By the way- this guys is killing it in NO QUITTERS.  I argued with him over two players' potential for the HOF.  My argument was for a SS/3B/2B vs. his CF.  I made the contention that my player had a better OPS, was a much better base runner (SB and on the paths during live ball), had much better career numbers that his CF, and could play multiple positions.  Additionally, the CF had won gold gloves in a league with less competition for that position.  He comes back with this and blows my argument away. 

This is all good stuff - I do a lot of these same things when evaluating players.  I think if you dug a little deeper you would see that the standard positional adjustments from MLB aren't accurate here.  Shortstops are considerably more important, and catchers are significantly less important.  The whole WOWY concept that positional adjustments is based on is easy to do in whatif.

I cleaned up the spreadsheet and put it up on zoho.com
Username: whatifsports
Password: whatifsports1

For anybody who is interested. The document include a 2nd sheet which is a glossary explaining each column. Should be pretty user-friendly, but let me know if you have any comments. Anybody is welcome to COPY and modify, just leave the original spreadsheet as is, so everybody can copy and use it.

As far as different values for league averages go, you definitely could do that. To be honest I just picked 4 random seasons in No Quitters and used that to guess values that seemed "about right." If somebody wants to figure out whether historical walk rate in their league is 8.8% or 9.2%, they are more than welcome to. In a sense, "league-average" is a year-specific concept, but it honestly doesn't really matter all that much. At least in my leagues, there hasn't been a "deadball era" or a "live ball era." Since the engine is mostly the same, there's not going to be huge fluctuations year to year that warrant the insane extra amount of work it would be to do this season-by-season for every player. Even the engine changes that affected home runs didn't do THAT much to total home run rate, it mostly just lopped off some home runs from the top guys. In a few seasons, I'll probably have to adjust that number down, as it's kind of the midpoint between S8-10 and S16-17. But, in all honesty, since you are mostly comparing guys who played in the same seasons anyway (not like comparing real-life baseball 1998 to 1968), everybody is running off the same baseline and at most you are going to make guys 2-5% more valuable or whatever. So I might play around to get some of the rates to be a little more accurate, but it's never going to warrant the time expense of doing it season-by-season.

And also, I have labeled the sheet "Win Shares," but the methodology is actually aping WAR or WARP, as Win Shares are a bit different. I just figured more people would understand that name.

I have no idea if the positional adjustments are accurate or not because I just took the MLB ones and threw them in here. Positional adjustments aren't based on a WOWY concept; WOWY is "with or without you" and useful for analyzing stuff like catcher defense. What you would do is figure out what the average or typical player at a position puts up offensively and compare that to league average. Upon a very cursory examination, it actually seems that (at least for this year for shortstops in No Quitters), they might be pretty close. I took the 13th, 15th, and 17th-ranked (by OPS) shortstops in No Quitters and found they were about (-5), (-7), and (-10) in offensive output respectively compared to league average. So it seems that +7.5 is a decent starting point for shortstops. When I have a little more time, I'll do a more in-depth analysis and possibly tweak the positional adjustments.

Actually, the positional adjustments most commonly in use on sites like fangraphs are in fact derived precisely from WOWY methods.  The good thing about WOWY is that it ignores the offensive contribution and focuses entirely on defensive differences from players that play multiple positions.  Offense shouldn't play into it, since it is independent of fielding.

http://www.insidethebook.com/ee/index.php/site/article/complete_war_2008/

http://www.fangraphs.com/blogs/index.php/position-adjustments/

Posted by kahrtmen on 11/16/2010 1:01:00 AM (view original):
Actually, the positional adjustments most commonly in use on sites like fangraphs are in fact derived precisely from WOWY methods.  The good thing about WOWY is that it ignores the offensive contribution and focuses entirely on defensive differences from players that play multiple positions.  Offense shouldn't play into it, since it is independent of fielding.

http://www.insidethebook.com/ee/index.php/site/article/complete_war_2008/

http://www.fangraphs.com/blogs/index.php/position-adjustments/

(1) Comparing player performance at multiple positions is not a WOWY method, but I really have no interest in debating those semantics and I'm sure nobody else cares.

(2) That process is necessary in real-life baseball, because in the MLB, centerfielders are better hitters than 2B. In HBD, they are not, so you are talking about a vastly more difficult process (and one that is significantly harder, do you really want to sort through a few hundred seasons of data from HBD players by hand to do these adjustments?)  for little to no increase in accuracy. In MLB, the question "is it harder to play CF or 2B?" is necessary to ask. In HBD, it isn't. The positional requirements are well-known in HBD, so a simple analysis of offensive production is a very reasonable proxy for positional scarcity.

(3) If you really want to do it the most accurate way, it would probably be to simply look up all players with league-average or better offensive production over a 5-season span and calculate what % meet the defensive rec's for SS, CF, etc., etc. Then you would know exactly how scarce each position was and use that to make your positional adjustments. That would take a while and I don't feel like doing that.

If you want to do your own positional adjustments, more power to you. I would love to see what results you come up with and if they actually differ from ones derived from relative offensive production.

Ok, so I finally got around to reading through this (for those who are wondering anyone should care if I read it or not, I'm the owner that was wondering if anyone thought that Thompson was Hall worthy).

I'm still thinking about whether you're double dipping on defense and positional adjustment when evaluating DHs. But I do have one comment to make about your baseline fielding percentages: .980 seems really high for a baseline fielding % for 3Bs. I'd expect something closer to .950.

My 3B's who have gotten at least 300 innings since S14 have been: .995, .972, .939, .977, .970, .950, .995, .970, .972, .952, .968.

So .980 might be a little high. Maybe .970 or .975. Remember that should be average, not replacement-level.

