So now lets try something different and test the simple moving averages double crossover not on an stock index, but something different, the foreign exchange rate of the euro and US dollar. Because trading on the EUR/USD on forex can also be an way to earn money ;)
So lets look at figure 1, where we used the data back from 4 years (1000 tradings days) to run this backtest. Again we used timewindows from 1 to 1000 days for the simple moving averages. Because we are dealing with an exchange rate here, it is of course not given that there will a longterm ‘positive’ movement, so you could invest sensefully with Buy & Hold. But in theory you can earn a good yield by timing the investment - what the SMAs can help you with.
Figure 2 shows a topview of the forex EUR/USD performance. In this backtest the exchange rate seems to undergo different fluctuations then the DAX or SP500 did. Because here you dont see a golden triangle with the middle simple moving average windows, but instead the more longer timewindow combination seem to have good results. But only until you look at the ‘Buy & Hold’ performance which is about 4.3% p.a. - about as same as the best simple moving average result!
If we would delete all yields which are below this benchmark, the graphic would be completley empty. Considering a smaller timescale - better yield for trading forex EUR/USD?
Because a longterm investment doesnt make much sense in forex trading, but shortterm trading should be the better solution. So we do some backtesting with a shorter timescale of 250 trading days (a normal year), the “buy and hold” yield was 6.5% p.a.
Figure 3 shows an 3D view of the yield performance for the shorter scale, and figure 4 shows the topview. The colorscale is different on the graphics because the very negative results (-50%) were not drawn in the 3D graphic, because the Z axis would be so high scaled that the differences in the positive domain couldnt be seen.
You see in this backtest that middle sized SMAs can be very bad, they form a deep ditch in the middle. But directly near that there is a height where longer sized simple moving averages perform very good, the best pair is 534⁄532 with 12.4% p.a.
Figure 5 shows the simple moving average pairs which have a performance higher then 6.5%, you see the success triangle is this time in the region of the high timewindows.