THE BOOK--Playing The Percentages In Baseball

In Google Reader, it looks like you were disagreeing with somene named Paula about Win Shares.  Then I noticed you used the word flavoring and thought you were talking about Paula Deen.  On the third try, I fugured out you were talking about Paula Abdul.  (I didn’t have you pegged for an American Idol fan, Tom.) I don’t watch the show, but I recall when she was big in the late 80s.  I was studying Arabic at the time.  Her surname literally means “servant(slave) of the...” It’s incomplete and it bothered me.

Back to the ultrastats discussion.

@gary

I laughed.

It always annoyed me that a Win Share was 1/3 of a win.  That annoyed me enough that I havent felt inclined to learn about them.

I used to be a big fan of Idol the second year, caught bits of the first year, and watched too much of the third year.  After that, I’d give it just a passing look.

But Paula’s been mocked mercilessly, and was an SNL staple for a while, with her “x factor” and “made it your own” ramblings.  Too many of my brain cells have memories of her.

So, brainstorming causes of differences:

* WSAB uses a higher baseline, so WAR will give more credit for playing time than WSAB.

* Fielding metrics.  This could be huge.

* Position adjustments.  I have no idea how WS does this, but I’m sure it’s not based on fielding talent disparities.

* Team wins vs. hypothetical wins?

Others?
-j

Doesn’t WSAB actually attempt to rank catcher defense whereas WAR (ironically) takes a Jamesian “pass” on the subject?

I wonder if a just catcher comparison would be as close.

Actually “WAR” doesn’t take a pass at it.

Fangraphs’ version does.

Rally’s includes something for catchers:
http://www.baseballprojection.com/war/c/cartg001.htm

Is it okay that I prefer WAR simply because it’s more straightforward? Maybe I just haven’t looked at Win Shares enough, but it seems like a lot of work to get to a less accurate result. It’s just isn’t transparent to me. I mean, I know WAR requires (among other things) some sophisticated work to calculate replacement level for here league/era, positional adjustments, appropriate linear weights, and, of course, the defensive numbers… But the sheer intuitiveness of, say, +2 offense - .75 position +.5 defense +.2 baserunning +2.25 replacement level makes more immediate sense to me than trying to make all the different components of Win Shares fit together, especially when you try to figure out how it’s supposed to accurately compare pitchers and non pitchers.

With WAR it’s basically just all on a runs saved/created above baseline scale converted to wins. As I recently blogged with reference to one of Neyer’s discussions (click my name—sorry if this is “spam"),I don’t know why writers like Neyer and Posnanski are otherwise so good at explaining things clearly, ever opt for Win Shares.

If there is someone out there that kinda understands WAR and kinda understands Win Shares, and he still prefers Win Shares, then I’d like to meet him.

This was devil’s link:
http://www.drivelinemechanics.com/2009/8/13/987017/cmon-rob-win-shares-on-rob-neyer

Win Shares and WAR aren’t exactly the same thing. Win Shares attributes all of a team’s wins to specific players based on an estimate of their contribution to wins.  WAR is an estimate of a player’s value, denominated in wins.  If you have WPA on one end of a spectrum and WAR on the other, Win Shares is in between.

For instance, the Runs Created version in Win Shares gives credit to batters for clutch hitting, a specific calculation that places it squarely between WAR and WPA.

I’m fascinated by the idea of finding a place in that middle.  If graphing Win Shares and WAR made it seem like I was presenting a choice, I regret that.

studes:

I hope I didn’t come off as critical of you or your article. And your explanation of WS as being midway between WAR and WPA actually makes a lot of sense. I like it when people explore different ways of doing research and stuff. That’s how cool new stuff happens. And WSAB certainly adds a twist that (I don’t think) makes Win Shares more useful for a vareity of reasons than James’ original version.

Is it possible to judge which stat--WAR or Win Shares--is better, or can that only be found by evaluating each method?

Studes/#11:

I thought Rally had made the offense in his WAR so that the total amount of BaseRuns/RunsCreated/whatever of the individual players added up to the total runs scored by the team.

Maybe someone can help me out here.

DF/8 - While I agree that WAR is much easier to understand and calculate, I think the name “Win Shares Above Bench” is better, at least for a mainstream audience like Neyer and Poz write for.

