One More Bat
With spring training a few weeks away, we have a pretty good sense about most of the roster.
Johjima and Clement are going to split time behind the plate. Depending on how much Clement shows defensively, the M’s may or may not carry Jamie Burke as a third catcher.
Beltre is at third, Betancourt is at short, and Lopez is at second. Branyan is penciled in as the first baseman, and Chris Shelton will fight Mike Morse for the right to be his platoon partner. Reegie Corona is currently the reserve middle infielder/pinch runner, but the team is looking to upgrade and might bring in a better player to unseat him.
In the outfield, Gutierrez is the center fielder and Ichiro is in right. Endy Chavez and Wladimir Balentien are the current candidates for the LF job.
However, Zduriencik has made it clear that his preference would be to add one more good major league hitter, preferably a lefty, before the club gets to Arizona. With the other six spots essentially taken, the new player will have to come from the LF/1B/DH pool. Here are a few of the more popular options, as they currently stand, if we allocate approximately 2000 plate appearances and 2,900 defensive innings to those three positions.
Stand Pat
In this scenario, the M’s would likely run different platoons with Chavez and Balentien in LF, Branyan and Shelton/Morse at first, and a rotating wheel of DH’s that would look something like this:
LF PA Innings wOBA UZR Chavez 300 850 0.304 10 Wlad 300 600 0.313 -5 1B PA Innings wOBA UZR Branyan 400 850 0.337 -3 Shelton 275 600 0.335 3 DH PA Innings wOBA UZR Clement 300 0.338 Wlad 150 0.313 Branyan 100 0.337 Shelton 75 0.335 Johjima 50 0.304 Lopez 30 0.324 Beltre 10 0.338 Ichiro 10 0.338 Total 2000 2900 0.325 5
The wOBA projections come from Sean Smith’s CHONE projection system – I put in the defensive projections. The offense would be -8 runs over 2,000 PA, the defense is +5 runs, the position adjustment for these three is a total of -35 runs, and the replacement level adjustment is +67 runs. Add it all up, and you get +28 runs above replacement, or +2.8 wins. That’s the status quo.
Trade for Nick Swisher
Here’s my preferred alternative – trading for Nick Swisher and giving him all of the PA’s that would have gone to Wlad (he’d either go away in the Swisher deal or in a separate trade) plus some of the PA’s that would have gone to Chavez.
LF PA Innings wOBA UZR Swisher 600 1100 0.360 5 Chavez 150 350 0.304 4 1B PA Innings wOBA UZR Branyan 400 850 0.337 -3 Shelton 275 600 0.335 3 DH PA Innings wOBA UZR Clement 300 0.338 Branyan 100 0.337 Shelton 75 0.335 Johjima 50 0.304 Lopez 30 0.324 Beltre 10 0.338 Ichiro 10 0.338 Total 2000 2900 0.340 9
The new wOBA from these three positions is .340, which translates to a 26 run offensive increase. There’s also a 4 run defensive increase, so the group goes from 28 runs above replacement to 58 runs above replacement, or +5.8 wins. Using Nick Swisher to replace Wladimir Balentien and cut Endy Chavez’s at-bats gives the Mariners a +3.0 win surge for 2009.
The other fun thing about Swisher is that his experience as a 1B gives you some serious flexibility. In late game situations with the lead, you could move Swisher to first and use Endy as a defensive replacement in the outfield, maximizing his defensive innings without giving him that many at-bats. You could also use Swisher as Branyan’s platoon partner instead of Shelton/Morse. Lots of options with Swisher on the roster.
Sign Adam Dunn
For those of you who are salivating over Adam Dunn sitting out there without a real contract offer, here’s the same analysis, except we sub in Dunn for Swisher.
LF PA Innings wOBA UZR Dunn 500 1000 0.373 -12 Chavez 150 450 0.304 5 1B PA Innings wOBA UZR Branyan 400 850 0.337 -3 Shelton 275 600 0.335 3 DH PA Innings wOBA UZR Clement 300 0.338 Dunn 100 0.373 Branyan 100 0.337 Shelton 75 0.335 Johjima 50 0.304 Lopez 30 0.324 Beltre 10 0.338 Ichiro 10 0.338 Total 2000 2900 0.344 -7
Instead of a .340 wOBA and a +8 defense, the team gets a .344 wOBA and -7 defense. That’s a seven run offensive increase and a 15 run defensive decrease. Instead of being +5.8 wins, the team would get +5.0 wins from these three spots. That makes Dunn a +2.1 win increase over the status quo and a -0.8 win decrease over trading for Nick Swisher. Considering Swisher’s contract (3 years, $22 million with a club option that could make it 4 years and $31 million) and Dunn’s reported demands (4 years, $56 million), in order for Dunn to make sense, he’d have to either cut his asking price in half or the Yankees would have to be demanding the moon for Swisher. Neither of those seem likely. As long as Swisher’s a possibility, Dunn doesn’t make sense.
