I noted that Gutierrez and Ichiro had disproportionally higher CLUTCH ratings than the other guys. Given that 'clutch' is generally considered not to be a repeatable skill... doesn't that indicate that Ichiro and Guti were in essence 'luckier' than the other guys with top WPAs? That perhaps their contributions should be taken with a larger grain of salt than the other guys, since their production benefited from more or less perfect timing (again, if we agree that this is not a repeatable skill)?
Obviously, you can't compare Guti last year to this year. But I examined Ichiro going a couple years back. His clutch is ALWAYS in the 0.4-0.6 range... excepting 2004, when it was flat zero. The guy had a component to his game this year that, while not under his control, greatly inflated his WPA. (I think. I could totally be misinterpreting the relationship between these things). He's always on the first page... but 2001 and 2003 were the only other seasons he ever cracked 1.00 (just barely in both). A 1.7ish level just doesn't seem repeatable... much less Guti's 2.44
Now, if the point is just to sit back and acknowledge that these guys were FREAKING AWESOME in 2009... no argument here! =) And it's cool to see that on Guti's side as well... I wouldn't have expected him to be in that category this year. But if we try to extend that data forward, I think we might run in to trouble. Am I anywhere near the mark?
Dr. Naka noticed that Ichiro had more Win Probability Added than any other American League player.
.............
1. For those who just joined us -- and many readers aren't heavily into sabermetrics -- you get (say) 0.4 wins added to your total if you hit a home run with two men on in the 9th inning with the score tied. Your team's chances to win that game went from 50% to 90%*.
You can do this all year long ... Joe Shlabotnik struck out leading off the 5th in a 6-3 game, lowering his team's chances from 78% to 77%.
This obviously refines "clutch hitting" and "RISP" which only track a player's hitting with men on base, or 7th inning +, etc. WPA tracks him with 2 runners on in the middle innings in 1-run games, etc etc.
..............
2. There's a certain amount of luck involved, since the WPA leaders' hits (mostly) happened to occur when they mattered most.
But take a look at the AL leaderboards and you'll notice it isn't Jack Hannahan and Willie Bloomquist in the top 10.
It's Ichiro, Bay, Teixeria, ARod, Mauer, etc.
WAR does in fact answer one important question: "Which batters actually did the most to push their teams into victories."
Notice that a .350 AVG doesn't necessarily tell you whether a batter helped his team win any real games. It just tells you that he probably would have. WPA tells you how many wins he actually pushed his team towards.
................
3. Ichiro wasn't necessarily the best or most-talented hitter in 2009. But he did help his team more than any other hitter, including ARod.
...................
4. Notice that Ichiro's 4.33 wins added were 53% more than the #10 hitter's WPA. Not 53% more than the last-place hitter. Not 53% more than an average hitter. Ichiro had 53% more WPA than Adam Lind, who himself was superb, was one of the ten most-productive hitters in the league.
....................
5. Leadoff hitters get more plate appearances.
Sabermetricians generally feel that hitters shouldn't get credit for the fact that they get more AB's. But I disagree. The practical fact is that a slow player, like Ryan Howard, will not bat leadoff in today's game, and therefore will help his team 50* fewer times than Ichiro will.
This isn't a discussion about skill level. It's a discussion about who changes the standings. If a leadoff hitter provides a 10% bonus in quantity, those are real baseball events.
......................
6. For Franklin Gutierrez fanatics... kudos. Gutierrez was #5 in the American League -- effectively a more productive hitter than Joe Mauer was.
Franklin Gutierrez had a whale of a lot to do with the Mariners' 85 wins in 2009, that's for sure.
Cheers,
Dr D
......
More interesting photos from Strobist at http://www.flickr.com/photos/davidhobby/
Comments
Gutierrez homered in a 1-0 game in the bottom of the 8th inning, changing the score from 1-0 (odds of a Mariner victory this late in the game: 14%) to 3-1 (odds of the Mariners winning this late in the game: 96%). In one swing, Gutierrez produced 82% of a win. He could have struck out in 40 consecutive at bats thereafter and he'd probably still have a positive WPA.
While I'm no major supporter of WPA, I do believe that there are potential uses for the approach that 'might' be able to shed some light in areas not yet plumbed.
In general I agree that WPA is effectively useless in a predictive way, (other than it tends to capture the best hitters as best and worst hitters as worst - just not as well as other advanced stats).
I think the assessment "Boy, did so-and-so do great for us this year," is exactly the right frame of mind when seeing a particularly awesome WPA. In a sense, WPA is the *ONLY* stat that in some way actually captures "the double". There's some value there, I believe -- though I think at present WPA is mostly misused.
That said - I just had a brief exchange with the patron saint of WPA, (Pirata), posing the idea that WPA might be a potential tool for examining Arm ratings for OFs. You might uncover that some (most?) OFs generate their assists when trailing by 5 runs.
In point of fact -- WPA might be a methodology to utilize in examining defensive value. If you are constantly in 1-run games, then the value of every defensive play, (good or bad), is sky high. And when up by 8 runs, it's to be expected fielders might slack off and allow some extra hits. Maybe Jeter's historically poor range values are 'tainted', because he was often playing when leading by 3-5 runs.
It might reveal nothing. It might show the same lack of predictability that offensive WPA tends to show. But, there's always the possibility that something new might be revealed.
What Sandy said.
Certainly there is some luck involved, but WPA is like RBI. You don't get 147 RBI without being very good, and you don't get 147 RBI if you are a 'soft' player who gets a lot of your production after games are already decided.
If I told you that Mike Carp was going to have 158 RBI next year, you could complain about luck imbedded in the stat, but you would also know that Mike Carp was going to be very good, and that he was going to hit when it mattered.
>think the assessment "Boy, did so-and-so do great for us this year," is exactly the right frame of mind when seeing a particularly awesome WPA. In a sense, WPA is the *ONLY* stat that in some way actually captures "the double". There's some value there, I believe -- though I think at present WPA is mostly misused.
Every batter has his herotic moments.
Some has more some less.
If you think that WPA measures only these herotic moments I think you are wrong. It measures the bad moments too.
Late inning close or tied game 1out runner at 1st and 3rd.
Ichiro and Guti seldom GIDP here. Beltre does.
These are skills I think. And it will effect WPA and can be measured.
And the first five guys on the list were Pujols, ARod, Ichiro, Teixeira and Dunn, I'd start using the stat. :- )
Any stat that collects all the best players in baseball is telling me SOMEthing.
Some stats tell more than others. We get to thinking that only the "best" stats should be discussed. But every statistic contains information.
...I'll discuss WPA...no problemo there. I *hate* the way some at MC uyse WPA though...I see people trying to use it to project the 2010 Mariner WPA values...using it as the only thing they look at in judging who had value and who didn't...basing their award-giving on only WPA...driving their whole statistical world view with the WPA engine...
It's dangerous and it has to STOP.
If folks don't realize the problemos with using WPA to predict the next season's WPA, sure...
Wondering idly. If you took the >0.00 WPA's from 2008 and from 2009 and weeded the players who weren't in the league in either year... what do you think the CORR would be? 0.50 maybe?
If you're trying to predict ERA, there are better stats for that than ERA itself. But as y'know, ERA still contains information that FIP doesn't...
And foud that WPA cross-correlated with itself (from one year to the next) with an R value of about 0.48 I believe. Pearson correlations use the rank (best, worst, or some discrete rank in between) for each pool being correlated together rather than the actual magnitude data.
A cross-correlation of 0.48 is significant statistically for sure - comparable to OPS...but pretty weak compared to some of the component statistics we use for making real projections.
Not a bad guess, that .50, what say :- )