OMG GET MANNY CASHMAN IS TEH SUCK!!!1!

We need Dunn too.

Finally, someone hears my pleas to make Melky a SS!

It’s possible.  But it requires the following:

Jeter continues to decline
Cano does not bounce back
Posada’s shoulder isn’t quite right (though he gets 500 PA?)
ARod doesn’t have his usual “odd year” performance
Tex has an adjustment year
Signs of real improvement from Nady were illusory
Swisher only gets 400 PAs, whereas Nady plays a full season in RF

Some of that will happen.  All of it, though?

As a BP non-subscriber, I guess my first question is, does PECOTA predict a larger downturn in overall offense than CAIRO and CHONE?  Is some of the discrepancy its assessment of game-wide trend, or is it straight-up hate?

Wow, Jeter and Cano are roughly equivalent to Gardner?  Ouch.

I think PECOTA’s numbers are heavily influenced by the aging curve without adjusting for outliers. My best guess is that this is what is keeping Rodriguez’s numbers down.

SG, can you check the following: have PECOTA consistently underestimated Rodriguez’s performances based on predictions and actual seasons? I would think that is the case.

I saw this yesterday, but I don’t think the conclusion that PECOTA hates the Yanks is correct. For example, it “hates” Boston even more. It has Boston’s highest projected VORP coming from Pedroia at around 32 and projects Ortiz to provide only around 28 VORP. It projects Bay for around 22. All told, in a similar number of play appearances, the Yanks’ starting lineup (or at least, a decent approximation) is about 30 - 40 runs of projected VORP better than Boston. The starting rotations are seen as being roughly equal.

In any case, my only point here really is that I do not think this pessimism is a Yankee exclusive problem.

In detail:
Varitek: 3.3 VORP
Posada: 9.3

Youkilis: 20.3
Teixeira: 36.2

Pedroia: 31.8
Cano: 20.7

Lowell: 10.6
A-Rod: 45.1

Lowrie: 14.8
Jeter: 25.4

Bay: 22.0
Damon: 20.9

Ellsbury: 11.6
Cabrera: 6.5
Gardner: 4.8

Drew: 12.7
Nady: 11.0
Swisher: 14.9

Ortiz: 28.7
Matsui: 12.3

As a BP non-subscriber, I guess my first question is, does PECOTA predict a larger downturn in overall offense than CAIRO and CHONE?

I thought of that as soon as I posted, and John touches on it.  I’ll see if I can figure out if they’re predicting a lower offensive environment and everything’s relative.

SG, can you check the following: have PECOTA consistently underestimated Rodriguez’s performances based on predictions and actual seasons? I would think that is the case.

Sure, I’ll take a look sometime today.

I saw this yesterday, but I don’t think the conclusion that PECOTA hates the Yanks is correct. For example, it “hates” Boston even more.

I think in general any projection system is going to seem to “hate” the team you’re a fan of.  As fans we are generally optimistic.  I got a lot of crap for the first set of CAIROs I ran on the Yankees, and compared to PECOTA they’re wildly optimistic.

Also, remember that these are the PECOTA weighted mean numbers. That means they include scenarios where, for example, A-Rod blows out a knee in spring training and is gone all year. You have to keep that in mind. If you ran PECOTA with the assumption that all the players would be healthy (which you shouldn’t), you’d likely get much better numbers across the board. I prefer to look at the PECOTA numbers once you can see the various percentiles for performance, especially the 50th percentile which I prefer to the weighted mean.

I think in general any projection system is going to seem to “hate” the team you’re a fan of.  As fans we are generally optimistic

Exactly. In particular, I think that we assume too much health for the players we like. The odds of one of the Yankees’ starting nine being gone for a few weeks or more due to an injury are pretty good, as they are for all teams.

What’s the kurtosis on these projections?  How much of the weighted average is dictated by these “catastrophic” events?

A tad OT, perhaps, but I’ve recently started reading the blog here on my phone, and I’m curious, SG…why do you post all your tables as images?  There’re perfectly good HTML ways of presenting tabular data, and having them as images really degrades the mobile browsing experience.  A lot of times, the phone will compress the image, and the text becomes kind of illegible.

There’re perfectly good HTML ways of presenting tabular data, and having them as images really degrades the mobile browsing experience.

That may be true in general, but it’s not true with the Expression Engine software we use for some reason.  I used to be able to post as tables without any issues, but it doesn’t work any more and I haven’t really had the time to look at why.  I’ll see if I can figure it out at some point.

