Quote:
Originally Posted by 702guy
So WAR is designed to be a comprehensive stat to determine how valuable a baseball player's production has been. And it does so by comparing each player's production to a hypothetical "replacement-level" player, which is defined as how good players that are relatively freely available might be.
Well that's the concept, but to compute WAR in reality, we have to put some numbers - how good are these replacement-level players at each position? Instead of trying to define this independently, basically all versions of WAR does some variation of this: they take average batting production, apply some discount (since replacement players by definition aren't as good as average), then apply positional adjustment and apply defensive performance. So WAR is:
Player's Offensive Production - Replacement Level Offensive Production + Positional Adjustment + Player's Defensive Production (above average).
So the idea of positional adjustment is tricky because it's trying to deal with the fact that equivalent offensive production at certain positions are more valuable than others, due to scarcity. Some positions are inherently harder to play and also certain positions exclude certain types of players altogether. 3B/SS are practically impossible to play for left-handed throwers at a high level and 2B/C are in some theoretical sense playable for lefties, but in practice this doesn't happen at higher levels of competition.
So one might ask, could we just use the average offensive production at each position? The problem is that this is highly volatile - you're penalizing players that happen to be playing at a loaded position. Also, it's fundamentally unfair because some positions might naturally have better athletes. For example, what scouts look for in SS and 2B has historically been similar, with the one key difference is that for SS, you want a better arm. But arm strength has some correlation with power, so you might end up in a situation where SS on average are better hitters than 2B in any given year, simply due to teams putting better overall athletes there, even though everyone would agree that 2B is a lot easier to play.
So people also looked at the relative difficulty of positions - for a set of players that switch between comparable positions (for example, 2B & SS, 3B & SS, 3B & 1B, CF & LF, CF & RF) and how do their defensive performance look like, to make sure that we don't accidentally penalize athletes playing a more difficult position simply because the position has better overall athletes playing them. The important thing is that this adjustment is not designed to predict how you would do if you were to move position. There's some interchangeability, but they aren't entirely fungible. There are also limits to this, so the positional adjustment has historically incorporated both this as well as relative offensive production.
For example, a SS can generally play 3B and 2B. a 3B can generally play 1B. SS/3B/2B can generally all play LF/RF. CF can generally play LF/RF. But there are a lot of position-specific skills, not to mention the whole left/right thing. Catchers generally won't really be good anywhere, though they can usually be good-enough at 1B. A lot of athletic, but smaller players can be adequate at 1B, but won't be elite as the positional adjustment might suggest.
But I think Lawnmower Man is trying to say that the positional adjustment is some sort of an effort at estimating how a player might do when changing positions:
Quote:
Originally Posted by Lawnmower Man
The relative values of the defensive positions were set long ago and assume interchangeability. In other words, the WAR model assumes, e.g., that an average 1B could play SS but would be a much worse fielder there on average. That's literally where the positional adjustment comes from.
But it's not - that's not the question it's designed to answer, but rather the other way around - some level of interchangeability does impact the position scarcity and the actual value of production. For example, even if the average SS hits better than the average 2B, if the average SS can just be moved to 2B and his defense would improve and the average 2B being moved to SS would see his defense suffer, the same offensive production is still more valuable at SS. And the replacement level should reflect that. But the converse is not true - simply knowing the positional adjustment says nothing about how a player might do when they switch position. And the idea that WAR as a model says that an average fielding left-handed 1st baseman would be around 20 runs below average as a SS because that's what the positional adjustment is - I don't really know what to say, that's just ridiculous.