Tangotiger Blog

And later in the thread, you see the confusion of oWAR and dWAR.  It’s really too bad that BR.com continues to do this.

I think the WAR framework is fantastic, and Rally’s implementation on his site is wonderful, as is David’s at Fangraphs.

Sean has done tremendous stuff in putting it out there as well.  But these little things simply are obstacles that don’t need to be there.  There’s confusion in the marketplace where none needs to be.

It’s really quite simple. Everything is relative to average. Everything. And then add everything up making sure that you are combining apples with apples, which isn’t too hard if everything is is runs above/below average (whether you are combining rates or total runs). Throw in anything you want. Base running, defense, catcher framing, SB/CS. You can even throw in veteran influence on other players if you can show some substantial evidence that it exists and that you can measure it in with some measure of accuracy and reliability. As I said, you can throw in anything you want. Just put everything in the same currency - usually runs/wins.

Now, you have a value which is relative to average in the league, by definition. You still have nothing to do with replacement and you shouldn’t care.

Now, someone might ask you, “How much should you pay so and so?”

I could answer that with no regard for the term replacement by saying, “Well, he is 2 wins above average, and we pay 5mm per win.”

Then they would say, “Well how much do you pay the average player?” I might respond, “10 mm.”

You would ask, “Why?”

Now, I would have to think a minute, and ask myself the same question. “Why are we paying average players 10mm. Why not 5mm? or 20mm?”

Now comes the point where you have to think about the replacement player. You don’t have to use that term. You simply have to ponder this: There are players who are on the fringe of being out of baseball and players in the minor leagues who are on the cusp of being major and minor league players, according to the Lords of Baseball. We have to assume that their collective talent is just below that of the worst MLB player - the 500th or so best pitcher or position player int the world. I know that they will gladly play in the MLB for the major league minimum or thereabouts. In fact, those minor leaguers have little choice and unless they become better players (not fringe) they are not going to get much of a raise if at all. They are going to bounce back and forth getting paid the MLB minimum when they do play in the show.

Now that I realize that, all I have to do is figure out what their value is relative to the average MLB player either overall or at each position (it gets a little fuzzy which way is better, but the answer is not going to vary all that much because of the fungibility of most positions). That is relatively easy to do.

So I do that and it turns out that the pool of those fringe players is 1.93 wins worse than the average MLB player. I pay them, say $518,000 on the average. So now I know how much to pay my average player or any other player, assuming I know how much a win is worth. And you can attach any number to what a win is worth - it doesn’t matter. It could be the league average (that may not be a good way because that includes lots of inefficient contracts and pre-FA and pre-arb contracts make it difficult, etc.), it could be what you think a win is worth to the average team in revenue, it could be what a win is worth to your team in revenue. It could what a marginal win is worth to your team given its current w/l record and chances of making the playoffs, etc. It doesn’t matter. The only thing matters is how many wins the average is compared to that player whom you can pay the minimum to play.

And the important thing is that it does not matter how that fringe player gets his value - defense, offense, base running, etc. Who cares? IT. DOES. NOT. MATTER. It only matters that they are 1.5 or 1.7 or 2.1 wins (per 150 games or whatever) worse than the average player. How that works out in terms of offense, defense, base running, probably depends on the position, the era, etc. But it doesn’t matter. I mean, if you are curious it might be interesting to see how the defense and offense and base running (and other things like SB/CS, catcher framing, etc.) shake out when it comes to fringe players, but it doesn’t matter. The concept of “replacement player” has nothing to do with component value - only total value.

