for different folks, right? Obviously his comments stroked you a bit differently than they stroked me.

But I appreciate the sarcasm.

You could kick it 17th-century style and write “cost’d.”

Do you think HRF has given any thought to that?

What makes you think I haven’t?

http://www.fangraphs.com/blogs/index.php/win-values-explained-part-one/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-two/
http://www.fangraphs.com/blogs/index.php/explaining-win-values-part-three/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-four/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-five/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-six/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-seven/
http://www.fangraphs.com/blogs/index.php/win-values-explained-part-eight/

but I’ll respond later.

you put a link to that post in your signature.

every time you feel the need to reopen this debate because someone uses a stat in a way you don’t like, just say, “please click on the link in my sig file.” you’ll be spared the typing, and others will be spared rehashing the same arguments again.

WOBA can’t "define" number of runs.

Except for the part that it can. Well, it can define average number of runs produced by an event. So context-neutral, yes it can.

Actually, just determining what wOBA is is quite difficult.

To calculate by hand, yes, but you just multiply events by average run values over an out. You’re operating backwards here— Runs above average (RAA) is scaled to wOBA, wOBA, isn’t converted to runs above average. There’s a difference.

Like so many new statistics, it is sold as a sort of magic number that encompasses all offensive contribution of a player.

What other statistic does this? Also, that is what it does, actually, context-neutral, of course.

I’m not sure if it includes HBPs, IBBs and ROEs where relationship to skill level is tough to determine.

I would suggest doing research to find out. I’ll help, though. Fangraphs does include stolen bases/caught stealing. HBP and IBB are, too, because we are measuring past value— those events have value. RBOE isn’t, I don’t think.

To say it isn’t an estimate is inaccurate. It uses weighted values for the events that it includes. Those weights are estimates of values. Yes, I have no doubt that great effort went into a complex formula that resulted in an average. Like a Russian doll those formulas could be opened but they’re probably secret and if not its too damned tedious.

They are all in The Book. It’s not complex at all— they picked a sample, determined average run values for events, and then constructed a model that estimated run values, and they came out almost the same.

They are also derived from whatever sample Tango or whoever chose meaning they are estimates when applied to a different sample. Still an average isn’t the same as a measurement of a single event. This collection of estimates is then in turn used to estimate a player’s performance and turn it into the magic number.

Yes, but then they constructed a model for expected runs, and applied it to different samples. They are not idiots. There is no better way to figure out value than this. Do you have a suggestion for a new method? I know you’ll say that just because it can’t be improved doesn’t make it good, but that’s missing the whole point.

It is further complicated by adjustment to the historical environment. Its adjusted to the year, which is a large division.

You’re mixing up wOBA with runs above average. wOBA is scaled to average OBP per year. There is very, very little difference in average OBP from year to year. Would you prefer to assume that the run environment is the exact same each year?

In its attempt to create an all encompassing stat, it limits its explanatory power. How Hart added or subtracted to team wins is lost. The number doesn’t include innumerable things like getting in a rundown to let a runner score, playing injured or geeking a runner.

Again, you’re kinda looking at this the wrong way in re:wins produced. Let’s say we measure value in garbatrons, where we multiplied total runs by 100 after multiplying by 20 per 600 plate appearances. Hart was worth 1300 Garbatrons last year. Pujols was worth 8000 Garbatrons. If we do this for every player— the scale is the same, in other words— it doesn’t matter if we call it wins. We’re measuring value on a scale.

If you want to measure things like that (and I’m surprised you’d go to that argument), read this. These things are context-driven, and they’re also not worth enough to really have much of an impact.

UZR has the potential to be completely wrong. I know I’m in a minority who think that’s pretty bad for a stat. It isn’t unusual to have defensive metrics disagree across players. Taking UZR apart is a monumental task in itself. Its loaded with estimates.

Then I don’t think you understand it. It measures the balls a defender reached, and if he made the out or made an error. Then it converts mistakes to average runs it cost the team and good plays to average runs saved. There’s range runs and error runs. Could be completely wrong? Based on what? Is everything we know about fielding wrong? Maybe the Earth is flat, huh? Defensive metrics only disagree if you use a good one and a bad one. PMR and UZR are pretty close, and UZR factors in errors, not just range.

Using positional adjustments is interesting as a valuation tool, but that doesn’t mean its logically applied or even appropriate. Do you think HRF has given any thought to that?

How the hell more logically can you apply it? Shortstop is more difficult than left field. Just think of it as part of their fielding value. Carl Crawford is a worse defender than J.J. Hardy, but Crawford had a better UZR. Just tack on positional adjustment to fielding if you don’t get this. I don’t care if HRF has thought about that, though it’s not clear what “that” even is based on this comment. You give an abstract with no details, and then ask if HRF has thought about it. I don’t get that part.

Translating that to wins? C’mon, do you really think its "close to perfect?" The number of wins different with a substitute player of whatever level you want and how that would change the number of games the team would winis a guess.

Wins is a way of expressing value. We can do it in Garbatrons if you want. They’ve used math to figure out how many extra runs get a team an extra win, on average, and it’s 10. Wins is a hell of a lot more practical than Garbatrons.

You can call it an educated guess, It is unquestionably a guess: good in ways, bad in ways, educated in ways and ignorant in ways.

If you say so. The point is that your assertion that there are a ton of voodoo estimates involved is plainly inaccurate, and I’m unimpressed with your continual bashing of HRF because you don’t think he’s put much thought into this. That’s fine, that’s what win values are for. If you want to keep evaluating players by your personal opinion, their batting average and RBIs, go ahead. I’ll be in the 21st century.

And none of them are listed in the above “autopsy.”

None of the widely cited statistics are constructed in a black box. I personally would never quote a statistic that I could not find documentation about. Tango et al are very forthcoming regarding the technical underpinnings of their work.

Limitations are put on the front page, not in the appendices (unlike my thesis!).

If you don’t address my post above, the next time this issue comes up we’re going to be right back where we started. Are you unable to acknowledge that the process is pretty sound? If not, what do you dispute? I’d genuinely like to know.

How? By using one of, if not, the best available, all-encompassing metrics? I’m not saying it’s an end all, be all. What I am saying it’s better than most of what’s out there.

I also love how your argument has evolved from “its a wild ass guess” and that I’m unreasonable for using it as evidence into “lol, well, I bet you don’t know the methodology behind it.”

You are a winner.