2008 WAR Totals of Five 2005 Draftees from www.baseballprojection.com

Jay Bruce -0.2
Ryan Zimmerman 0.9
Troy Tulowitzki 1.2
Ryan Braun 2.1
Alex Gordon 3.3

6 months ago devil_fingers 36 comments 2 recs |

Story-email Email Printer Print

Comments

You've got to be kidding

Braun has a smaller WAR than Gordon? Who did you pick, Ryan Z. Braun? I find Tulowitzki hard to believe too.

by AxDxMx on Apr 29, 2009 4:44 PM EDT reply actions 0 recs

I guess he was a butcher at 3rd.

And now he is an outfielder, so I guess 37 homers and crappy defense gets you that?

I’d still take him over Gordon.

by AxDxMx on Apr 29, 2009 4:50 PM EDT up reply actions 0 recs

Tulo in 2008

Crappy hitting plus average fielding plus playing in only 101 games is going to add up to a pretty low WAR total.

The immoderate moderator

by NYRoyal on Apr 29, 2009 4:52 PM EDT up reply actions 0 recs

What accounts for the large difference between those numbers and Fangraph's WAR?

Is it the use of different defensive measures? According to Fangraphs:

Braun 3.9
Gordon 2.6
Zimmerman 2.2
Bruce 0.9
Tulo 0.8

The immoderate moderator

by NYRoyal on Apr 29, 2009 4:52 PM EDT reply actions 0 recs

Defense is the probably biggest difference in general, and certainly for these particular players — TotalZone versus UZR. TotalZone like Gordon’s 2008 D better than UZR (although it doesn’t like his 2007 D as much… usual defensive metrics stuff). TotalZone hates Braun’s OF defense.

Rally probably uses his own park adjustments.

Positions: I don’t know what positional adjustments Rally uses for contemporary baseball, but they won’t be that much different, if at all. I know he adjusts them by decade.

Replacement level: FanGraphs doesn’t adjust for the relative difficulty of the AL and NL, Rally does.

Baserunning: FanGraphs doesn’t include baserunning beyond the SB/CS stuff included in their implementation of wOBA (for their hitting weights above/below average in the "Value’ section), while Rally not only does basestealing separately, but includes baserunning in general (taking extra bases, advancing on wild pitches, etc.)

Double Plays: Rally includes the run values grounding into double plays above/below average given an equal number of opportunties. He also compensates for contact hitters by giving the linear weight value of grounders that advance (or potentially advance? I don’t know) runner, although I don’t know if that’s in the GDP section or the “Bat Runs” section. I read him commenting on it elsewhere. It’s sort of complicated, though, since I have Gordon has very good at avoiding the DP according to b-r’s situational stats, while Rally has him as exactly average. Must be the groundball thing.

Bat Runs: I don’t know what all Rally includes or doesn’t include that FanGraphs doesn’t in their “Hitter Value” (which is just a certain wOBA turned back into linear weights: so the various kinds of hits, walks, HBPs, SB, CSes). I covered the park adjustments above. He does say that he adjusts the offensive output of the each player to the runs actually scored by the team so that they don’t get credit for runs that were “created,” which I think it pretty cool (and more accurate, when you’re accounting for “value” rather than ability).

I’m not advocating one version of WAR or the other, I was just messing around, and given recent discussions, thought this was very interesting.

….

The truth is, Alex Gordon sucks, but I’m so in love with him that I totally made up that whole websiet and all the stats contained therein to justify my “Gordon is a decent player” claims. This slanderous page is the tell.

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 5:10 PM EDT up reply actions 0 recs

I'm glad you finally came clean

The truth is, Alex Gordon sucks, but I’m so in love with him that I totally made up that whole websiet and all the stats contained therein to justify my "Gordon is a decent player" claims

I knew it.

