I’ve found some new stats in the press, again contradicting the ATP stats.
This boxscore was in the Western Daily Press:
PETE SAMPRAS (USA) v ANDRE AGASSI (USA)
17 Aces 5
5 Double Faults 6
60 First Serves In (%) 43
88 1st serve points won (%) 74
49 2nd serve points won (%) 50
43 Unreturned Serves 36
8 Forehand Winners 9
4 Backhand Winners 6
13 Volley Winners 6
1 Passing Winners 5
4 Return Winners 4
61 Returns In (%) 51
27/50 Net Points Won of Approaches 10/12
3/9 Break Points Won 0/4
I got the same numbers of Unreturned Serves, if I include aces along with return errors.
I got the same numbers of Return Winners.
It looks like the Passing Winners are non-return passes, and my own count agrees for Sampras: 1 such pass in the whole match. I have Agassi at 7, though, rather than 5 (it might depend on whether you count passes only as ground strokes hit near the back of the court; I count any groundies even if they're hit from near the net).
For Agassi I got 10 rather than 9 Forehand Winners, and 7 rather than 6 Backhand Winners.
I got 18 volley/overhead winners for Sampras (including 5 overheads), rather than the 13 Volley Winners in the box. And for Agassi I got 5 volley/overhead winners (including 2 overheads) rather than the 6 in the box.
The net stats are lower than NBC’s. Two games before the end, NBC already had Sampras at 41 of 62, Agassi at 9 of 13.
The break points line up with the ATP stats, but again it looks like the ATP figures for Total Points Won are wrong.
By my count Sampras made 31 return errors, Agassi 27. So if I use the % of “Returns In” provided by the boxscore, it looks like Sampras made 49 successful returns, while Agassi made 28. Add all these numbers to the aces and doubles, and you've got 167 points played in the match, compared to 223 reported by the ATP.
167 total points is more realistic, because it would yield 5.4 points per game. The ATP’s figures would yield 7.2 per game, which is just too high considering how well both players were serving -- particularly Sampras.
(The only good way of calculating the actual number of points played is to do your own count, so 167 can only be treated as an approximation. But to get to 223 the boxscore figures for % of "Returns In" would have to be seriously off).
Finally, the boxscore has Sampras serving at only 60% rather than the 66% he has at the ATP. The success rate on first serve, for both players, is close to the ATP’s. But on second serve Agassi moves up to 50% rather than 45% at the ATP, and Sampras moves up to 49% rather than the wholly unrealistic figure of 39% that the ATP gives him.