Tim Lincecum in 2010

After Tim Lincecum got off to a rough start, I wrote that perhaps 2008 would be his peak season. No shame in that. Only 20 individual pitching seasons had accumulated at least 7 WAR, one of those was Lincecum’s 2008. Naturally, he responded by pitching even better in 2009, and finished with a WAR above 8. So, with the odds on my side, I feel comfortable taking a shot at this again. What are the odds that Lincecum makes me look like a moron again? Well, here’s a list of every pitcher season to post at least 7 WAR during a single season in our database that data available for the season afterwards:

Curt Schilling 2002
Randy Johnson 2002
Pedro Martinez 2002
Roy Halladay 2002
Roy Halladay 2003
Pedro Martinez 2003
Mark Prior 2003
Esteban Loaiza 2003
Randy Johnson 2004
Ben Sheets 2004
Johan Santana 2004
Curt Schilling 2004
Johan Santana 2005
Johan Santana 2006
Brandon Webb 2006
CC Sabathia 2007
CC Sabathia 2008
Tim Lincecum 2008
Roy Halladay 2008
Cliff Lee 2008

That list is full of some all-time great pitchers pitching some all-time great seasons. The average WAR of those spectacular seasons is 7.8, with a spread of 2.9 wins (9.7 being the max, 7 the min). In the year after, the average fell to 5.5 wins with a spread of 6.8 wins (8.2 by Lincecum being the highest, and 1.4 by Esteban Loiza being the lowest). Roughly two-thirds of those pitchers failed to post 7+ WAR in consecutive seasons, which shows just how rare Lincecum’s feat was, and only one pitcher (Johan Santana) pulled the feat in three straight years.

Assuming my SQL database skills are still in the best shape of their life (meaning replacement level as a ceiling) there were 660 individual seasons of 25+ starts between 2002 and 2008. We had 20 seasons with 7+ WAR, 6 with 8+ WAR, and 4 with 8.3+ WAR. By those numbers, and completely ignoring that Lincecum is clearly better than most, if not nearly all of those pitchers, he has something like a half of a percent’s shot at reaching 8.3 WAR next year. That’s without figuring in that he just had an 8 win season, and that nobody who recorded at least 8 WAR in a single season has replicated that success (although Pedro Martinez did reach 7.9, which is basically the same thing).

Some encountered more bad luck, others injuries, and yes, Lincecum is a fantastic arm and talent, but so were most of those guys. So yeah, I feel comfortable betting against Lincecum being even better this season. And hey, if he does make me look bad once more, then we’re all winners by witnessing The Freak at his best.

For extra credit I’d suggest reading this Book Blog thread. It’s about that Stephen Strasburg cat and projecting his ERA. Even if you don’t understand most of the comments at first read, and heavens I sure didn’t, you should still absorb something worthwhile.





47 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
kingstephanos
14 years ago

I understand the uniqueness of Lincecum’s WAR values, but to use 2002-present as a marker for this evaluation uses an extreme SSS.

With recent history as a comparison, Pedro Martinez’s 1998-2002 years are empirically equal to or greater than any modern era pitcher (including Lincecum), WAR not withstanding.

toastyfire
14 years ago
Reply to  kingstephanos

Someone should try to calcuate what Maddux’s WAR would be from 92-98 when his highest FIP was 2.85 in those years.

Steven Ellingson
14 years ago
Reply to  toastyfire

http://www.baseballprojection.com

He used a different metric than FIP (something like ERA with an adjustment for defense), but it should give you a good picture.

92 – 98

8.4
6.2
7.8
8.8
6.3
7.3
6.1

Alex
14 years ago
Reply to  toastyfire

And the most amazing thing about those numbers is that the 7.8 and 8.8 came in strike shortened seasons. He was on pace for over 10 WAR in each of those seasons (based on a 162 game schedule).

B
14 years ago
Reply to  toastyfire

“And the most amazing thing about those numbers is that the 7.8 and 8.8 came in strike shortened seasons. He was on pace for over 10 WAR in each of those seasons (based on a 162 game schedule).”

Not sure how those are calculated, but wouldn’t “replacement level” have to be set lower for shortened seasons, too, meaning that kind of calculation wouldn’t be correct?

alex
14 years ago
Reply to  toastyfire

My understanding is that those WAR calculations are done by comparing a run rate (I believe its era at that site) to a replacement level and then using innings pitched to determine WAR. Since maddux would have pitched significantly more innings (he only started 25 games in 94 and pitched over 200 innings), his WAR would have been significantly higher.

Honestly I’m not quite sure why you would think there would need to be an adjustment to replacement level in a strike shortened season. Its not like the level of replacement players changes significantly. There are just fewer games, and therefore fewer wins above replacement to be had.

B
14 years ago
Reply to  toastyfire

“Honestly I’m not quite sure why you would think there would need to be an adjustment to replacement level in a strike shortened season. Its not like the level of replacement players changes significantly. There are just fewer games, and therefore fewer wins above replacement to be had.”

Because there are fewer games for replacement players to play, too. If you had a team of all replacement players, they would win fewer than the ~46 or so games a replacement team wins in 162 games, because they would play less games (the ’94 Braves played 114 games, which turns ~46 replacement wins into ~32 replacement wins, for instance). WAR is created relative to a baseline set by other players, and if the entire population of players loses PT….well, both the nominator and denominator are affected by that change. I don’t know the specifics, but I would suggest we need to know whether that’s accounted for or not before we know if it’s appropriate to make that claim about Maddux.

Alex
14 years ago
Reply to  toastyfire

My understanding has always been that replacement level is a rate, not a counting number. Thus, WAR is simply going to be a function of the difference between a particular player’s rate and the replacement rate multiplied for the amount of playing time said player had.

Look at your 1994 example. You assume that a replacement level team would win ~46 games, so in 1994 they would win ~32 because of fewer games. All you did there was take the original winning percentage and multiply it by the new number of games. Same thing with WAR. The season being shortened doesn’t fundamentally change the replacement player, so he performs at the same rate. The same is true of the player we are looking at. Therefore the only difference is playing time, so we just take the difference in the two rates and multiply out for the proper playing time.

BTW, what fraction are you talking about when you bring up the numerator (not nominator as you said) and denominator changing?