SG, how does your version of wOBA differ from FanGraphs?  I know they include SB/CS and I don’t think you do, but besides that?  I ask b/c you have Swisher’s wOBA at .357, while FanGraphs has it at .375, which looks like 2 runs per 650 PA.  Not sure if that’s significant (doesn’t appear so).  But anyway, Swisher didn’t have any SB, nor did he have any attempts.  I’m sure FanGraphs version isn’t park-adjusted (they do that later), not sure if it is league-adjusted.  Best I can find for their formula is this:

(0.72xNIBB + 0.75xHBP + 0.90x1B + 0.92xRBOE + 1.24x2B + 1.56x3B + 1.95xHR) / PA

Which is from Inside The Book.  I’m always curious when numbers with the same name and same meaning come out differently.  As always, great stuff!  Replacing Damon and Matsui will be a big deal for this offense.

SG, how does your version of wOBA differ from FanGraphs?

SB is the big difference, although like you say with Swisher that shouldn’t be the issue.  I don’t take out IBBs, and I don’t have ROE in there, but the rest of the weights are the same. 

A pure runs analysis says DNYS was a pitcher’s park, but if you look at the component stats it actually plays more like a hitter’s park.  So I’m going to assume that park factors are the primary difference.

But yeah, two runs over 650 PAs isn’t a big deal I don’t think.  As long as the formula is consistent, for the purposes of these comparisons it should be fine.

(Damon + Swisher + Matsui) * (Relative Health + Exceeding Projections) = 2009 Yankee Bottle Lightning.

What would a zero BR/650 correspond to? a player who always gets out?  I’m asking because it seems that wOBA should be (absolute batting runs/PA)*scaling, but trying to compute BR/650 from that gives numbers about 100 runs too high.

What would a zero BR/650 correspond to? a player who always gets out?

Not quite.  It’d be a player whose negative contributions equal his positive contributions.  If you look at the weights of the events, some are positive, some are negative.  So if the total of your weighted positive events (H, BB, SB) is less than or equal to the total of your weighted negative events (batting outs, CS, DPs), you’d show as zero or negative even if you didn’t get out all the time.

I’m asking because it seems that wOBA should be (absolute batting runs/PA)*scaling, but trying to compute BR/650 from that gives numbers about 100 runs too high.

Yeah, you need to factor in the outs.  Looks like something around -0.435 times outs subtracted from wOBA BR gets you pretty close.

The primary way I use wOBA is to compare two players in the batters box.  Since it doesn’t include things like SB/CS, DP, it’s not quite as robust as full linear weights so I prefer not to use it when looking at total run values.

Aren’t batting runs scaled to have zero = average?

SG- What’s the past season weighting formula that you use?

Aren’t batting runs scaled to have zero = average?

Not necessarily.  Depends what you are trying to do.  If you want to look at how many absolute runs a player produced, you scale it so that everything is related to no runs by subtracing something like -0.1 times outs (out value depends on the run environment).  If you want to compare to other players, that’s when you want to shift the scale to zero = average.

It might be clearer if I give an example.

Derek Jeter: 745 PA, 100 BR
Brett Gardner: 735 PA, 87 BR
Mark Teixeira: 725 PA, 120 BR
Alex Rodriguez: 715 PA, 124 BR
Jorge Posada: 705 PA, 96 BR
Robinson Cano: 695 PA, 100 BR
Nick Swisher: 685 PA, 93 BR
Juan Miranda: 675 PA, 84 BR
Melky Cabrera: 665 PA, 74 BR
Starter Total: 6345 PA, 878 BR

Derek Jeter: 745 PA, 9 BRAA
Brett Gardner: 735 PA, -4 BRAA
Mark Teixeira: 725 PA, 31 BRAA
Alex Rodriguez: 715 PA, 36 BRAA
Jorge Posada: 705 PA, 10 BRAA
Robinson Cano: 695 PA, 14 BRAA
Nick Swisher: 685 PA, 8 BRAA
Juan Miranda: 675 PA, 0 BRAA
Melky Cabrera: 665 PA, -8 BRAA
Starter Total: 6345 PA, 97 BRAA

In the first set of numbers, we’re looking at how many total runs the Yankees as a team would score using their 2010 CAIRO projections, by scaling batting runs to absolute runs. 

In the second set, we see how many runs above average their offense would be by scaling batting runs in terms of above/below average.

Obviously their starters aren’t going to play 162 games each, so don’t take that 878 run total as saying they don’t need to add any hitters.

