Calculate and plot the exponential moving averages: Go back to Chapter 3, Get into Terms with Commonly Used Functions if needed and review the exponential moving average algorithm. Calculate and plot the exponential moving averages of 9, 12 and 15 periods.

emas = []
for i in range(9, 18, 3):
weights = numpy.exp(numpy.linspace(-1., 0., i))
weights /= weights.sum()
ema = numpy.convolve(weights, close)[i-1:-i+1]
idx = (i - 6)/3
ax.plot(dates[i-1:], ema, lw=idx, label="EMA(%s)" % (i))
data = numpy.column_stack((dates[i-1:], ema))
emas.append(numpy.rec.fromrecords(
data, names=["dates", "ema"]))

Notice that the plot function call needs a label for the legend. We stored the moving averages in record arrays for the next step.

Find the crossover points: Let's find the crossover points of the first two moving averages.

first = emas[0]["ema"].flatten()
second = emas[1]["ema"].flatten()
bools = numpy.abs(first[-len(second):] - second)/second < 0.0001
xpoints = numpy.compress(bools, emas[1])

Annotate the crossover points: Now that we have the crossover points annotate them with arrows. Make sure that the annotation text is slightly away from the crossover points.

for xpoint in xpoints:
ax.annotate('x', xy=xpoint, textcoords='offset points',
xytext=(-50, 30),
arrowprops=dict(arrowstyle="->"))

Add a legend: Add a legend and let Matplotlib decide where to put it.

leg = ax.legend(loc='best', fancybox=True)

Make the legend transparent: Make the legend transparent by setting the alpha channel value.

leg.get_frame().set_alpha(0.5)