Reading Sherman's book, I got to wondering how many analysts would choose 1996 as Rivera's best season. It's chief rivals are 2005 and 1999. It would probably be easiest to sort through the cases of the closer seasons, and then compare to the set-up year in '96.
We can easily rule out 2002, when injuries limited him to 46 innings; his debut season of 1995, when he was a starter with a 5.51 ERA; 2000, with his career-high relief ERA of 2.85; and both 1997 and 2006, which were fine seasons but boasting no advantages over his very best seasons.
2005 was the year Rivera had his best conventional rate stats: a career low 1.38 ERA (and a career high ERA+ of 323). He allowed his smallest percentage of baserunners (.235 on base percentage against), and was just as effective in denying power (hitters slugged .230 off him, missing his career best by only .002). He also had his best ever-rate of stranding inherited runners, allowing only 2 of 18 to score. He threw 78.3 innings, went 7-4, and saved 43 against 4 blown saves. In an unspectacular postseason, he allowed 1 run in 3 innings with two saves against the Angels.
In 1999, he was nearly as sharp in 69 innings, allowing a .239 on base percentage and a .237 slugging percentage. He went 4-3 with a 1.83 ERA with 45 saves against 4 blown saves. Of 27 runners inherited, he stranded 22 and permitted 5 to score. He also was the most efficient he had ever been, notching an out every 4.8 pitches, a career best. At this point, you would still have to put 2005 first. But then you get to the postseason. 1999 is in his top three: 12.3 scoreless innings, with 2 wins and 6 saves (and the World Series MVP Award). Counting their postseason and regular seasons stats together, 2005's ERA advantage shrinks to 1.44 to 1.55. And there's more: Rivera allowed only 1 unearned run in 1999, whereas he allowed 6 in 2005. If charged for all runs, inherited, earned, and unearned, in the regular and postseason, Mo was less scored upon in 1999 (2.21, 81.3 innings) than 2005 (2.32, 81.3 innings).
Rivera allowed an even smaller average of total runs the year before, in 1998. In 61.3 regular season innings, he allowed only 13 runs, all earned, and allowed only four of 24 inherited runners to score. In 13.3 postseason innings, he allowed only 6 hits, 2 walks, and no runs at all. In 74.6 combined innings, he allowed 2.05 runs per game, a career best. He also his highest percentage of good results: 3 wins, 36 saves, and 6 postseason saves, against no losses and only 5 regular season blown saves. Overall, though, it is hard to pick 1998 as his best year. The results may have been the cleanest, but he was hit harder (.270 on base percentage, .309 slugging percentage) than in 1999, and he didn't pitch as much.
2001, when he saved 50 games and got more batters out (236) than in any other year besides 1996, was his career high in win shares. I'm a big win shares fan. Asking them to pick a closer's best season, however, is among the last things win shares should be asked to do. It was a really good year, but he wasn't as dominating as in other seasons, and given that, I wouldn't pick the year he got beat by the Diamondbacks in the World Series as the best of his career.
2003 rates a mention. After going 5-2 with 40 saves and a 1.66 ERA in the regular season, he had arguably his best postseason: 16 innings pitched, 7 hits, 1 run, 0 walks, and 14 strikeouts. This includes his epic three-inning stint against Boston in ALCS game seven. On the downside, though, he was asked to deal with a career high 35 inherited runners, and almost half of them, 17, scored. In terms of his own runs, the regular season and postseason combined gave him about as good a mark as 2005, 1998, and 1999, but I wouldn't want to pick a year when he had such a hard time with inherited runners. The postseason record is a temptation, but he shut out postseason opponents in 1999 anyway. I think 1999 is still the top candidate.
There is a case for 2004 based on the sheer volume of good results. In the regular season, he had 4 wins and 53 saves against 2 losses and only 4 blown saves. In the postseason, he pitched well, gaining a win and 2 saves while allowing 1 earned run in 12.7 innings. It was also his best fielding season, making outs on a career-high 41 plays against 1 error. But although he pitched well, he had three blown saves in the postseason, one when he came back to get the win against Minnesota, and two against Boston in the ALCS. The second there was an impossible save, but the first was the disaster that let Boston back in the series. He also blew the emotional regular season game when the Red Sox attacked Alex Rodriguez. For me, it is too many black marks to pick the season over 1999.
1996 was Mo's best strikeout year by far: 130 Ks. He held one of the hardest hitting leagues in history to a career low .228 slugging percentage. Compared to 1999, he walked more hitters (.258 on base percentage vs. .239 in 1999), and hence had a higher ERA 2.09 to 1.91. He did worse with inherited runners, stranding 16 and letting 6 score. His postseason was excellent, 1 run in 14.3 innings. Overall, slightly sharper in 1999, but then there is the big ticket item: in 1996, Rivera pitched 107.7 innings, getting 316 batters out, 112 more than in 1999. And that's a lot mo' Mo.
The fact that he was not yet the closer is not a big thing to me. I have no doubt whatsoever that Rivera's 7th and 8th innings could have been 8ths and 9ths if he had been used that way. We've seen enough of this guy to know that. I also suspect that the added situational importance of his 1999 closer innings is not that large, and not at all larger than the 112 extra outs. Baseball-Reference.com (one of the world's greatest baseball websites), has now merged in data from Retrosheet.org (the other one) to provide access to an amazing amount of game information that it would have taken a huge amount of work to compile before. With a few clicks of a button, you can see what the score was every time Mariano Rivera entered a game, and what it was when he left. These summaries of Mo's usage pattern are from B-R:
Rivera made 61 appearances in 1996, and 66 in 1999. In 1996, he had more games in which he pitched more than an innings, and in 1999, he had more games in which he pitched on consecutive days. As the setup man, he pitched in 17 games when the score was tied or the Yankees were behind, and as the closer, he pitched in only 8 of those. But overall, in both seasons, he made 37 appearances when the score of the game was within two runs. I can't see that situational importance in 1999 outweighs the extra quality pitching in 1996.
No, the objection to choosing 1996 as Rivera's best season isn't the role, it's the style. It is hard to choose as Rivera's best year one before he was regularly throwing the cut fastball, his signature pitch. Reliance on the cutter turned him from a high strikeout pitcher into one who would overpower hitters right on their bats, often splintering them in the process. From 1996 to 1999, he cut almost a full pitch off his pitches per out rate, and he became a different pitcher.
So who would you want, the guy who made that amazing first impression, blowing everyone away, oblivious to the whole mid-90s batting storm? Or the guy who carried you through the last decade, making short work of the last three or four outs, picking hitters apart with a few cutters? In the end, I can't choosethey're my two favorite players.