With regard to defensive metrics, I wish to hell the development of these metrics would move towards a consensus. Unfortunately the difficulty in measuring fielding, and thus the sometimes wildly different values associated with a particular player’s fielding, leads to some very disparate WAR measures. It’s going to be difficult to convince many people that WAR is a great total value stat when one version of WAR says Braun was merely average in 2008, while another says he was very, very good.

The immoderate moderator

by NYRoyal on Apr 29, 2009 5:18 PM EDT up reply actions 0 recs

I agree on the defense thing

I think short-term there will be big improvements, perhaps even in the stuff available publicly for free to people like Rally (and while UZR is better than TotalZone, I would guess, I love that Rally does so well with just freely available RetroSheet stuff — not just the defense thing, but everything. Basically, one guy spent the last couple years learning stuff, and now outdoes WARP, etc. by a long shot for everything past 1955 or so. Well, at least BP still has the best projection system…). I think it will be a long time until the defensive metrics are a a point where they can judge with anything approaching the precision of offensive stats. I haven’t thought that much about it, but I’m not sure it’s possible in principle.

In the meantime, I think it’s a good thing that different people are doing it in different ways (not just defense, but baselines, etc.) to generate discussion, further research, and refinement and stuff. That’s one of the smarter things BP did — having both RARP and VORP (EqA and MLV) on the site. Yes, I think that EqR is way better than MLV, but having individuals pursue different lines of thought, research strategies, etc., fosters creativity and whatnot.

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 5:26 PM EDT up reply actions 0 recs

In the meantime, I think it’s a good thing that different people are doing it in different ways (not just defense, but baselines, etc.) to generate discussion, further research, and refinement and stuff.

This is true. Of course it makes the whole “how good is Player X” question still often difficult to answer. I mean it’s always difficult, but it is more than a little frustrating when two different versions of the “same” total value metric take you in two very different directions. I guess for the time being in addition to averaging projections and average defensive metrics, I’m going to have to start averaging WAR values too (and hope that the truth lies in the middle).

Hopefully Hit f/x data will help improve PBP defensive metrics until we finally get some sort of Field f/x.

The immoderate moderator

by NYRoyal on Apr 29, 2009 5:36 PM EDT up reply actions 0 recs

If I was "averaging"

anything, in this case it would only be the defensive metrics. It’s just me. That’s because of the vagaries and limitations of defensive numbers and data sources.

As for the other stuff, it tells you different things. Positional adjustments, replacement level, etc. aren’t different interpretations (I mean, they are that) of empirical data as much as they are, well, okay, they are, but they’re more different “philosophies.” Or with the Batting Runs, you have to decide how much a player’s value is determined by his team offensive context or not. Or whether grounding into double plays should be involved. Or baserunning (you could just add EQBRR-SBRns to FanGraph’s figures).

Personally, I think FanGraphs should be prorating replacement level and position adjustments per 700 PAs rather than 600, and that they should have different levels for the leagues. I guess I’ll have to start my own site!

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 5:43 PM EDT up reply actions 0 recs

As for the other stuff, it tells you different things. Positional adjustments, replacement level, etc. aren’t different interpretations (I mean, they are that) of empirical data as much as they are, well, okay, they are, but they’re more different "philosophies."

Isn’t this also a fairly accurate description of the differences between various projection systems (PECOTA, CHONE, ZiPS)? And, instead of picking one as the best (which is very difficult) it seems like it makes sense to average them. I think that works for the two WAR’s as well. Of course a third choice would be to pick the pieces that you like from each philosophy and create your own WAR, but that’s a bit more labor intensive.

The immoderate moderator

by NYRoyal on Apr 29, 2009 5:56 PM EDT up reply actions 0 recs

My problem is that some of metrics are set by human judgement.

Errors, zone of hit, Type of Hit (LD vs FB).

I think these measures need to be standardized first before moving on with the metrics.

by Jeff Zimmerman (TucsonRoyal) on Apr 29, 2009 7:21 PM EDT up reply actions 0 recs

Thankfully some of it will become standardized and objective

Hit f/x will allow all of the batted ball data to be standardized. Zone of hit is going to be human until some sort of Field f/x is developed (unless that can be determined by the Hit f/x data). But I think this is the element of WAR that is least susceptible to human judgement. Errors, I think, will forever be an iffy judgement call.

