Correlation = a coefficient that ranges from 0 to 1.
1 = totally correlated
0 = no correlation at all
... and anything in between (like 0.35 and 0.68) is the degree of correlation.
Another meaningful number is r^2 (r x r) too.
Matt's example is one where r = 1, like he says ... another would have been taking NYG +2.5 and NYG +121 on the ML last night ... anyone who wins the moneyline bet would also win on the pointspread, so you would want to parlay that if you could ... but you can't.
There are other examples of correlated parlays, and one is in MLB ... where you take the home team -1.5 and the under ... that is because if a home team is winning big, then they probably won't bat in the 9th inning ... and for a 9 inning game with a total of 9, that means 0.5 runs is scored every half inning and you only have 8.5 innings played. BetJM and Pinny used to not allow you to parlay run-lines with totals for that reason (not sure if they still don't) ... and that is why they don't post run-lines in Cubs games until they post the totals ... 1.5 runs is worth a lot more in a game with a total of 7 than it is in a game with a total of 11.5.
There are other correlated parlays in football, like Alabama -34.5 and over 41 ... but while the correlation isn't 0.99 here, it is still very high.
Ganchrow could explain correlated football parlays to you a lot better than I could ... and I hope he does.