And finally…
Sign Ken Griffey Jr
It won’t help the team. I promise.
LF PA Innings wOBA UZR Chavez 300 600 0.304 8 Wlad 250 500 0.313 -4 Griffey 150 350 0.332 -4 1B PA Innings wOBA UZR Branyan 400 850 0.337 -3 Shelton 275 600 0.335 3 DH PA Innings wOBA UZR Griffey 250 0.332 Clement 150 0.338 Wlad 100 0.313 Shelton 50 0.335 Johjima 30 0.304 Lopez 25 0.324 Beltre 10 0.338 Ichiro 10 0.338 Total 2000 2900 0.326 0
The team would get an extra run on offense and lose 5 runs on defense compared to the status quo. Yep, that’s a downgrade. They’d be a half win better by not adding Griffey. Whatever money you’d spend on Griffey would be wasted, both in terms of improving the club on the field and in limiting the playing time of players with some career ahead of them past 2009. Toss in the opportunity cost of not acquiring a real hitter to fill the void, and it’s a move that doesn’t make any sense at all.
So, to sum this up:
Make no more moves, platoon Wlad/Chavez in LF: +2.8 wins
Trade for Swisher: +5.8 wins
Sign Dunn: +5.0 wins
Sign Griffey: +2.4 wins
One of these options is clearly superior to the rest of them. Nick Swisher please.
Not true. Jose Lopez’s CHONE and BJ wOBA projections are identical and CHONE projects a higher batting average.
Yes I just made that up because clearly I know sweet bugger all about projection systems. I have been caught out.
Wait no.
Thanks for the responses, Dave.
Re: wOBA, I must confess. If you have a list of numbers in the same column with the same label, but they actually represent different statistics, then I am confused. I guess part of my confusion is that projected SLG, OBP & OPS follow the same pattern as wOBA in Griffey’s case. Surely projected SLG, OBP and OPS mean the same things to CHONE & BJO.
I am also confused by the way you are defining environment. In most social sciences, environmental conditions are by definition consistent across co-located individuals (e.g., players on the same team). Environmental characteristics can interact with individual differences, but that just leads back to my original assertion that CHONE & BJ have different assumptions about the roles of individual characteristics.
Finally, I will say that I agree that I have made an untested claim that CHONE probably underestimates the performances of players that have been injured in the previous season. However, you have also made an unsubstantiated claims about the differences between BJ’s projections and CHONE if you have neither specific knowledge of Bill’s formula nor results from multiple regression estimates.
Re: wOBA, I must confess. If you have a list of numbers in the same column with the same label, but they actually represent different statistics, then I am confused.
They don’t represent different statistics. This is why I asked you to read the post I did about this at FanGraphs. I’ll try an example to see if it’s easier to understand, though.
Let’s say Bill James and Sean Smith sit down and decide to project the 2009 performances of 500 players using their respective systems. The results are like this.
400 players – BJO has higher wOBA
50 players – same wOBA for both
50 players – CHONE has higher wOBA
The average of all the BJO projections is clearly going to be higher. Let’s say that the average wOBA of the 500 players for BJO is .340, and for CHONE, it’s .330.
Therefore, according to BJO, a player with a .340 wOBA will be a league average hitter. However, in CHONE, .330 is league average, so a .340 wOBA would be a better than average hitter. In terms of relative value (which is all we really care about), .340 and .330 would be equal.
Because BJO projects so many more players to do better overall, the effect is simply to increase the projected run environment of baseball as a whole. Therefore, when you look at the projected wOBA for each player, you have to keep in mind the context of the projection system’s definition of league average.
That’s why I kept pointing you to wRAA. That was showing you that the difference in wOBA between the two systems wasn’t an actual difference in projecting the relative value of Ken Griffey Jr for 2009. The two systems are in total agreement – Griffey will be a barely above average hitter. They just have a different opinion of what average is.
This is the point we’ve been tyring to drive home to you, and it’s all laid out in the comments section of the fangraphs post I’ve linked to. This isn’t something we’ve ignored or been unaware of. I was the one who pointed out to everyone “hey, these projections have different baselines, act accordingly”.