WAR implies a replacement level baseline, which a lot of people don’t like for whatever reason, while WSAB compares against a less theoretical concept.  Also, the “Shares” part sounds good to me for some reason.

Aaron, you are correct, I make the runs add up to the team level using custom base run values for every.  I don’t look at individual clutch hitting, like RISP, like win shares does to get to that result.

Is it okay that I prefer WAR simply because it’s more straightforward? Maybe I just haven’t looked at Win Shares enough, but it seems like a lot of work to get to a less accurate result.

Actually, I have the opposite reaction.  One of the reasons I still like Win Shares is because the concept is simple: take a team’s wins and break them down into offense and defense, pitching and fielding.

Yes, the math is complicated (and at some points makes you scratch your head. Hard.) but conceptually to me it’s more appealing than “we use linear weights based on RE tables for individual teams, add them to ...”

See what I mean?

Personally studes, I use WAR in a lot of my writings and it helps that it is easier to calculate so I can show I got where I got. 

For example,

“x player is projected to have a .350 wOBA the rest of the year, which translates to x runs above average.  He is also an x run fielder.  Then you add the positional and playing time adjustment, divide by 10, and player x is worth x WAR!”

That allows the readers to input their own values for fielding or offense and be able to figure it out on their own.  You couldn’t do that for Win Shares.

studes #15: an earlier post is still in the moderation process (I wonder if it’s my handle?), but anyway, I do see your point on the “comprehensibility” scale, although I think (as far as I understand Win Shares) the moves from coinverting team wins to player contributions is at least as complex as the RE tables, etc. I want to make sure you know that I do like the stuff you do with Win Shares for a number of reasons, not the least of which is that I think in about an field, multiple “research programmes” (philosophy of science jargon—shudder) are a good thing.

Rally #14: I like how you do the offense on your site, I meant to mention that elsewhere… I think I do in my earlier post re: Neyer. That just shows how flexible and adaptable WAR is. Maybe this is more appropriate for another thread, but do you do that using BaseRuns or a more “traditional” lwts formula using the team as the “entity?”

Nick #13: I think we should make sure and see how many message boards we can spread this discussion across (winky face)! But seriously… another thing about WAR is that you can easily break is down into runs/wins above/below average (of which it is “parasitic"), which makes it more easily understandable for folks who don’t like the replacement-level baseline.

That allows the readers to input their own values for fielding or offense and be able to figure it out on their own.  You couldn’t do that for Win Shares.

Good point, Nick, and I agree it’s useful.  I think the difference is in the audience.  Most baseball fans don’t want to, and won’t, get into substituting their own stats or that kind of thing.  I assume you consider your audience to be real stat heads.

My audience (at least the one in my head) has typically been people who like stats but who aren’t likely to spend time creating their own or worrying about linear weights vs. runs created.  That’s the same audience that Neyer writes to.  They aren’t the ones to worry about how James does everything in the details.  But I think the notion of splitting wins between players is something that can appeal to them, at least to some of them.

That is why Win Shares continues to appeal to me, and why I try to improve Win Shares within its structure.

I have always wanted to tackle the fielding portion of Win Shares on my own, but just haven’t had the time and/or insight.

Rally #14: Ah, that explains it.  I remember a few months ago I noticed that you and fangraphs differed significantly on Mauer’s batting value in 2008 (45 runs vs 28), which seems bigger than slight differences in LW methods would normally give.  I guess the fact that the Twins outperformed their component statistics last year to is the explanation.  As a sanity check, you get the same thing for Morneau (42 vs 26.)

Studes, I actually write for you at THT Live (and BtB which is a little more saber), so I hope that out intended audiences are the same

Anyway, I agree with you.  The concept of Win Shares is easier to understand, mainly because of the name which I think is genius; however, it’s clearly not as accurate as WAR and harder to calculate.

Not to worry, Nick.  I know who you are. 

People can write on the same site and have different implied audiences.  I like to think that’s one of the things that makes THT work as a general baseball site of interest.

I still wish James would have used decimal for Win Shares instead of dividing by 3.  I love trying to explain to someone WinShares and you get to the division of 3 and the person asks why it is done that way. 

The only response is “Bill James says so.”