turn one of the few sites that seem serious about EA testing.
  
• The author, Jerry Brunet, provided a mathematical proof of the system efficiency, more a sketch of a proof than anything else, but way better than the usual screaming marketing material extolling the virtues of other EAs.
  

I am an engineer, and not afraid of mathematics, and I had started looking into quantitative finance, but was not much acquainted with FX at the time. While I could glimpse the principles between the system, I wanted to somehow reverse engineer the description to convince myself that this was not a bogus claim.

This led into a rather detailed analysis of grid systems and related aspects, including historical simulations of grid based strategies, and some empirical relations allowing to predict which currency pairs are best candidates for grid systems. It turned out that the best pairs derived from this study were also the recommended pairs for Robominer: NZDAUD and EURCHF, validating my own approach.

The backtest results are mostly in line with the claimed yearly doubling goal discussed (based on 6% per month growth rate). This is both incredibly high for any financial instrument and low compared to the claims made by other vendors.

The results were good enough to convince me that this was a valid money investment. At about the same time, I stumbled on Collective2 that seemed an other possible alley to get some money back for the time spent on this by selling the results of it.

And this led to this blog that explains the system, and also shows how open source and open data allows anybody to perform on his own some moderately serious number crunching with practical results.