NOTE: Apologies to all. In the comments, jco properly pointed out that my numbers for run differential in the original posting of this were off. Sorry, I don't know exactly how that happened. My previous conclusions about Darwin Barney and Lucas Harrell leading the NL in NWAR have been removed, as they no longer apply. The extremes are certainly toned down, but I hope you latch on the premise of my proposal for normalize war described below.
In my last FanPost regarding WAR, I mention that the variation among replacement team wins, or run differentials, indicates that the stat might need some tweaking if it is to really represent what it claims to, which is the amount of wins (runs) a player adds to a team over a replacement. As I've discussed, WAR actually is calculated from runs, both saved and created either offensively with the bat, in the field with the glove, or with pitching, using a conversion that approximates at 10 runs equals 1 win. With that in mind, a team's total WAR should reflect the difference in run differential (Runs - Runs Against) from a replacement team.
What is a replacement team's run differential? Fangraphs specifies a replacement team as a team that wins 43 games, so the run differential would be calculated as (43-81)*10 = -380. Baseball Reference uses 52 wins for a replacement team, making its run differential -290.
Using these values and comparing with actual results, we can come up with an expected fWAR and bWAR for each team. As an example, the Angels run differential was +68. This is 448 runs above a Fangraphs replacement team, giving us an Expected fWAR of 44.8 (or Expected fWAR = 38 + RD/10). Their Expected bWAR is 35.8 (= 29 + RD/10).
I propose that in order to make WAR match reality, we normalize it by dividing a team's Expected WAR by its actual total WAR to get that team's Normalizing WAR Factor (NWF). We then calculate each player's WAR in the normal way, but then multiply his WAR by the normalizing factor. If we do this, the team's total WAR will equal 1/10 the deviation from a replacement team's run differential, or, the number of wins above a replacement team (using the same 10 runs = 1 win ratio).
The table below shows both the Fangraphs and BR NWF for each team.
|Team||Total fWAR||Total bWAR||R-RA||Exp. fWAR||Exp. bWAR||fNWF||bNWF|
The Blue Jays have the highest fNWF; their 23.4 total fWAR accounted for a -68 run differential, which is 312 runs over a replacement team, meaning they got more for their fWAR than expected. The smallest extreme is the Indians bNWF. They had a 13.4 total bWAR, but were only 112 runs over a replacement team.
So let's apply this to the players. I've recharted the top 60 batters and pitchers in each system, but for brevity, I will only post the leaders for the BR system. First, let's look at batters.
You'll be happy to know that Mike Trout still leads both leagues in normalized WAR (NWAR) in both systems, although knocked down a bit. Andrew McCutchen now has a higher bNWAR than than Buster Posey, as does Ryan Braun.
Let's take a closer look at Andrew McCutchen's new bNWAR to how it applies to real life. His 7 bWAR is 33% of his team's 21 Total bWAR. The Pirates had a run differential of -23, which is 267 runs over a replacement team's run differential. Therefore, he was responsible for 33% of the 267 runs, or 89 runs, which matches his 8.9 bNWAR.
For pitchers, let's look at fWAR. There is one particularly interesting result there. Justin Verlander no longer leads the majors in fNWAR. His 6.8 fWAR is reduced to 6.46 fNWAR, while Felix Hernandez leaps over him a considerable amount, posting a 7.86 fNWAR. A similar examination as above shows that he is responsible for 22.6% of his team's 348 runs over replacement team.
|Zack Greinke||Two Teams||---||5.1||---|
|Chris Sale||White Sox||1.218||4.9||5.97|
|Jake Peavy||White Sox||1.218||4.4||5.36|
With my edits in place and accurate data being considered, this is probably not as controversial. However, I think my conclusions for the most part remain.
All I'm doing is matching the stats with reality. I don't think bNWAR is necessarily a good way of comparing players with each other in the sense of who is a better player when considering trades and such. That is, it might not say much about what a player would have done for your team, or what that player may do in the future. But I think it can be argued that these values give you a true sense of a player's value to his team last season, in the way MVP is classically considered. The fact of the matter is that WAR doesn't actually relate to real-life team runs, unless it is normalized in the way I've described.