Comment by lesuorac

It's only a 75% success rate if you include the superbowls prior to the claim being made. Obviously the guy looked at the results of the superbowl prior to making his claim.

It drops down to 67% [1] when you look at only after 1978 and since 2000 its 10/23 (43%) which really implies its headed towards 50% (i.e. uncorrelated). The big difference here is that you can get outliers in your data when you have a small sample size (i.e. 50) but as the number of football games approaches 5000 those outliers go away.

But in support of your point: https://www.tylervigen.com/spurious-correlations

[1]: https://en.wikipedia.org/wiki/Super_Bowl_indicator