Rodriguez’s PECOTAs vs. Actuals from 2006 through 2009:

pNN = projected, aNN = actual. 

Doesn’t look like he’s been consistently under-projected, but they’re projecting a big drop off this year for whatever reason.  In general, players decline by about 1/2 win per season in their 30s, but PECOTA’s saying Rodriguez is going to decline by 2 full wins.

Oh, maybe now I see. Are they using a 4-3-2-1 weighting scheme for last 4 years? That gives 40 percent weight to the last year. Perhaps that explains why their predictions show the same sine-curve like swings from one year to the next: Rodriguez had alternate good years. This, with an aging curve that predicts decline could explain the rate number sequence of 119, 111, 119 and 106.

Although, this year it seems like it is really predicting down. .508 SLG for A-Rod? I think he beats that handily.

From a comment on the baseballthinkfactory, seems relevant to the discussion.

DKDC Posted: February 02, 2009 at 12:26 PM (#3065718)

TM   Wins   Hit Rank   Start Rank   BP Rank    
NYA       95   19   1   1    
BOS       94   10   2   2    
TBA       87   12   7   8    
TOR       79   30   3   7    
BAL       76   11   29   16    
               
CLE       86   13   9   4    
MIN       80   20   21   3    
DET       78   21   17   14    
CHA       75   28   12   24    
KCA       73   29   14   23    
               
LAA       81   18   10   26    
SEA       74   26   16   28    
OAK       74   25   22   19    
TEX       72   23   26   15    
               
NYN       96   1   4   5    
ATL       89   3   8   12    
PHI       86   4   15   27    
WAS       80   7   18   30    
FLO       74   9   30   18    
               
SLN       87   5   13   10    
MIL       87   2   23   6    
CHN       84   8   19   9    
CIN       81   14   11   25    
HOU       72   22   28   22    
PIT       69   24   27   29    
               
ARI       85   17   5   11    
LAN       81   6   25   13    
COL       79   16   20   17    
SFN       78   27   6   21    
SDN       77   15   24   20

Looks like they are projecting the Yankees to have 5th best offense in the AL, with the best SP and BP. I´ll take that.

I don’t see how Boston, with the second best pitching and the best offense in the AL finishes second to the Yanks with the 5th best offense and the best pitching.
In real life I can’t imagine the delta between 1-2 in pitching will be greater than 1-5 in offense.

I would tend to agree, but since the delta in pitching are for the extreme points in ML, it could be larger than the delta in offense from the middle of the pack projections for both teams. Completly POOMA, I´d say the difference between the 10th and 19th best offense is around 40 runs, and the difference between the 1st and 2nd pitching is around 30 runs.

The Yankees are going to have only the 19th best offense in the majors?  I find that hard to believe.

Completly POOMA, I´d say the difference between the 10th and 19th best offense is around 40 runs, and the difference between the 1st and 2nd pitching is around 30 runs.

So you agree with me then?

Yes, I do.

I knew the method was very crude, but by rereading the results maybe we should just wait for the properly done simulations. The numbers are league adjusted, so the best offense relative to league does not equal best runs produced. I also think the defensive value is included in the offensive projections, since the defense supervorp belongs to the hitter. Finally, there is no way in hell that Texas is falling all the way down to the 26th place in “offense”.

May it just shows that PECOTA believes offense will be down across the board in the AL so perhaps, as John Lynch suggested, they don´t hate us all that much.

As long as we’re talking about mobile browsing a little today, an internal link to the bottom of the comments page will come in handy for mobile browsing during a game (when we’re most likely to want to refresh comments and look at the last few).

For reference, here’s how each player stacks up if you rank them by projected VORP according to position. Note that PECOTA lists each player according to last year’s team. I tried to correct for that with the AL ranks, but I make have screwed it up, so these are obviously just quick and dirty.

Posada (C): 25th MLB, too lazy to count AL rank; again, projection is low because of playing time concerns
Teixeira (1B): 7th MLB, 2nd AL
Cano (2B): 9th MLB, 4th AL
A-Rod (3B): 3rd MLB, 1st AL (remember when the AL had all the good SS?)
Jeter (SS): 8th MLB, 1st AL
Damon (LF): 11th MLB, 3-5th AL (depending on Manny and Dunn and other issues)
Cabrera(CF): 35th MLB, too lazy
Gardner (CF): 47th MLB, too lazy
Nady (RF): 26th, too lazy
Matsui (DH): 5th MLB, also 5th AL (shocking!)

