Not sure what you're after. Win Shares already has a "built in" replacement level of .200, which is already taking some criticism as being way too low. How could you possibly want to go even lower than .200?

I don't think we were referring to that KJOK. We don't want to adjust for it being lower, we need to adjust for not really accounting for a replacement level.

Take a look at the article (not sure where, don't have the book handy right now) where he discusses replacement level. I wouldn't really say that it's .200 (yet, need to read more), so much as he doesn't really account for it.

If it's .200, then we need to account for moving it up to .300 or .325, which is more realistic.

If it's .000, then we need to account for that also. I guess the key is figuring out where we are right now.

On another thread (the WS thread on Clutch Hits that's going right now), I posted some initial calcs, assuming WS repl. is set at .000. Those gave us about 6.5 WS/375 outs as replacement level for hitters, and about 7.5 WS/200 IP as replacement level for pitchers.

But Craig pointed out (in an e-mail I think) that we should be using PA as the context, not outs, based on the Crossing the Rubicon essays. I'm still trying to work through all of that in my head also.

So I guess I'm looking for some help in coming up with these answers. Sorry if this doesn't make a lot of sense, that's probably because this thread is really a spin off from the Clutch Hits thread.

KJOK -

Win Shares doesn't rely on replacement level - it relies on the empirical observation that so-called "marginal runs" have an approximately linear relationship to wins (for all except *extremely* bad or good teams, which would include only teams like the 1899 Cleveland Spiders).

In a sense one could consider the definition of marginal runs as pegging a replacement offensive player at .52 times the league average of runs created, and a replacement defensive player (pitcher or fielder) at 1.52 times the league average of runs allowed, since these are the levels at which a player is credited with zero win shares.

Yes, James SAYS that WS doesn't have a replacement level, but in reality "marginal runs" can only be above a certain 'replacement' level, otherwise it would be absolute runs...

Well, taking Robert's cue I am also now looking at the spread of performances for 19th-century baseball.

I am finding (very preliminary right now) that "replacement levels" for the 19th century are in line with 20th century levels but haven't done the calculations to my liking.

To revive an ancient thread--I've been looking into the question of historical replacement levels recently. My methodology has been to use the average production of the three worst starters at a position in the league as a baseline for replacement level and then tweaks. I've found that for the 1893-1906 time period (I haven't gone back yet into the pitchers' box era), the outfield positions are backwards: the worst three center fielders in the NL each year averaged 0.9 wins per 162 games below a player hitting and fielding at the league average, while the worst three corner outfielders in the NL averaged 2.2 wins per 162 games below a player hitting and fielding at the league average. I am totally baffled by this result and have no idea how to deal with it. My questions are the following:

1. Has anyone else tried to compare replacement levels for center and corner outfielders during this time period? Is my finding empirically correct?
2. What could be the potential reasons/explanations why the worst three CF would be consistently better than the worst three corner OF?
3. If this is the case, what is the best way to set replacement level for this time period? Should I just use one replacement level for all OF?

I thought we discussed this somewhere a long time ago, but it's roughly the same reason that the replacement level at SS in Little League is higher than CF.

The best players/athletes in the 19th century generally played SS. Fielding wise, after Catcher, you had 3B/2B/1B as the more important defensive positions.

The OF mainly for guys that couldn't play the IF, but teams did tend to put their best fielder/athlete in CF, who would also tend to be their best offensive player.

You might want to break down RF & LF and see what you get. Teams generally put their absolute WORST player in RF as RF tended to be the short field in most parks back then (teams even played good hitting pitchers in RF just a few years earlier).

That definitely makes sense. So given that knowledge, what's the best way to calculate replacement level? I guess you would need to know the defensive "translations"--eg, a -5 in CF is equal to a +3 at RF or whatever--to put all outfielders on an equal footing, and then use the same rep level for all of them. Does anyone have that kind of data available?

I see that according to Tangotiger the gap is about 9 runs a season now...but that has probably changed over time, no?

Also, KJOK, is that only true for the outfield, or all across the diamond? Would replacement level for SS in the 19th century be "artificially" high for the same reason?

Some interesting questions:

That definitely makes sense. So given that knowledge, what's the best way to calculate replacement level? I guess you would need to know the defensive "translations"--eg, a -5 in CF is equal to a +3 at RF or whatever--to put all outfielders on an equal footing, and then use the same rep level for all of them. Does anyone have that kind of data available?

I don't know of any data that would be readily available for this analysis.

I see that according to Tangotiger the gap is about 9 runs a season now...but that has probably changed over time, no?

Extremely likely that it has changed over time - but to what?

Also, KJOK, is that only true for the outfield, or all across the diamond? Would replacement level for SS in the 19th century be "artificially" high for the same reason?

Depends I guess on your philosophy of player analysis - but I would tentatively say yes, if you're measuring league replacement level vs. positional replacement level.

Would replacement level for SS in the 19th century be "artificially" high for the same reason?

Yes - unless you think SS's weren't good fielders then. I don't have access to the Sabermetric Encyclopedia right now . . . but IIRC guys like Wagner, Davis, Tinker, Jennings (at SS) have 'positional average' OWPs (using entire ML, not just NL for NL/AL years) of around .470, whereas even with the SS hitting boom of the 1990s-2000s, Jeter's is something like .425.

When I get a chance, I'll post the numbers across positions - but as late as the Cobb/Speaker years, CF league average OWPs were typically in the .575-.580 range, for the corners it was around .600. I think that's one sign that we overrated the pre WW-II CFs (Averill, Carey) a little bit back when we considered them.

The CF as very important defensive position thing didn't really start to take hold until after WW-II . . . and what we remember from the 1970s and 1980s certainly isn't the historical norm. That was brought about by the huge stadiums of that era and artificial turf - you can already see CF becoming a less defensive and more offensive position over the last decade now that the stadiums are more 1950's-ish than 1980's-ish.