Use probability to make predictions

[christianbundy]

This PR makes two changes:

• A random prediction is made when m.f and m.d are equal (resolves #4)
• A probabilistic prediction is made when m.f and m.d are unequal (resolves #5)

It's now very improbable for "malicious" patterns like dddddffffffdffffddfffdfdfffdddff to hit 0%. I ran the new algorithm against the above pattern, and these were my results:

n	p
1	33%
2	29%
3	40%
4	37%
5	44%
6	40%
7	37%
8	44%
9	29%
10	40%
11	11%
12	40%
13	29%
14	37%
15	18%
16	33%
mean	34%

It definitely isn't perfect, but a mean of 34% is a step up from 0%. It seems to work well with npm test, but please let me know if you notice anything funky. Cheers!

[willy-b]

This is awesome. I was talking with someone about this early this AM on Twitter. Was going to do a PR but glad you beat me to it (I wasn't planning the nice table). Thanks for really improving the demo

[christianbundy]

@willy-b Hell yeah, glad you got something out of it.