CHONE and BJO are in agreement on Griffey. They both think he’s a league average hitter. Thus, any argument built on the premise that BJO has some kind of advantage over CHONE in terms of it’s Griffey projection is faulty just by the nature of the argument. I don’t need to do a multiple regression to disprove your theory.
Thanks again, Dave, for your patience. First, let me apologize for confusing my original assertion by opening up a can of worms about the CHONE-BJ differences. That is only indirectly related to my initial contention.
I do see your point about the BJ-CHONE difference with regard to Griffey. When you average everything out, the difference could be the difference in scale. That’s why I did not deny that BJ’s projections are generally higher. I would be interested in just how much of a general scaling difference there is between the 2 methods. The difference in Griffey’s CHONE & BJ wOBA projections is .22, not .10.
Still, if a scaling difference is the case, then that implies that all of BJ’s projections should be scaled down in a proportionally (or nominally) consistent manner. Yet, a difference in scaling cannot explain all of the variance in the projections. As you said, many of BJ’s projections are equal to or less than the CHONE projections. That means there are differences in the methods that go beyond scaling. I hypothesize that one of those differences is that BJ places greater weight on stats before the previous year. (Sorry, I don’t have a good database to prove or disprove that using a proper method.)
The real issue I raised about Griffey was whether the projections adequately explained variation due to unexpected injuries. Using multiple years of data certainly helps smooth over problems created by idiosyncrasies in any particular year. However, without some specific knowledge of particular injuries, no general statistical method can be completely accurate.
There are certainly examples of what I am saying. Randy Johnson (2003-2004), Varitek (2006-2007), Furcal (2005-2006), B. Giles (2007-2008), Lidge (2007-2008) are all guys who seemed to benefit from knee surgery. There are also guys who seemed to go the other direction after knee surgery (e.g., Berkman, Ortiz, Bay, Castillo, Mueller) so surgery could be, as you suggested, an excuse or a cause of other negative issues.
Either way, it seems to me that projections of previously injured players really need asterisks. Qualitative data (e.g. trustworthiness of the player) becomes much more important. This is one reason why the “make-up” argument can never be entirely dismissed. So in my mind, you have to ask yourself whether Griffey is intentionally or unintentionally misleading he public by using his knee as an excuse.
By the way, Dave, I encourage you to drop the term, “environment,” when describing the differences between CHONE & BJO. The term has a meaning in social science statistics that is different from the way you are using it. I encourage you to simply say that the mean of BJ’s projections is higher than the mean of CHONE. That is very different from saying BJ’s projections are ALWAYS higher than CHONE.
Yes. Also, I encourage you to stop using the term “offensive” as it sometimes means causing displeasure or breeching moral boundaries.
“Offensive” has two meanings. Environment does not.
Like many of the factual assertions you’ve made so far, that is incorrect.
Merriam-Webster Online has three!
Interestingly enough, your definition of environment isn’t specifically called out there. Must not be a website about social sciences.
Sabermetrics is a social science. If you don’t want to use the terminology of social science, then I guess that is up to you.
And by the way, the relevant definitions in Webster are consistent with the social science definition.
Let me know when you wear out the shovel you’re using to dig yourself down, I’ll have someone lower a new one to you.
I’ll admit, Merriam-Webster isn’t always that reliable.
I was looking for a specific definition of this word but couldn’t find it!
LOL… Webster is entirely consistent with the social science definition.
Cute, Jeff. Regardless of how it seems to you, I promise I am not trying to pointlessly whip up a debate.
It’s fine to say that BJO is higher on average than CHONE. However, when you use the term, environment, you imply that the differences are exclusively due to factors outside of the abilities and efforts of the hitters. Obviously, none of us know that because none of us have specific knowledge of Bill James’ formula.
I would be interested in just how much of a general scaling difference there is between the 2 methods. The difference in Griffey’s CHONE & BJ wOBA projections is .22, not .10.
Just find players with similar wOBA projections and similar playing time estimates, then look at their wRAA. That will give you an idea of the run per PA difference. With Griffey, we know the projections are no more than 3-4 runs apart, because the wRAA difference is just 1.1 runs and CHONE only projects 40 more at-bats than BJO.
As you said, many of BJ’s projections are equal to or less than the CHONE projections.
Many is probably an overstatement. I can find a handful, but certainly they are in the minority.