Swisher was projected at 1B, which isn’t too useful since he doesn’t figure to play there much.

So, again, looking at those ranks, it’s clear that PECOTA does not hate the Yanks. Look at the projected AL rankings for the infield positions. I think if we were given the option to guarantee that we would have those rankings at the end of the year, we’d all take it in a heartbeat.

It’s more accurate to see that PECOTA is simply pessimistic about individual player performance in general. I have no idea if that comes from projecting less offense this year (doubtful) or simple from a more conservative projection algorithm (likely, imo).

And, because I care too much, I will note that Damon and Teixeira see a nice bump if you go by SuperVORP. Teixeira rises to 3rd in MLB, 1st in AL at 1B. Damon rises to 5th in MLB, 2nd or 3rd in the AL in LF, depending on where Manny signs.

For full disclosure (and to avoid any confirmation/optimism bias), Jeter falls according to SuperVORP, down to 12th in MLB and 3rd in the AL.

The basic point remains: PECOTA does not hate the Yanks (relative to other teams).

FYI, in 2008 the difference between 10th and 19th best offense was 54 runs.  In 2007 it was 63.
From first to second pitching was 38 in 2008 and 9 in 2007.

With regards to the BUI, how viable are Teixiera and Rodriguez as backup/emergency options at 3B/SS respectively?

how viable are Teixiera and Rodriguez as backup/emergency options at 3B/SS respectively?

My first inclination is to say, “Not at all.”  Thinking about it a little further, I’m still inclined to say, “Not at all.”  It doesn’t make a lot of sense to do that kind of musical positions thing just to get Jeter a day or two off to rest a little owie, or to move Teixeira across the diamond to sub for A-Rod under similar circumstances.  So I think somebody would have to miss a lot of time before it would even be considered.

Hit submit too soon.  I guess I could imagine Rodriguez and Teixeira going to Girardi and volunteering to switch positions like this after watching a month of Cody Ransom at SS.  But I would expect Girardi’s response to be to tell them that they haven’t played those positions for like five years, and it isn’t exactly like riding a bicycle.

Teixeira’s total 3B experience is 15 games in his rookie year.  He doesn’t even have much experience at the position at lower levels, since he made it to the majors so fast.

Changing the subject, Melky Cabrera is officially out of options.  I know we’ve speculated about him possibly sneaking through some loophole or other, but apparently that’s not the case.  So whatever Cashman and Girardi may be saying about an open competition, I’m guessing that barring a trade before the end of ST, Cabrera is on the opening day roster and Gardner is in SWB no matter how they perform.

If someone on the bench could play emergency 2nd, then the above would allow dispensing with a UI for in-game injuries, and the next day one calls up whoever from the minors.

Too many ifs, and probably a dumb idea anyway.

Moose on Torre.

Looks like they are projecting the Yankees to have 5th best offense in the AL, with the best SP and BP. I´ll take that.

Keep in mind that the marginal value of saving a run is greater that marginal value of scoring a run.

The difference is pretty marginal.

...eh, FWIW, PECOTA also has Chipper Jones posting the second-highest EQA at .342, out-VORPing A-Rod by 20 points.

“Keep in mind that the marginal value of saving a run is greater that marginal value of scoring a run.”

How so?  The pythag formula doesn’t care, and except for tails due to blowouts (but no, that should average out to zero too) I don’t see where this could be asymmetric unless good teams are adopting inefficient lineup construction strategies.

Make that, unless teams are adopting inefficient roster construction strategies.

Lets say you have a team that scores roughly 5 runs a game and also allows 5 runs a game. The run differential is even.
Now let’s say you have a choice of making the pitching better so that the team prevents one more run per game or making the offense better so that the team scores one more run per game.
The pythagorean formula is;
Winning % = 1/(1 + (RA/RS)^2)
In the first case RA/RS = 4/5, or .8 and in the second it’s 5/6 or .8333. Plugging those numbers into the equation yields winning %s of .609 vs .590 respectively. Each team scores on average 1 more run than the opposition per game but that run is more valuable in the first situation because there are less total runs in each game.

Unless I´m mistaken, if MRS is the marginal value of a run scored, and MRA is the marginal value of a run prevented, then MRS/MRA = RA/RS, so that for winning teams a run saved is better than a run allowed. So, for the Yankees of 2008 going into this offseason, a run saved was about 8% better than a run scored. For Boston it was 21% and for the Cubs 27%.

