Here's how the party holding the convention did the election prior to the convention and during the election in which the nominee ran. This is what statisticians would call "not revealing at all."
On average, those holding conventions in Cleveland saw a dip of 1.6 percent in the margin between the two candidates from the prior cycle. But that's an average of plus 15 and minus 18. So: a level of variability that it is silly to extrapolate from.
Extrapolating from that, the Republicans could expect to lose Ohio by 4.6 points, instead of the 3 they lost Ohio by in 2012.
Using the same "math" for the Democrats, they can expect a 5.3 point boost in margin. Meaning that Hillary Rodham Clinton (or whichever of the hundreds of viable Democrats makes it through the party's scathing primary process) could expect to win Pennsylvania by nearly 11 points. Wow!
Conventions are good for raising money. They are poor for raising excitement. They are pointless as predictors of who will win elections. The end.