If you prefer, you can play with binomial distributions to calculate P(side X wins with N votes && side X wins with N-1 votes) [1]. Using the normal approximation and ignoring the normalization prefactor (an overestimate) exp[(n(p-0.5)^2 / 1(np(1-p))] odds of having your vote matter. For p=0.501 (i.e., voters are split 50.1%, 49.9%), that works out to about 10^{-87}. For p=0.51, double floats are incapable of representing a number that small.
I stand by my statement. Your vote doesn't matter.
[1] Conceptually, this is slightly different from computing the odds that you cast the deciding vote, but I realize it is more useful from a decision theoretic perspective.
[0] http://redbluerichpoor.com/blog/2008/10/what-is-the-probabil...
[1] http://www.stat.columbia.edu/~gelman/research/published/prob...
(By "significantly", I mean "increases the odds of your vote mattering to a number greater than, say, 10^{-86}. )
The odds that the election would be decided by a single vote have not been considered.
This assumption doesn't make sense