[BobC]

Originally Posted by nat

"That last one in some ways makes WAR pointless as it's typically figured.
If it was done based on comparing not to the league overall, but to other pitchers in a similar position - like only including first or first and second starters- it would probably be much lower."

People have looked into starter matchups, and have found that once you're more than a few weeks into the season, aces don't match up against aces with any regularity any more. (Nor 2nd starters with 2nd starters, etc.) Different teams have different days off, different pitchers get rested at different times, some teams have a rookie they want to see, so they slot him into the rotation for a couple weeks, and so on. Once any of those things happen, the aces don't match up with the aces any more. Once you're more than a few weeks into the season, who the opposing pitcher is, is mostly just random.

As for a recency bias in WAR for modern pitchers: totally not. Here's the all-time top 10 in pitching WAR (baseball-reference version):

Cy Young
Walter Johnson
Roger Clemens
Kid Nichols
Pete Alexander
Lefty Grove
Tom Seaver
Greg Maddox
Randy Johnson
Christy Matthewson

Give or take a player here or there, that's the list basically anyone will give you of the greatest pitchers of all time. By my estimation we've got players who peaked in the:

1890s
1910s
1990s
1890s
1920s
1930s
1970s
1990s
2000s
1910s

The next ten feature Tim Keefe, Eddie Plank, and John Clarkson, and Pud Galvin is 21st. The advantage that the old guys had is that they pitched tons of innings, and they're getting credit for all of those innings that they pitched. Basically, if a modern pitcher is pitching five innings, and an old guy was pitching nine, at the same rate of performance, the modern player is going to accumulate only 5/9ths the WAR. (Pitchers do, on average, pitch better in shorter stints, but as the list above indicates, not enough to make up for the lower workload.)

The reason that WAR allows cross-era comparisons is that it compares players to how well they performed against their contemporaries, and you can compare those comparisons against each other. For example, newly elected HOFer Jim Kaat's best season (1975) was worth 7.7 wins above replacement; this means that if you dropped him into an American League team in 1975, you could expect them to win about 8 games more than they would have had he not been on the team. This is a pretty good match for Tim Lincecum's 2008. What that means is that you should expect Kaat's pitching in 1975 to win as many games for a team as Lincecum's pitching would have won for a team in 2008.

That is, you're comparing Kaat against other pitchers in 1975, and Lincecum against other pitchers in 2008. You find that in their respective contexts they were each worth about 8 wins to a team. And looking at that, you can see that, in their respective contexts, they were about equally valuable.

It doesn't tell you what would happen if you put Kaat in a time machine and sent him to 2008. You really can't know that with any certainty, and that's the kind of "cross era comparison" that WAR can't (and doesn't try) to do. When people talk about modern players being so much better than the old guys, this is what they have in mind. In Honus Wagner's day players often didn't have proper nutrition, they certainly didn't have kinesiologists plotting out optimum workout routines, and medical care didn't compare. Wagner was probably as naturally talented as any modern player, but if you put an adult Wagner in a time machine and told him to suit up for the Pirates, he wouldn't be a superstar, just because he wouldn't have the advantages of modern training and nutrition. That's what people are talking about when they say the old guys weren't as good. But that's not very interesting - it's just a remark about how science and technology have advanced, it doesn't really tell you anything about baseball players. So it's really not a weakness of WAR that it doesn't allow THAT kind of cross era comparison. The kind it DOES allow - about how much a player meant to the league he played in - is important and interesting, from a baseball perspective.

It is all part of what makes the debate fun because you can only do so much with pure statistics, and it is literally impossible to effectively measure and account for all the variables that are out there. I've just seen how modern statistics for pitchers are based, or at least seem to be skewed, on today's starting pitchers not going for complete games anymore, and disregarding wins as an important statistic. Some of today's statisticians then claim that modern pitchers are all so much better than pitchers from earlier eras because today they're all bigger, stronger, and throw faster than pitchers ever used to. I've questioned though how these modern pitchers would fare if they had to pitch complete games like their predecessors, and how well they would hold up with the added stress, wear, and tear their bodies and arms would face pitching like the old-timers did. But statisticians can't really account for that in their formulas, nor do they have a way to give more credit to old-time pitchers for their wins that they were certainly way more responsible for by pitching complete games, than their modern counterparts who may only pitch 5-6 innings before turning it over to their bullpen and defense to get them the W.

If you hadn't seen the Greatest Lefthander of All Time thread from a couple months ago, go check it out and you'll see how some some statistical experts were blatantly saying how pitchers like Grove and Spahn would barely be just a little above average compared to today's pitchers. So their point was that WAR was not a good cross-era measure at all, and Spahn being the the all-time winning-est lefty in history, by a wide margin despite losing three prime years in the service, basically didn't mean anything. That is where I'm kind of coming from.