I’m experiencing some brain fog this morning courtesy of Roger Federer, who came back from two sets down to win his quarterfinal match at the U.S. Open last night against Gael Monfils. The match ended around midnight, and for Fed fans like myself it was hard to turn off the adrenaline and go to sleep. Roger saved two match points in the fourth set, and those two points were the turning point in the match.

It was Federer’s ninth career comeback to win in five sets after being down two sets to nothing.

I’m not exactly sure what the odds are in that situation, but I’m guessing that it must be around a 5% probability because in virtually every case there’s a very good reason why the player who has won the first two sets is likely to win the match. To state the obvious, the player with the two set lead is likely to be 1) the better player and/or 2) playing better on that particular day and 3) the beneficiary of confidence and momentum that accrues to the player in the lead. Of course, the better the player, the less they conform to probabilities suffered by lesser players, but any way you slice it, being down 0-2 is not where you want to be in a tennis match.

So it is interesting to look at Federer’s win last night in the light of those conditions. Over the entirety of Federer and Monfils’s careers, Federer is clearly the better player. Federer has won 81% of his matches over his entire career (he’s 33) while Monfils, who just turned 28, has won 63%. But Monfils was certainly playing better last night than Federer for two sets, and arguably playing better than Federer for just shy of four sets, until he caved in after failing to convert the two match points. But where Federer won the match was in his ability to expropriate what should have been Monfil’s confidence and momentum.

There was an article on the ESPN website this morning that reviewed each one of Federer’s comebacks from an 0-2 position. (There have now been nine over the course of his career.) Discounting the first one, where Roger was losing in the fifth set when his opponent had to retire, presumably from an injury, the interesting thing to me is the average number of games that Federer’s opponents have won in the fifth set. Over the remaining eight matches in our sample, the average number of games won by Federer’s opponents in the fifth set (when Federer wins, of course) is two-and-a-half games. In other words, the fifth set has typically been a rout, an example of Federer grand larceny when he has stolen all the confidence and momentum available on court.

That confidence and momentum stuff is valuable. It’s no wonder that the great players know how to steal it.