SG- What’s the past season weighting formula that you use?

7/5/4/2 for hitters, 8/5/3/1 for pitchers.

[9] Thanks. The reason I asked is because it’s obvious whatever happened with Swisher in Chicago looks like the outlier, but it was weighted most heavily in his projection. I guess that’s the ‘grain of salt’ factor with all projection systems.

I think of the broken wrist play when I think about Matsui too.

Of course, I also think about the grand slam in the snow on his first opening day.  Loved watching him play!

so don’t take that 878 run total as saying they don’t need to add any hitters

That list doesn’t even include Adam Dunn and Matt Holliday, so it’s not really indicative of what we can expect they’ll do in 2010.

Matsui’s double off Pedro in the middle of the 8th inning rally in the 2003 ALCS was another memorable moment.

[2] Thanks!  Looks like if you take the ROE and IBB into account (2 of each) it’s a difference of about .5 runs.  So sure, park-factors are probably the difference, and FanGraphs doesn’t apply those until they determine offensive-runs for WAR.

I will always remember an absolute bomb that Shemp hit off B.J. Ryan in September 2003 (?). It was a pretty meaningless game, but he just annihilated one of the toughies lefties in baseball.

The primary way I use wOBA is to compare two players in the batters box.  Since it doesn’t include things like SB/CS, DP, it’s not quite as robust as full linear weights so I prefer not to use it when looking at total run values.

Thanks.  The reason I was looking at it is that wOBA annoys me because it looks like a rate but it isn’t really, not in the way AVG, OBP and even SLG are.  I guess you could use BR/PA to get a decimal rate, but that gives a pretty small number and it can be negative, which is unattractive.

A pure runs analysis says DNYS was a pitcher’s park, but if you look at the component stats it actually plays more like a hitter’s park.

I’d love to understand what this means…

[1] In simple terms, I think it just means that the component stats should have resulted in more runs than were actually scored.  So DNYS must suppress clutch hitting.  Or something.

I’d love to understand what this means…

Teams were less clutch at DNYS. This is a serious reply, by the way. Their components (AVG/OBP/SLG/etc.) would lead one to believe a lot more runs were scored than were actually scored. I could be wrong though.

[18] Hey, I was going to answer that!  Yeah, I think you’ve got it exactly.

I guess you could use BR/PA to get a decimal rate, but that gives a pretty small number and it can be negative, which is unattractive.

I’m pretty sure that’s what wOBA is.  They use some constant (maybe variable for run-environment?) then to scale it to OBP so it is a familiar number

[18] In even simpler terms it means Robinson Cano’s home park is DNYS

I’d love to understand what this means…

The standard way you calculate a park factor is to compare the runs scored in all home games and road games by both teams.  There were 819 runs scored in DNYS this year compared to 849 on the road.  So that would lead you to think it suppresses scoring.  You’d divide 819 by 849 and get a park factor of something like 96 or 97, which means it’s 3-4% less conducive to scoring.

But there is a lot of noise in those numbers, especially in a single year.  You have factors that can influence scoring that are not park-related, likes baserunning, timeliness of hits, etc.,  The other issue is distribution of PAs.  I don’t have the time to look up the numbers now, but I’m pretty sure the games on the road will sum up to more PAs than the games at home, since the Yankees won enough road games that the home team had to bat in the ninth more frequently.

So as far as I’m concerned, if we want to guesstimate a park factor for DNYS in 2009, we need to look at the difference in context-neutral offense on a rate basis to smooth out the things that can skew the run scoring distribution.  When you do that, you get a park factor of 102, which means it boosts overall scoring by 2%.

[22] So what you’re saying is that it is Swisher’s fault for not hitting HRs at home?

Can we save the post-mortems on guys until they’re actually gone? I’m still hoping Zilla takes a one-year deal.

[22]  Building on this…do we do park-factors for SB?  How about things like foul-outs?  Non-SB baserunning?  I think the 102 is probably on the mark, but I’m wondering if there are things we don’t typically take into account that could help account for the differences in expected runs and actual runs at the park.  E.g. if the base-paths are a slow running surface, it may suppress SB and the ability for runners to advance an extra base on singles (or doubles in some cases).  I’m wondering if that has ever been studied?  I’ll Google it when I have a chance, but any info you can provide would be fantastic.

In even simpler terms it means Robinson Cano’s home park is DNYS

Well played.

