Let this be a classic lesson of trust, but verify. It does not matter if you spent $1 million hiring Russian programmers to cook up trading systems on hundreds of years of data from Global Financial Data if you are ultimately clueless about how to analyze the end results. To that end, I will present a chart of the model's buy/sell signals:
[align=center]
[/align]Looks great, right? Serendipity dictated that since I'm too po' to buy the expensive historical data from Global Financial Data, I was forced to use free historical data which only started in 1979 for all of the four currencies, not 1971. As a result, the performance of the model is absolutely terrible, earning 1.92% CAR and -73.06% MaxDD since 1979 and this is without transanction costs or earning interest while in cash. It also misses the huge gains in the current bull run because it is chopped.
But, that does not compute! How can the purveyor claim that you can earn around 20% annualized (non-leveraged) while being invested in gold with this model?
Answer: The out-performance is extremely front-loaded at the beginning with one upleg before gold became legal to own in 1975 and then during the final upleg of the gold bubble after the 50% peak-to-trough drawdown (which incidentally, would all but wipe you out if you were stupid enough to use a double leverage gold ETF which conveniently lets the purveyor highlight the around 40% annualized gain and conveniently ignore the maximum adverse excursion or maximum drawdown).
My apologies to everyone!