The immoderate moderator

by NYRoyal on Apr 29, 2009 7:25 PM EDT up reply actions 0 recs

The batted ball data from BIS, while not perfect, is probably the best available now

Unlike the 30 different official scorers (plus the replacements when the official scorer goes to a wedding)
BIS has its own staff who follow BIS’s own protocol on how to classify the batted balls, so the consistency will be much greater. I believe Fangraphs uses the BIS data.

Unfortunately, the BIS batted ball data only goes back to earlier this decade and costs money to use.

by Gopherballs on Apr 30, 2009 12:34 PM EDT up reply actions 0 recs

yup to FanGraphs and BIS

What does Stats, Inc., do?

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 30, 2009 12:36 PM EDT up reply actions 0 recs

No idea

Dewan makes the rounds promoting the Fielding Bible and BIS, so we know what they do, but Stats seems to have a different business model and does not promote to the general public. Stats might do the same or similar things, but as far as I know, they do not publicize it.

by Gopherballs on Apr 30, 2009 1:55 PM EDT up reply actions 0 recs

The consistency matters

Typically I find the subjective data BIS collects is quite consistent, which is what makes their pitch type classification more reliable than the pitchfx algorithm right now. Sure it doesn’t have all the goodies, but you can actually tell when a pitcher is throwing more or less of a certain pitch without manually futzing around with the data.

by dkappelman on Apr 30, 2009 3:52 PM EDT up reply actions 0 recs

Do you know if Fangraph's pitch type classifications and data comes from BIS as opposed to Pitch f/x?

The immoderate moderator

by NYRoyal on Apr 30, 2009 5:02 PM EDT up reply actions 0 recs

Both

In the regular stats pages it comes from BIS. In the special Pitch f/x sections it’s all pitch f/x data.

by dkappelman on May 1, 2009 1:49 AM EDT up reply actions 0 recs

per 700 PA?

Just stumbled upon this conversation. I’m just curious what you mean by prorating Replacement or positional adjustments by 700 PA instead of 600?

The Replacement section is just 20 runs / 600 PA or .0333 runs per PA , and the positional adjustments are what they are at a particular position per 162 games. If you change what they’re prorated at you’re changing what replacement level is, or what the positional values are.

by dkappelman on Apr 29, 2009 11:17 PM EDT up reply actions 0 recs

hey!

glad to see you stopping by, Dave. I was half-joking given the context (the variance between versions of WAR), but while we’re on it, I thought that the positional adjustments were done per 600s PAs, too, like replacement level. Doing the positional adjustments that way makes more sense (and that’s the way I did it on my own spreadsheet before FanGraphs started posting WAR values).

Can I just ask why 20 runs/600 PA rather than say 22.5 runs/2.25 wins per 700 (or 20/700 for the NL or 25 for the AL)? Just curious. It’s not a huge deal, I’m just following what I understood from How to Calculate WAR (contains older positional adjustments).

Thanks for the great site, Dave. I’ve lots of annoying suggestions… But also, you once personally told me the easy formula for runs-wins conversion, which has been very helpful for, um, whatever it is I do.

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 11:52 PM EDT up reply actions 0 recs

not sure

There’s no real reason why we don’t do different replacement level for different leagues. Just something I really haven’t gotten around to yet.

20/600 is basically 22.5/700, it’s off by like .001 runs per PA which is half a run per 600 PA. Not really any difference there. Even at the team level it’s only a half a win difference.

by dkappelman on Apr 30, 2009 3:38 PM EDT up reply actions 0 recs

thanks

like I said, no big deal, just curious

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 30, 2009 4:01 PM EDT up reply actions 0 recs

ooo, ooo, and while you're here

can you explain how you park-adjust the “Batting” line in your hitter value sections. Am I right that the wRAA numbers aren’t adjusted in the “Advanced” sections, but that the “Batting” in the “Value” section is? I’ve tried applying a couple different factors (including the same ones listed in Cameron’s series on value), but it doesn’t make sense. I know how to convert wOBA to wRAA easilly enough, but the park-adjustments are something I haven’t done to my satisfaction.

