shows what happens when you mix large and small sample sizes

flaw.py

import matplotlib.pyplot as plt
from pylab import savefig
import numpy as np
from numpy.random import rand, randint

def flipsome(n):
	flip = lambda: randint(0,2)
	return sum(flip() for _ in range(n)) / float(n)

interval = 1000
begin, end = -450000, 0
length = end-begin
intervals = length/interval

a = np.linspace(begin,end,intervals+1)

for i in range(1,21):
	b = [flipsome(interval) for _ in range(intervals)]
	b.append(flipsome(65))  # final data point
	fig, ax = plt.subplots()
	fig.patch.set_visible(False)
	ax.axis('off')
	plt.plot(a,b)
	savefig('figure %s.png' % i)