For losing teams it works the other way around.
In the previous example, and keeping in mind that a run per game is not a marginal difference, if your team had to downgrade itself due to economic reasons, and you could choose between saving one less run per game or scoring one less run per game, than you should downgrade pitching, not offense.

Changing the subject, Melky Cabrera is officially out of options.

Thanks MC.  That’s too bad because it does tie the Yankees’s hands a little bit.  I don’t think it *completely* means Gardner won’t be the starter - Melky could look completely lost in ST and Gardner could look fantastic, making the Yankees decide to DFA Melky.  There also is a *chance* they could decide to keep both on the roster.  However, it does seem likely now that Gardner will start in AAA, and be the first call-up if an OF goes down.

I wouldn’t doubt though if they both have strong ST’s, that Cashman will be fielding offers.

keeping in mind that a run per game is not a marginal difference

Of course a run per game is not a marginal difference—it’s a huge difference.  If the Washington Nationals could somehow improve by a run per game they’d go from losing 100 games to having a realistic shot at a .500 record.  A 5 or 6 WAR player is a pretty rare thing.  A one run per game improvement would require adding three of those (and losing 3 replacement level players).  How many all-start level pitchers would that hypothetical team in jyjjy’s example have to add to allow 162 fewer runs over the course of a season?  You’re talking about adding three CC Sabathias and losing three Sidney Ponsons.  In the real world, you’re almost never in a position to make such big upgrades or downgrades that you have to worry about how the changes in the run-scoring environment affect the marginal value of those upgrades or downgrades.

Melky could look completely lost in ST and Gardner could look fantastic, making the Yankees decide to DFA Melky

On the bright side, if Cabrera looks that bad in ST, he might clear waivers.

That’s the bright side? 

Oh, Melky.

Well, yeah, isn’t it?  I’d have to think that Melky would get claimed right now, or during ST if he’s holding his own.  He’s a perfectly useful role player, especially if you never let him bat against a LHP.  He’s only scary as a starter on a team that wants to contend in a tough division.

I was just thinking it might be better for the team/organization if Gardner got the first shot at the job.  We’d be more likely to see things like Damon or Swisher getting a few spot starts in CF against lefties.  And Melky just might tear up AAA for a couple of months and establish some trade value, or at least deserve another shot if Gardner flops.

  but PECOTA’s saying Rodriguez is going to decline by 2 full wins.

Must because PECOTA is taking Torre’s book into account…

Just read Mussina’s quotes on Mariano, ridiculousness.

http://www.yanksblog.com/item/2009/2/2/mussina-said-what-about-mariano-rivera

MC, I agree that a run per game has nothing masrginal attached to it, and that is why I remarked it. I was just using the same example to show that the relative value of offense would be larger than pitching for the same case if you were downgrading the team. For the example Jyjjy gave, the marginal values of both RS and RA would be equal.
But it is still true that the marginal value of preventing runs is larger for teams with a Piagorian winning record, and this in no way assumes changes in the run-scoring enviroment.

If I may link to my own blog, I examined the question of runs scored versus runs allowed here, here, and here. I wouldn’t claim that these are authoritative by any means, but I figured I’d throw them out there for those interested in some food for thought (plus, feedback is always nice).

But it is still true that the marginal value of preventing runs is larger for teams with a Piagorian winning record, and this in no way assumes changes in the run-scoring enviroment.

No, the marginal value of preventing runs is larger precisely because of the change in the run scoring environment. to quote jyjjy again: “Each team scores on average 1 more run than the opposition per game but that run is more valuable in the first situation because there are less total runs in each game.”

My point is just that if you need to be talking about a full run per game to see a substantial difference between improving offense and improving pitching/defense, then it’s hardly worth talking about.  The chances of changing your team’s overall performance by a full 162 runs on just one side of the ball are virtually non-existent.  The difference between scoring 162 more runs and allowing 162 fewer runs is about 3 wins, which sounds pretty darned impressive, and could make you want to go out and sign every above average pitcher and flashy glove man you can find.  But the difference between scoring 20-30 more runs and preventing 20-30 more runs is pretty negligible, so in general you should just go out and get the best players you can find without worrying too much about whether they’re good players because they help you score or because they help you keep the other team from scoring.

in general you should just go out and get the best players you can find without worrying too much about whether they’re good players because they help you score or because they help you keep the other team from scoring.

Exactly. If you could choose to deduct 25 runs from your runs allowed or add 25 runs to your runs scored, you’d choose the deduction every time. Unfortunately, you never have that option. You have to choose between real baseball players and the differences between those players are likely to dwarf the difference between the marginal value a run scored and a run allowed. If you have to choose between a pitcher who can be expected to save you twenty additional runs and a hitter who could add twenty-two additional runs, you gotta take the hitter because those two runs are much more important the which side of the ledger they appear on.