All the moments folks mentioned for Matsui are legit, but to me, the signature Matsui moment is every single time he threw the ball in from left field so hard he did a forward somersault.

I just loved that.

I was sitting in the right-field bleachers when he hit the grand slam on opening day. That was a great great moment to be at any ballpark…..

[24] Then we may not be doing a review of say Hinske until February?  SG needs to do season reviews now; soon he’s going to be doing more important things like analyzing potential trades, FA’s, reviewing contracts, etc.

I’m pretty sure that’s what wOBA is.  They use some constant (maybe variable for run-environment?) then to scale it to OBP so it is a familiar number

Sort of, they shift the values so that an out is worth zero, and then scale it, but that prevents it from measuring anything that’s tangible in the game.

In even simpler terms it means Robinson Cano’s home park is DNYS

The house that Cano botched.

No, I meant the “favorite moments in his Yankee career” stuff

Building on this…do we do park-factors for SB?  How about things like foul-outs?  Non-SB baserunning?  I think the 102 is probably on the mark, but I’m wondering if there are things we don’t typically take into account that could help account for the differences in expected runs and actual runs at the park

I wouldn’t be surprised if someone out there has done something like this with retrosheet, but I haven’t seen it.  I think a lot of it would be captured in other areas.  For example, if a stadium is above average in terms of foul outs, we’d see impact in the hits and runs factor, etc.,

For example, if a stadium is above average in terms of foul outs, we’d see impact in the hits and runs factor, etc.

Thanks.  I found an article by MGL (that he wrote) a few years ago and he did include a foul-outs factor.  Nothing on base-running.  I’m just wondering if we can get it down to be more than just some randomness - hitting w/RISP for example - that would explain the difference between component factor and runs-factor for a single season.  Chances are there’s just too much noise and at most we could find maybe 10 of those 40-some runs in “other” factors.  Oh well.

Pujols rightfully took home the NL MVP unanimously.

What I love about MVP voting is that the ballots are big enough that you see some REALLY funny votes.

To wit, one voter felt that Jeremy Affeldt deserved recognition as being among the 10 most valuable players in the National League in 2009. Nice.

As for the AL MVP ballot, which vote is funnier - Placido Polanco or Jason Kubel?

Or Robinson Cano tied with Zach Greinke?

What I love is that, even with 30 names, there are no Mets on the list.

Cairo is deadly accurate on Damon and Swisher, at +1 standard deviation.

Actually, I am surprised that Zips loved Swisher so much.

The standard way you calculate a park factor is to compare the runs scored in all home games and road games by both teams.  There were 819 runs scored in DNYS this year compared to 849 on the road.  So that would lead you to think it suppresses scoring.

Thanks to all for the answers, esp. SG.
What does the number for runs scored “for both teams” “on the road” mean, though, exactly?  What NYA scored on the road is clear enough.  How do we get the number for what their opponents score “on the road”?  Runs scored at non-NYS-games by each opponent throughout the season, pro-rated for the number of games they actually played at NYS?  I apologize to the majority of you to whom this is self-evident…

OT, but since there’s basically a news vacuum:

Davidoff:

I spent some time with Johnny Damon today, and he spoke forcefully about his free agency and his future. I asked him specifically about Scott Boras’ comments from a couple of weeks ago, and Damon backed his agent 100 percent.

Damon’s history shows that he’s not going to bend in his free agency. It’s how he became a Yankee in the first place. It’s why I think he’ll be elsewhere, come next season.

“OT, but”

Far as I know there’s no site policy about thread topicality, or even any opinion about preferred practice.  Of course I say this as probably one of the least topical commenters.

Incidentally, Rilke, did you know that the Bey of Algiers has a wart under his nose?

WP, at the library the other day I found Rilkekind a book about rare and endangered creatures, all from Australia, but it’s not worth mentioning.

That wart or boil or lump is still the subject of translatorial dispute.

Very cool.
You’re quite right, Rilke.  Actually, it’s indubitably incorrect, but for the current application all that matters is that is be comprehensible, and I think that’s the version that would come up most quickly on the web.

What does the number for runs scored “for both teams” “on the road” mean, though, exactly?

It means look at all the runs that scored in games that involved the Yankees, and split them into games in the Bronx (home games) and everywhere else (road games from the Yankee perspective).  So if the Yankees scored 500 runs and allowed 300 runs at home, that means there were 800 runs scored in their park. Then, you look at how many runs the Yankees scored and allowed on the road.  You divide the Yankee Stadium numbers by the road numbers, and you have your park factor.  If the numerator is bigger than the denominator, it’s a park that favored hitting.  If the denominator is bigger, it’s a park that favored pitchers.  If they’re the same, it’s neutral.