That means there are differences in the methods that go beyond scaling. I hypothesize that one of those differences is that BJ places greater weight on stats before the previous year. (Sorry, I don’t have a good database to prove or disprove that using a proper method.)
Sure – I’m certain there are minor differences in the projections themselves. However, you don’t need a database to prove or disprove your theory. Simply pick ~50 players randomly, go to their FanGraphs page, and copy down the projections for both CHONE and BJO. Then, calculate the actual differences in projections and find out if the players that BJO is higher on than CHONE (beyond the scaling difference) are indeed players whose performance was better in ’06/’07 than in ’08.
It might not prove your theory is entirely true, but at least then we’ll have some reason to think that there might be reason to look into this further.
However, without some specific knowledge of particular injuries, no general statistical method can be completely accurate.
No statistical method can be completely accurate even with specific knowledge of injuries. Perfection is not the measuring stick. The question here is whether reported injury information makes the projections better. You think it’s certain that it does – I think it’s just as likely that most of the injuries of Griffey’s variety (played through the problem, talked about it when season was over) won’t have any actual predictive value.
Just like with all the “so and so reported to camp in the best shape of their life” stuff, it’s better ignored than trying to be factored in. Those kinds of reports make projections less accurate, not more accurate, so they’re best left ignored. I think it’s quite possible that injuries of the Griffey variety are the same.
So in my mind, you have to ask yourself whether Griffey is intentionally or unintentionally misleading he public by using his knee as an excuse.
I don’t think that’s the question at all. I’m sure Griffey believes it’s part of the reason he had a bad year. After all, his knee hurt, and he had a bad year, so therefore, hurting knee must have caused bad year. People draw these conclusions all the time. Doesn’t make them true, though.
I’m not arguing that Griffey’s misleading anyone. I’m just saying that there’s no evidence to compel us to assume that a player who says he played hurt will then resume playing like his pre-injury self. Remember, he’s going to be 39 this year. Injury or no injury, the idea that he should be projected to hit like he did when he was 37 flies in the face of what we know about how player’s age. That kind of assumption requires evidence.
By the way, Dave, I encourage you to drop the term, “environment,†when describing the differences between CHONE & BJO. The term has a meaning in social science statistics that is different from the way you are using it.
I didn’t create the term run environment. In the statistical community of baseball analysts, run environment is widely understood as the way I’m using it. I’d encourage you to read more about them, starting here.
However, when you use the term, environment, you imply that the differences are exclusively due to factors outside of the abilities and efforts of the hitters.
Every year, BJO projects ridiculously high offensive levels. It’s fair to say that the reason for this is not that his projections are overly optimistic because he just thinks this current crop of players is that awesome, but because there’s a flaw in his formula that causes offensive levels to consistently be overstated.
That’s an environmental factor that affects every player in the projection and is not specific to the player themselves. Thus, the usage of the word fits the context.
Gosh, it seems like the folks over at Fangraphs such as Tom Tango, are also using the term “environment” in a way that you would also consider inconsistent with social science statistics (if you read Dave’s link and the comments under it). It’s almost as if it’s nomenclature that sabrmetricians can understand that’s specific to their field, but that can’t possibly be true, could it? I suppose it’s a pity that Jack Zdurencik hired someone with such imprecise understanding of the language of his craft.
Thanks, Dave. Two points –
1. I agree with you that Griffey might really believe his knee was an issue. That’s I why a made sure to say “intentionally or unintentionally misleading.”
2. In the first two links you provided about “run environment,” Tango is using the term, environment, in a manner that is consistent with the social science definition. I didn’t check the other links. Notice how in both postings, he is discussing the use of proportional linear weights. In the second posting, he discusses ways to fragment the effects of the run environment, but he is still talking about the use of proportional weights to control for environmental factors.
Now look at the differences in the BJ & CHONE projection across players. There is a noticeable difference in the mean so, as I previously suggested, part of BJ’s “flaw” could be that he has put the wrong weight on the influences of the environment. So it is probably correct to say that that the BJ-CHONE difference is partly due to the environment. However, there is also a lot of variation across individual players’ BJ-CHONE differentials… even across players on the same team. Those differences must be attributed to individual differences, not environmental differences (unless Bill is taking an extraordinary amount time to measure each player’s different environmental characteristics, which is unlikely). So there is more to the story than just the influences of different assumptions about the run environment.
Saying things are so doesn’t make them so, as you have consistently and generously shown us all.
That’s why I clearly described the definitional difference in the subsequent message. I encourage you to discuss the substance of the definition if you are interested.
