“So out of 87 games that went to overtime from 2010-2017, the results were:
45 first possession won
37 first possession lost
5 still tied after overtime
I.e. there was a first possession advantage in current NFL overtime during that time period.”
That data shows 45 wins and 42 no wins. T test produces a p-value of .482579. The result is not significant at p < .05 That “first possession advantage” is not large enough to be significant. In other words an apparent advantage of that modest magnitude it could easily occur by random chance.
Since ties are valid results and are not wins, throwing out the ties and comparing only wins and losses is not correct. But if we try that anyway (45 wins vs 37 losses) just to see what it looks like, we still do not get a significant result ( p = .451092).