My point was that for x/y, x~=y, we have marginal change dx/y vs x/y - x/(y-dy) ~= dy/y ~= dx/y unless I’ve forgotten how to do a taylor expansion.

I think what Mussina was trying to say is: hey, look.  Nobody would ever criticize Mo, but even Mariano the Great has messed up.  He closes those games out and some “goats” would instead be remembered as heroes.

Mariano doesn’t walk Kevin Millar.  He closes the game.  The Yankees sweep the Sox.  ARod’s a hero.  Clutch, even.

See it?

Or doing the calc in % as discussed here recently yields the same result.  It is true that not scaling yields asymmetry.

A no-calculus argument: x/(x+y) = 1-y/(y+x).  Under variation the 1 doesn’t matter, so the formula doesn’t favor run scored or runs against.  Or consider losses.  It does matter for teams not near 0.500 - but differently depending on which side of the tracks they are on.

Mariano doesn’t walk Kevin Millar.  He closes the game.  The Yankees sweep the Sox.  ARod’s a hero.  Clutch, even.

I think that’s exactly it.  I also think Moose would have a much stronger case for the HOF if Mo had closed out Game 7 in 2001.  Not that he would be any more deserving, but having that WS ring would remove one of the non-stat arguments.  Everyone is mortal, even Mo.

[54] Well, that is my point from [47].

For me, it is computed by noting that W% =g(u), u= RS/RA, so that
W_RS/W_RA=u_RS/u_RA.

[49] The difference between the marginal values is there not because of changes that the delta would bring to the run scoring enviroment, otherwise it wouldn’t be marginal. They are there because of the non-even enviroment of the team that wants to make these decisions, and this is a well known variable, you do know a priori how your team performed in the previous season.

And for teams that already were playoff contenders, the difference in value may not be negligible. I mean, 30% is significant, no matter how you slice it.

I also think Moose would have a much stronger case for the HOF if Mo had closed out Game 7 in 2001.

Indeed.  Moose was, of course, also the starting pitcher for game 5 of the ‘04 ALCS that Gordon (not Mo, in my book, no matter what the save rule says) blew.  6IP, 6H, 2BB, 7K, 2R.  So that’s 1 ring and another WS appearance that could have been but weren’t.  Though Moose obviously also benifitted from some generally strong teams, so I’m not sure that his case overall is a sob story.

I mean, 30% is significant, no matter how you slice it.

30% of what is significant why?  Say you have a choice between adding pitcher x or hitter y.  Each projects to be about a 20 run improvement over the player they’d be replacing.  You’re seriously going to tell me that the marginal difference in the value of a run prevented is larger than the uncertainty on those projections in the first place?

Well, I don’t know hoq much is the uncertainty in the projections, but I can tell you that, in the situation you described, I’d certainly choose the 4 to 6 equivalent runs the pitcher would provide me if I was the 2008 Cubs. If you believe in the Pythag formula, why wouldn’t you, all other things being equal?

Just look at the numbers: 2008 Cubs RS 855, RA 671.

Using John Lynch tables from his blogs:

Team A RS 875 RA 675 Winning percentage = .615
Team B RS 850 RA 650 Winning percentage = .619
0.004X162= .648 wins.

2008 Red Sox RS 845 RA 694

Team C RS 875 RA 700 Winning percentage = .600
Team D RS 850 RA 675 Winning percentage = .603
difference .486 wins

These half wins are not without value.

A winning team also needs to consider short-series strategy for post-season play, which argues for maximizing dominant pitching.

OK, I wrote and deleted three different responses, but I’m just going to skip the details.  You’re obviously devoted to the numbers on this one, so all I’m going to say is that all other things are never equal, and for my money I want my GM doing his best to figure out those differences instead of falling back on the idea that a run saved is worth more than a run scored.

for my money I want my GM doing his best to figure out those differences instead of falling back on the idea that a run saved is worth more than a run scored.

Again, I concur entirely. If it actually turned out to be the case that the *only* difference between two players was which side of the ledger their contribution appeared on, then go with the runs saved. This is likely to almost never happen.

“then go with the runs saved”

if you’re a winning team.

Another extremely fiddly point to consider is evaluating salary offers - the hypothetical +/- 0.5 wins above are relevant to that calculation (along with the various uncertainties on the estimates).