I’ve read 2 articles by Tango about run environment. He does not use run environment to explain individual differences in either article in the way that Dave used projected offensive environment to explain variation among individual players’ wOBA projections. So as far as I can tel, Tango’s definition is consistent with mine. Maybe you can show me a specific blog entry where Tango uses run environment to explain individual differences, but I have not seen it yet.
Also, if you could refrain from using the term “player” in the future as it can mean a person who pursues and engages in as many sexual encounters as possible outside of monogamous relationships, that would help too.
DMZ, I think we should note that the term “run” refers to the act of scoring in baseball, as opposed to the mode of locomotion where both feet leave the ground during strides. This would no doubt help to keep the discussion clear for everyone who is following it.
Perhaps we should also clarify that the term “term” in this case refers to a name used to describe something, and not the assigned time in office for an elected position.
Wow, Crack anaysis AND high comedy. This blog is unreal.
Understanding what Dave means by environment is important because it helps explain why he thinks variation between individual players exists. The fact is that BJ’s projection might not be a little higher because of environmental factors (e.g., park factors, pitching quality). It might be because BJ is simply putting different weights on multiple individual characteristics, including historical stats. If that is true, then the predictive value of just slapping a proportional weight on BJ’s projections to bring the mean into line with that of CHONE is lower than it would be otherwise.
Dave’s point about Griffey’s injury is well-taken. We don’t have any clear evidence about how such injuries affect the validity of predictions. However, it is totally reasonable to conclude that the injury creates a level of uncertainty about Griffey that does not exist with the other options. And none of this discussion addresses the assumptions that Dave’s original analysis makes about the way Griffey would be used if he was on the team. I guess we will just have to wait and see. Regardless, I have always agreed with Dave’s conclusion that Swisher is probably the best option assuming salaries are in line with most expectations. I doubt we will see Griffey as an M.
When you say crack, you should be more specific, as this can be confused with ass cleavage.
…or the stuff you smoke when you come up with your jokes…
However, it is totally reasonable to conclude that the injury creates a level of uncertainty about Griffey that does not exist with the other options.
It’s also just as reasonable to stick with the negative hypothesis (as Dave has, that there’s no predictive value in special pleading Griffey’s 2008 injury into evaluating his performance projection for 2009 upwards) until such time as someone proves your assertion.
And none of this discussion addresses the assumptions that Dave’s original analysis makes about the way Griffey would be used if he was on the team.
You know, nothing’s stopping you from doing your OWN analysis. I’ll go out on a limb and say that at best, if you do a bunch of wishful thinking where you throw away 2008, you still end with a relative ranking of something like this:
Swisher>Dunn>Raul>Griffey
(Just eyeballing the OPS’s, that seems to be how it would work, and Dunn, Raul and Griffey are all varying shades of awful in the OF.)
You’re still talking about someone who has zero defensive value and can’t hit LHP (he hasn’t since 2005)- even IF we are generous with his 2009 projected performance. That’s simply not a valuable package, given that Russell Branyan can do everything Griffey can at this point in his career (and realistically, that’s what you would use Griffey for on a good team- 200-300 PAs off your bench).
Thanks for the comments, eponymous.
It’s also just as reasonable to stick with the negative hypothesis (as Dave has, that there’s no predictive value in special pleading Griffey’s 2008 injury into evaluating his performance projection for 2009 upwards) until such time as someone proves your assertion.
The injury issue might not change the nominal projection for Griffey, but it does significantly lower the certainty about that projection.
You know, nothing’s stopping you from doing your OWN analysis.
I wish. Time and a proper database would help.
Swisher>Dunn>Raul>Griffey
I generally agree with your analysis… although you can’t really separate the projections from the costs of acquiring each of those players.
Where I doubt Dave’s analysis is the suggestion that Griffey would make the team worse… especially if he was used exclusively at DH for 400-450 PAs against righties. To the best of my knowledge, the CHONE and BJ projections assume KG would be hitting against both lefties and rightes like he did in 2008. But your point about Branyan is well taken. Despite the nature of Safeco, I tend to think this team could use a good right handed bat more than a left-handed bat.
So how does Abreu compare? Looks like he is out for a one year deal as well. You run his numbers?
I CAN NOT shake the “bring Griffey back!” feeling BTW… no matter HOW much evidence I see (and knowingly suspend believe). This is why I will never be a good SABR guy.
Thankfully this new managment team looks to be saving me from myself.
Dave what were the lgwOBA and wOBA scale that you used for this analysis?