Bear in mind this is the simplest park factor you can do, and it’s probably not very useful when looking at a single season.  You can create park factors for just about every event, and you can split them by handedness as well. 

You shouldn’t necessarily use component/platoon park factors when looking at value in-season, but it can be useful when trying to project how a player moving to a new stadium may do.  In-season, a run specific adjustement is enough.  Even if a player gains a particularly large advantage from his park, it doesn’t change his value relative to the run environment of the park itself.

What NYA scored on the road is clear enough.  How do we get the number for what their opponents score “on the road”?

Runs allowed by the Yankees’ pitching and defense = runs scored by the opposition.

Runs scored at non-NYS-games by each opponent throughout the season, pro-rated for the number of games they actually played at NYS?

No, just runs scored/allowed in games where the Yankees are on the road.

This is a pretty good explanation of one way to work through park factors.

BBTF: Park Effects

SG, thanks much!  That, of course, clears it all up.

[43] Without having thought at all, I wonder if it is really the case that teams try equally hard to win games away and at home.  Can’t see off-hand if that would affect pitching or hitting more - I thought the former at first, since that’s a larger effect, but it’s easier to rest regular position players away than to mess with the rotation.  Of course that shouldn’t affect park factors since it ought to be symmetric to first order (well, perhaps the Yankees face few house-money lineups in YS) but maybe it’s a significant factor in home field advantage, in which case I would think the post-season HF factor should be reduced, if it’s based on regular season data.

[42] I couldn’t even manage to learn the Cyrillic alphabet.  What is it with Gogol and noses anyway?

hmm, I think it would be possible to check if away lineups are typically weaker than home lineups (average projected wOBA, I guess), which would explain at least part of home field advantage.  And it would be interesting to compute home field advantage using just post-season games.

One factor that could cause a team to play worse on the road may be the fatigue caused by traveling and changing time zones.

According to Baseball America (h/t LoHud), IPK’s FB is maxing out at 93.4 and averaging 90.66.

[45] I really enjoyed learning the Cyrillic alphabet during my two semesters of Russian. Sadly, due to the fact that there was only one Russian professor at my school I was unable to continue due to scheduling conflicts. But strangely enough, I found Russian to come much easier than Spanish ever did.

[49] Russian was the only language offered at my high school, so I took 4 years of it.  I only remember the curse words at this point, but that has more to do with the fact that high school languages aren’t taught immersively and less to do with the difficulties of the Cyrillic alphabet, which you start to get the hang of after a semester or two.

[47]
Rich, couldn’t THAT be determined…. well, not determined, but suggestively indicated, by comparing actual road performance to predicted performance over a sufficiently large sample size of season performance?  &, if we wanted to be more certain still, comparing it to the same for home actual-vs-predicted performance?

[50]
The ONLY language offered?  What high school was that, Deviance?  (Serious question.)

Rilke - if you’re really asking, I believe I can answer that question, but the answer would be too long to be appropriate here.

[50] I actually took two years of HS Russian, but the first year was partially derailed b/c our teacher got into a skiing accident and missed about 6 weeks or so.  I then took two semesters in college, but never quite got up to “conversational” Russian.  If you put the whole alphabet in front of me I could probably tell you what all the letters are, though after 14 years I couldn’t write it off the top of my head.

That and French (which I took 3 years in school) I always want to get back into, especially as they’re teaching some French words/phrases to my son in 2nd grade.

[51] The school is called Staten Island Technical High School… it’s a NYC public school, founded in the late early 80’s, that required certain grades to get into (I think now it might be part of the Stuyvesant test, along w/ BK Tech, Bronx Science, etc).

As I understand it, they only offer Russian because, at the time the school was founded, Russia and the US were the two world super powers, and it was thought that learning Russian would be beneficial to us future American architects, engineers, scientists, etc.  (The school also required everyone to take 3 years of electrical engineering, 2 years of CAD and architectural drafting, stuff like that.)

By the time I was at Tech in the late ‘90’s, the Soviet Union had of course collapsed, and the Russian language was no longer seen as being particularly relevant, so there was talk about changing the school’s language to Japanese, though that change never did happen and Russian and Russian alone continues, I believe, to be taught there.

*that should be “early ‘80’s”, not “late early 80’s”