This is the formula I use. It might look a bit complicated but once you've used it a few times it's pretty easy.

Predicting the number of losing runs of a given length, against a given strike rate:

X = N x S^2 x (1 - S)^L,

where
X = number of losing runs of length L
N = number of results
S = strike rate
^ = to the power of
L = length of losing run you wish to investigate.

nb. S is in decimal format eg. s/r of 75% would be 0.75

I find if you make N 10,000 it's pretty accurate.