replacement level doesn’t have a solid definition, it moves around a lot depending on what people try to argue.

but most consider replacement to be league average defense

Here’s how I’ve seen it explained, so here’s how I think of this. Note that this is the Tango/MGL/Fangraphs crowd version of replacement level position players, not BP’s. They both are about hte same thing, but define it differently.

There is no such thing as a “replacement level hitter” or “replacement level defender,” only replacement level players.

The value of a player is relative to league averages, and is a combination of his offensive and defensive runs above/below average and what position he plays. If that combination hits 22.5 per full season (howver that’s defined - say, 600 PAs or 160 games), then he’s replacement level.

So let’s take a player as an example, say Endy Chavez. CHONE projets him as a -18 hitter/150 games. That’s pretty horrible. Let’s say we consider him a corner outfielder (although he can probably player center). The positional adjustment used at FanGraphs (and not everyone agrees with this) is -7.5/150, since corner outfielders are relatively easy to find on the replacement market.

Now, if we just consider those two things, they he’s a 25.5 player - below replacement level. That’s the case if he has average defense. As Zep says, it is generally assumed, on the basis of some studies, that a replacement-level player (freely available AAA talent/journeymen, etc.) to have league-average defense.

But what about defense? CHONE sees him as a +20/150 defender on the corners. If that’s the vase, then he’s a -5.5 run (we’d rather say -.55 wins, for the sake of simplicity, we’ll stick with a 10:1 runs to wins conversion) playres, or +19.5 above replacement.

Make sense? For pitchers it’s a bit different, but the general lines for replacement is 2.25 wins average. Some adjust differently for different leagues in certain eras. For example, Tango goes through an argument to concretely demonstrate the superiority of Al pitching, so he concludes that the relative replacement levels for the AL and NL to reflect that - so he gives AL nonpitchers a replacement level of -2.5 wins below average, and NL nonpitcher -2 wins.

You’re smarter than me, you tell me. That’s why I qualify my statements above by “this isn’t how everyone does it.” It’s how Tango does it.

We assign a larger penalty to DH because we know that someone who plays 1B, SS, or anything else, has the defensive “ability” to handle DH, and not vice-versa.

I used the generic DH penalty there for the sake of this article, Tango, et. al. give an after the fact “pinch hit” bonus of 5 runs to DHes to account fo the difficulty of hitting off of the bench, so after the fact, the DH penatly is -1.75 wins.

But I do think that using generic positional adjustments across the board makes more sense in general than adjusting from positional averages for the season (as VORP) does, otherwise you get into certain situations, like in the 1950s, where CFers hit on average better than LFers. Does this mean that LFers were more scarce and harder to replace? I’d say no - because we know that CFers can play LF, but not vice-versa.