To tell you the truth, I am not even crazy about using “R” as the baseline in stats (as in WAR). It is really only necessary in terms of figuring out the monetary value of a player, and to some extent whether one below average player is “worth” more than another below average player if they have different playing times. And even then, who cares? No one really cares about the bad players. All of the debates center around the good players. You never read about a “WAR” controversy about who is the worst player. If it were me, I would center all the stats around average. I have been working with valuing players for 25 years, and I have nothing in my database which uses “replacement” as a base line. I don’t need to. Why would I? I can’t add “2 wins” to a player’s wins above/below average if I want to know how much to pay a player? I mean attacking WAR because someone does not understand the concept of replacement or because they think it is a fuzzy or even a wrong concept is a complete straw man. That is because you can simply change WAR into a WAA and it won’t change a thing! If Trout were .8 wins better than Miggy in WAR he is going to be exactly .8 wins better in WAA (OK, maybe that is not true if the WAR people are using different replacement levels for different positions - are they?-  but it is close enough)!

I agree with MGL’s basic points:

- what we really care about is comparing to the average
- the issue of replacement level only comes into play when playing time doesn’t match the player’s talent level

The Trout v Cabrera issue shouldn’t be complicated by talking about “replacement level”, since their playing times are going to cancel out.

Strasburg pre-2013, Ben Sheets, etc… those guys is where it matters, when their playing time is not consistent with their performance levels.

Replacement level really helps in trying to frame the monetary issue… but it’s NOT NECESSARY. It is helpful, and it will save you from making crazy suggestions.  Should I say “12MM$ for Francoeur”?  Yes, I should.  That’s when replacement level helps, as it makes you stop doing something silly.

I’m not sure what you mean when you say that it is helpful when playing time does not match talent level. Say that Trout only plays half the year and Miggy plays the full year. And let’s say that they play at the same rate. Trout will have 4 WAR and Miggy 8 (or whatever). At the same time, Trout will have 3 WAA and Miggy 6. Has Miggy provided 3 or 4 wins of value more than Trout? I don’t think that has a clear cut answer.

Taken to the extreme, player A plays a whole season at league average. He has 0 WAA. Player B plays a week at above average, and has .1 WAA. Again, I don’t think that it is clear who provided more value, other than in a monetary sense. Obviously it costs a lot of money to provide almost a season worth of league average play.

But the question, “Who provided the most value to their team,” I don’t think is a specific enough question to have a clear cut answer.

Am I wrong?

This may be of interest:

http://www.baseballprospectus.com/article.php?articleid=21773

MGL: this is why having a single-dimension answer will leave us wanting.  Hence, my preference to show things as an Individualized W-L record (i.e., The Indis).

In an extreme example,
league average player in 162 games
v
the guy who has 1 PA all year: a grand slam walkoff

The league average player will have an Individualized Won-Loss record of 5-5.

Moonlight Graham over there would say have a W-L record of 0.2-0.

Which record is “better”?  I dunno, depends what you are asking.

The guy who is 5-5 is 0 wins above .500, but 2 wins above .300.

The guy who is 0.2-0 is 0.1 wins above .500, and 0.14 wins above .300.

And what happens to the 9.8 other “games”, the difference between the guy who plays the whole season (5 wins, 5 losses = 10 “individualized” games), and the lone ranger’s 0.2 individualized games?

Hence, I’m happy simply reporting 5-5 v 0.2-0, and I’ll let everyone else decide their questions and their answers.

“a grand slam walkoff”

ehhh… I added too much information there… let’s just say he hit a random triple

And we already have experience with these kinds of questions and answers.  CC’s 11-2 run, compared to whoever he was up against for the Cy.

Or a reliever with a 1.something ERA v a starting pitcher’s 2.something ERA

etc, etc

We just don’t think of it in the same way for hitters, when really it’s the same idea.

“You never read about a “WAR” controversy about who is the worst player.”

No controversy, only consensus.  His name is Willie Bloomquist.

If Trout were .8 wins better than Miggy in WAR he is going to be exactly .8 wins better in WAA (OK, maybe that is not true if the WAR people are using different replacement levels for different positions - are they?-  but it is close enough)!

This is completely in error. Everybody’s WAR implementation is sensitive to playing time. There’s a 50 PA differential between them. (And WAA and WAR will return different ordinal rankings as well, when used to evaluate real performances.)