I posted this over at Bullpen Banter, but since it is obviously sabermetrically based, I figured the saberists here would be interested in it. To be clear off the bat, this construct doesn't change the WAR output, it just cleans up the inputs (in my humble opinion).
Update: With a little help from Renè Saggiadi and Colin Wyers, I made a small update regarding fielding. Renè first mentioned, there is no such thing as a "replacement level fielder." And Colin elegantly mentioned, that "replacement level teams in the wild are average on defense." Thinking about this further, both Renè and Colin are absolutely correct. Mathematically, one would only need to adjust RAA to RAR to leave the fielding metric as average.
Today's most popular statistic derived from sabermetric principles is Wins Above Replacement (WAR). In reality, to call WAR a statistic is misleading, because it’s really a concept which can be applied in many different ways. Tom Tango, the creator of the concept said, "WAR is wins above replacement. Replacement is … the talent level for which you would pay the minimum salary on the open market, or for which you can obtain at minimal cost in a trade."
Today, there are two widely available WAR constructs, which can be found on Sean Forman's Baseball-Reference and on Fangraphs. Similarly to Fangraphs, Baseball Reference uses a construct of WAR, created by Sean Smith of BaseballProjection.com, that adds together a player's runs above an average player in the areas of batting, base running, and fielding with adjustments for position difficulty and something called "Rrep" (on B-REF) or "Replacement" on Fangraphs. While some find the difference between these two statistics bothersome, I find it quite refreshing that two separate sites can use the same concept as a platform and have varying result.
I am not going to get into what replacement level should be. Rather, I want to discuss an alternative to the use of (or lack of use of) "Replacement" and "Rrep" runs within the WAR construct. Before I dive into this any further, I want to share where I first read Tom Tango musings on the subject. He writes on 2/06/2008 (emphasis removed from the original):
"To convert wOBA for a hitter into wins: (wOBA - .338) / 1.15 * 700 / 10.5 will give you wins above average. (The .338 is whatever the league average wOBA is, which is EXACTLY equal to whatever the league average OBP is; 1.15 is the relationship between wOBA and runs; 700 is the number of PA per 162 games; 10.5 is the relationship between runs and wins.) Adding in the wins above average at the position plus the positional adjustment gives you wins above average per 162 games. Add in the replacement level, and that gives you WAR per 162 games. Simply multiply that number by the percentage playing time you expect (no more than 90%, typically at 85% for regulars), and you have your WAR."
There have been some variations from this blueprint. First, and most notably there is no mention of fielding runs. That is something that the sabermetric community has added as it has become more comfortable with its fielding metrics. Secondly, because we aren't projecting WAR we don't need to multiply by a percentage of estimated playing time, we can instead replace "700 PA per 162 Games" with the actual number of plate appearances within the formula. Lastly, and I'm not sure where, when, or why it happened, but the standard amount of runs per win has been decreased to 10.*
Why don't we cut out the middle man? What I mean is, let’s stop calculating Wins Above Average (WAA) first and instead jump straight to WAR. Theoretically, the change would be simple. All one would need to do is, subtract replacement level from a given player's wOBA rather than subtracting the league average wOBA from a given player's wOBA (which gives us Runs above Average or RAA**). And… that’s it. Well, that’s not really it, but now, there is no need to add "Replacement" or "Rrep" into the equation and we are left with:
Batting Runs above Replacement (RAR)
Fielding Runs Above Average
Positional Adjustment & League Adjustment Runs
Why go through all the trouble if the results are the same? Before I answer, let give you a hypothetical.
When an organization calls up a replacement player he is a 0 WAR player. He has 0 Batting Runs Above Average, 0 Fielding Runes Above Average, 0 Replacement Level Runs Added (for playing time) and 0 positional and league adjustments, for the time being. Theoretically, after any given period of time he'll be +X for his playing time and -X for the other components (batting runs above average, fielding runs above average, and the adjustments). Thus, keeping him a replacement level player.
Under my WAR construct, a replacement level player's plate appearances and fielding opportunities are inherent within the components.
I can see a strong rebuttals. First, the league average is an objective standard. Using runs above average gives the saber community a consistent starting point from where we can then apply our subjective replacement level.
I would counter with an extremely strong arguments with simplistic rationale. Wins Above Replacement should use components that use Above Replacement baselines. Remember, this is just an idea, the current construct works fine. The change would, in my eyes, make thing more efficient and easier on the eyes of saber beginners.
* Does anyone know when/why this happened?
**To be clear, I'm not advocating to get rid of Batting RAA because it is an extremely valuable statistic. It’s easy to calculate, manipulate and use across leagues. I'm just advocating not using it as a basis for WAR.
Ed: JD Sussman is one of the best saber minds out there, and can be read at bullpenbanter.com. Thanks for the contributions, JD! -jbopp