Pitcher Win Values Explained: Part One
Since we released the Win Values for hitters here on the site, one of the main questions was when we were going to add them for pitchers. The answer: today. If you go to a pitcher’s player page here on FanGraphs, you’ll see the newly added Value section down at the bottom.
For example, here’s Johan Santana‘s Win Values for the past five years:
2004: +7.6 wins
2005: +7.2 wins
2006: +7.1 wins
2007: +3.4 wins
2008: +4.6 wins
During his stretch of dominance with the Twins, he was consistently amazing. He took a step back in his last year in Minnesota though, and while he rebounded somewhat this year, he hasn’t been the same elite guy that he was in his prime during the last two years. Still very good, certainly, as a +4.6 win pitcher is among the best in the game, but not quite the guy he was from 2004 to 2006.
Other fun pitchers to look at: Brad Lidge (+3 wins from a closer in ’08 – quite the addition for Philly), Barry Zito (the Giants should have seen this coming), and Ben Sheets (seriously, someone should give this guy some money).
So, now, for the obvious question – how on earth did we come up with these things?
Starting tomorrow, we’ll do a week long series explaining the calculations behind pitcher win values and the questions that arose during the process. They’re far more complicated that hitter win values, honestly – there’s issues of run environments, leverage, and context that had to be accounted for, and in many cases, the decisions of how to handle these things aren’t cut and dried. So, over the next few days, we’ll dig into those issues and talk about how we arrived at the values we did, and what the positives and negatives of those decisions are.
It appears the dollar values may be incorrect for pitchers. Perhaps you didn’t account for inflation in the same way you did on position players..it looks like all the dollars for pitchers are in 2002 dollars. Or is there some other explanation why Sabathia and Chipper Jones would have the same Value Wins in 2008 (7.6), but dollar values of $19.7 and $34+ million, respectively?
Good catch – David will get that fixed shortly.
Fixed!
Is the reason that win values are only shown from 2002 on also relating to the dollar values that have been previously calculated? Or is there certain information used in the run calculations that are not available prior to 2002? But this is great, another fun thing to play around with on Fangraphs
Yep, there’s certain information that I have only going back to 2002, mainly for the batter win values. UZR only going back to 2002 is the main reason.
We could do pitcher values going back to 1974 right now, and possibly 1954 later on, but there is a leverage index component in the pitcher values which makes going back before that not possible.
Not possible for now. If Retrosheet is able to get all of the play-by-plays then you guys go and get the LI’s.
First impressions -
Man, Roy Halladay is grossly underpaid.
It’s strange living in a world where Javy Vasquez is worth more than Jake Peavy, but I’m sure grateful that such a world exists.
I’m sure you’ll get lots of questions about the use of FIP in this.
I’m personally more interested in what you’re using for park factors.
How do you guys adjust for a pitcher having a huge impact on his own run environment?
And any chance of putting a wRAA next to FIP for pitchers?
I wonder… do the pitchers receive the same amount spent per win that hitters do? If so, should that be the case? We’ve seen baseball teams more often grossly overpay for pitching then we have for hitting. In that regard, pitcher values should be subject to their own sample size.
We’ve seen that, but does it make any sense? A win is a win is a win, right?
Though, I suppose that would impact the market for pitching vs. the market for hitting, so perhaps pitchers simply can’t be had for the same $/win rate.
Then again, the actual salaries paid to players suggest that no player can be had for the actual $/win rate – it’s just that some are grossly underpaid (like Halladay) and some are grossly overpaid (like Jose Guillen).
We’d need to measure it to be sure.
Tim Lincecum is very good at baseball.
Am I reading it right that Javier Vazquez was worth more $ in 2008 than in 2007???
Why didn’t you guys use a more advanced stat like tRA? Sorry, but FIP is not by any stretch of the imagination the best pitching metric out there. So why did you stick with FIP?
Also, why use FIP when looking at past value? Zito was a bad value because of the runs allowed, not the FIP runs allowed.
I see FIP as sort of a middle ground between ERA and tRA. I think that works out nicely for a stat like this. Basically we’ll give a pitcher the benefit of the doubt on the LOB% and BABIP, but a home run problem (which is typically the main culprit for ERA), we’re not willing to ignore.
Last time I checked FIP and tRA were at least similar with about a .75 r^2, or something like that.
David, isn’t tRA a middle-ground between ERA and FIP? FIP assumes all balls in play are the same. tERA gives different credit/blame for liners and fly balls for example. Now, there may not be much of a year-to-hear correlation for LD%, but if a pitcher gave them up, one could argue that’s part of his value. Taken a step further, you’d use PZR, which is the pitcher’s half of UZR. With the UZR methodology, you know the expected run value of each batted ball. Assign that to each pitcher. (Fielders get the difference of the expected run value of the batted ball given up by the pitcher and the actual outcome of the batted ball.)
It’s my understanding that tRA regresses LD%, so players aren’t getting fully docked for their LD% in tRA and they’re not getting docked for HRs at all.
Anwyay, upon further research, the 2008 tRA to FIP r^2 for “qualified players” is about .82, so in practice they’re still pretty close.
Maybe middle ground isn’t quite right. There are differences on both ends, but tRA for qualified players doing a straight average is about .4-.5 points higher than FIP, which does match up with ERA.
Out of curiosity, has anyone done a study on tRA that shows it to be more useful than FIP/DIPS/whatever else is out there? I’d be curious to see one.
It’s a good point, and I agree, that FIP/tRA/xFIP/etc are strongly correlated. And are better measures than ERA.
Graham has done some preliminary work with tRA/ERA/FIP correlations and they’re promising for tRA. He hasn’t published it or finished it, yet, though.
Any possibility of PZR for the site, David? You already have the UZR methodology implemented…
Great work on this.
Two quick questions:
1)Any chance you will be adding a 2009 projection for value? Using either Marcel or BJames, or combination of both?
2)Is there a “leaderboard” function that we can sort through values by position, by team, etc?
Combining 1 and 2, it would be interesting to see projections for teams, both in terms of wins and $ value.