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 30, 2009 12:06 AM EDT up reply actions 0 recs

Right

wRAA is not park adjusted, but Batting is.

Batting is park adjusted using the “odds ratio” method. Where you add in the difference between league average R/PA and the park adjusted R/PA and then multiply by PA.

For instance let’s say a player is 10 wRAA with 500 PA and league R/PA is .12 and his home park R/PA is .10

You’d do wRAA + (.12 – .10) * PA

This works quite well and is really simple to do for run stats based on PA.

by dkappelman on Apr 30, 2009 3:45 PM EDT up reply actions 0 recs

thanks so much

do you use the same adjustments you do for pitchers? (Patriot’s, I think, correct me if I’m wrong)

thanks for taking the time

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 30, 2009 4:02 PM EDT up reply actions 0 recs

the same ones

Yep, we use the same PF for batters and pitchers and we use exactly the same methodology as Patriot does, but I’m calculating them on my own, so there might be some extremely slight differences. For 2009 we’re using 2008 except with the new parks where we’re just using 1 until there’s some data.

by dkappelman on Apr 30, 2009 4:57 PM EDT up reply actions 0 recs

What about 2006-2007?

Relive Royals History at royalsretro.blogspot.com

by RoyalsRetro on Apr 29, 2009 4:53 PM EDT reply actions 0 recs

drafts or these players?

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 4:54 PM EDT up reply actions 0 recs

These players

Zimm and Tulo were much better prior to 2008.

Relive Royals History at royalsretro.blogspot.com

by RoyalsRetro on Apr 29, 2009 5:02 PM EDT up reply actions 0 recs

no doubt

Not saying that I think Gordon is a better player, just thought given all the… unhappiness… and draft talk, that Rally’s figures were interesting… Gordon’s a good player

Let’s do a Rally/FanGraphs combo for 2007, since Gordon wasn’t up in 2006, and I’ll sub in Justin Upton (#1 in 2005, although he’s really young — again, this isn’t a “projection” thing) just for completeness. Keep in mind that Ryan Braun didn’t start the season in the big leagues in ’07 (if I remember rightly).

First number will be Rally’s, the second FanGraphs, for 2007 WAR

1. Justin Upton -0.8/-0.6
2. Zimmerman 3.2/5.1
3. Tulo 5.6/5.3
4. Ryan Braun 1.3/3.1
5. Alex Gordon 1.1/2.1

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 5:18 PM EDT up reply actions 0 recs

I think Zimmerman v Gordon is a pretty interesting debate

by royalsreview on Apr 29, 2009 7:03 PM EDT up reply actions 0 recs

how so?

Just curious. I’d just assumed that Zimmerman had been tons better the last could seasons. I was surprised that Alex was as close to him as he was.

I don’t follow the NL that closely, so I didn’t realize Zimmerman had such a poor season last year, at least accordingn to Rally’s defensive system. But even with a more favorable system to Zimmerman (FanGraphs/UZR), Alex outplayed him overall — better hitting. And that isn’t adjusting for league difficulty, either.

We’ll see going forward. The Projection systems all see Zimmerman as the better hitter at this point, and I think overall, his defense has been better (TotalZone is the only dissenter — UZR, Dewan’s, Fans Scouting Reports all like Zimmerman’s D much better). So you’d have to give him the edge.

On the other hand, they are both young, Zimmerman didn’ t exaclty dominate the league last season, and so on, so we’ll see.

I guess I was more interested in your thoughts than mine, though, although I blabbed and blabbed…

I'm not a sabermetrician, but I do play one at Driveline Mechanics.

by devil_fingers on Apr 29, 2009 7:36 PM EDT up reply actions 0 recs