Thanks, I like statistics and I agree with all of that.

The danger is in assuming a statistic is saying something it isn't saying. For example, many people took Clinton's approval rating of high 60s or 70s or whatever it was a year before the primary as evidence of very high vote number she would have during the general election 2 years later. As we know, that didn't happen. It's common for non-election period approval ratings to have low to no prediction power of the coming election.