We have more and more data in our database, in order to produce more or less relevant statistics.

We have registered 1228 matches, with their results and the rankings of the 2 players involved (ranking of the week of the tournament). We also know if it was a best of 3 or 5 sets match.

For this sample, we are able to determine how often the higher ranked player wins the match.

Best of 5 sets matches : in 74% of the matches the higher ranked player wins
508 matches (the 4 last Grand Slams, from Roland-Garros 2014 to the AO 2015)

Best of 3 sets matches : in 69% of the matches the higher ranked player wins
720 matches, mainly M1000, mainly hard court

It could be interesting to add several criteria : the conditions (hard, hard-indoor, grass, clay), the tournament category (250, 500, 1000, GS, Davis Cup, ...), or the gap between the players (50 ranks, 10 ranks, 5 ranks among the top 10, ...).
There are more "surprises" (the lowest ranked player defeats the higher ranked) on clay than on hard court, because there are more ATP tournaments on hard courts. It's easier to have a good ranking if you're better on hard court.
Including the tournament category will let us know if there are less "surprises" in the toughest "best of 3 sets" tournaments (M1000) than in the smallest tournaments (ATP 250).

I think this "higher-ranked-player-wins rate" will decrease if we consider the whole year. Because our sample does not consider enough small tournaments, and not enough different conditions.

With our database, in around 70% of the matches "the higher ranked player wins". The real rate will certainly be around 66 or 67% at the end of the season.

It could also be interesting to study this rate for the Davis Cup matches.