Television audiences for the College Football Playoff semifinals declined sharply in Atlanta, although not as sharply as in many other places.

Nielsen ratings for Thursday’s two semifinal games dropped an average of 15 percent from last season’s semis in the Atlanta TV market, while the ratings plunged a drastic 36 percent across the nation.

The biggest reason for the fall: The semifinals were played on New Year’s Eve this season and on New Year’s Day last season.

Clemson’s 37-17 victory over Oklahoma in Thursday’s first semifinal, played in the Orange Bowl, posted a 17.8 rating in the Atlanta market, down 16 percent from 21.3 for the first semifinal last season.

Then Alabama’s 38-0 victory over Michigan State in Thursday’s second semifinal, played in the Cotton Bowl, drew a 20.2 rating in Atlanta, down 14 percent from 23.4 for the corresponding semifinal last season.

This season, Atlanta had the sixth highest rating among the nation’s 56 metered markets for the Orange Bowl and the fifth highest for the Cotton Bowl. Birmingham had the highest rating for both semifinals and Greenville, S.C., the second highest for both.

Nationally, the Orange and Cotton bowls posted ratings of 9.7 and 9.9, respectively, down from 15.5 and 15.3 for last season’s two semifinals, which were played in the Rose and Sugar bowls in the inaugural season of the four-team playoff.

This season’s ratings weren’t helped by the one-sided scores, although one of last season’s semifinals was a rout, too.

The College Football Playoff rotates the semifinals among six bowls, playing them on New Year’s Day only in the seasons the Rose and Sugar bowls are the hosts. Next season’s semis are scheduled for New Year’s Eve, which will fall on a Saturday, in the Peach and Fiesta bowls.

Outside the playoff, Georgia’s victory over Penn State in the TaxSlayer Bowl on Saturday posted a 13.3 rating on Atlanta TV, down from 15.7 for the Bulldogs’ Belk Bowl victory over Louisville the previous season.

The rating is the percentage of TV households tuned in on average.