But yeah, a lot of the values were kind of formulated off a few minutes of rough estimation. A lot of it could definitely be improved.

Posted by jtrinsey on 12/19/2010 8:50:00 PM (view original):
My 3B's who have gotten at least 300 innings since S14 have been: .995, .972, .939, .977, .970, .950, .995, .970, .972, .952, .968.

So .980 might be a little high. Maybe .970 or .975. Remember that should be average, not replacement-level.

But yeah, a lot of the values were kind of formulated off a few minutes of rough estimation. A lot of it could definitely be improved.

I checked the lifetime percentages for my current regular 3B in 3 different worlds. They are .969, .969 and .979. BUT ... each one of them has better than recommended fielding rates (and one of them has won 2 GGs). In fact all 3 of them are (or were) recommended by the AI to play SS.

Unfortunately all 3 of my worlds are in that gray area between rollover and the start of the regular season, so I don't have any current "live" sim stats to analyze to see what's a typical fielding %. However, it's relatively easy to check RL stats on bbref. For the last decade the AL and NL have averaged anywhere from .947 to .960, with the .947 being an extreme outlier (the next lowest is .950). So, in RL I think that the average would be something like .955.

EDIT: In real life no one has every recorded a 3B career fielding percentage of .975 and only 4 have reached .970 or higher.

Wheres the park factor adjustments? Lot more work needed on this. Will you use 3 year averages for the parks or just go season by season? You'll have to determine the players ball park factor for singles, doubles, triples, and homeruns for each season of his career and adjust.  I think tweaking the grey ink test would prove much easier.

Love the work that you put in on this. 2 questions...

1. Why not adjust for replacement player level to get a true VORP? Based on the reading I've done, you're probably talking about somewhere around 80% of the league average, depending on position http://en.wikipedia.org/wiki/Value_over_replacement_player.
2. Have you done anything like this to evaluate pitchers?

Re: clutch

You are welcome to do park adjustments by each player. I'm way too lazy to do that, and/or don't have the programming skill to find a way to do it that wouldn't require inputting each season in manually. To get a truly accurate account, you would have to do that. The grey ink test is interesting, but it's also kind of arbitrary and doesn't really mean anything for HBD. The grey ink test means something because there are already tons of players in the real HoF, so saying "90% of players with a Grey Ink score over X are in the Hall" means something. But in HBD, where nobody is already in the Hall, it's kind of hard to do so.

Re: dw

(1) You say, "why not," but I say, "Why?" comparing to a replacement-level baseline is great in real life where there is no salary structure, thus it is illuminating to see the "value" of a marginal win. In HBD, teams often play with drastically different salary baselines, but not necessarily at a disadvantage. In No Quitters, I generally play with a self-imposed "cap" of about 60M, because I like to transfer a lot of money into IFA. Thus, I am not willing to pay nearly the same amount for a marginal win as I am in Happy Jack, where my payroll is like 100M. Also, I believe that, for HoF purposes, calculating the value over an average player is more informational than calculating the value over replacement. To a GM, a player who posts 200 innings of slightly-below league average ERA, or a 3B who gives you average defense and slightly-below league average OPS can be somewhat valuable. However, I don't think that "value" should be counted as a credit to that player's HoF career. Doing things in this way slightly penalizes players who get carried on past their prime and have some bad seasons at the tail end of their career, but I don't see that as a huge problem.

(2) Yeah. It's really simple. I just use:
[ FIP * (IP / 9) ] - [ 4.5 * (IP / 9) ]
To get runs saved above average. 4.5 being about a league-average FIP. Actually it is a little more than league-average, but I bumped it up a little bit to account for the fact that relievers generally have lower FIP than starters. That may or may not be actually true in HBD, and I have a feeling that it isn't true to the same degree in HBD as it is in real life (in real, generally all pitchers are more effective coming out of the bullpen, whereas in HBD they are exactly the same, but I figure there are still some super-effective 30-STA guys who bring that FIP down), so I didn't bump it up by the same margin. If I really wanted to get accurate, I would do a study and calculate starter FIP, and use that, and then relief FIP and use some sort of leverage adjustment when doing relievers. And technically I should adjust for the fact that a run saved is slightly more valuable than a run earned and not just divide by 10 to convert runs into wins. But I'm way too lazy for that.

In general, pitchers are less valuable (by this metric) than position players. I think that 50ish wins above average for a position player should be a lock and 40 should be borderline, whereas it seems 40 should be a lock for pitchers and 30ish is borderline.

Carny,

That's interesting. A couple 3B from No Quitters:

Alex Flores: 72/69/72/75 fielding ratings. Or, compared to recs: +2 / -1 / -3 / +5, which seems to me to be right around average, maybe a LITTLE above. His career Fielding % is .976

Ricardo Chavez: 68/73/71/72 or -2 / +3 / -4 / +2, which seems right around average, although slightly above in the "error" categories of glove and arm accuracy. Career fielding % is .964

Bud Henry: 68/70/76/69 or -2 / 0 / +1 / -1, which is almost exactly average. Career fielding % is .962

Based on that, it seems like: (a) my guy is getting kind of lucky, and (b) I would be comfortable moving down the 3B fielding % to .970 or .965. I based .980 off just sorting league leaders in 3B innings and "eyeballing" fielding %'s. I was probably been thrown by the fact that a lot of people seem to like to play 2nd SS's at 3B.

Let me begin this by stating that I am a novice. And also, not sure that I mentioned it before, but love the work and thinking that you've put into this. Even if I don't agree with 100% of it, it's really solid stuff, and much further than I've gotten on the subject.

Regarding league average vs replacement level
I guess it depends on your objective. I had completely forgotten that you were using these stats as a way to evaluate HOF candidates. With HOF evaluation as your goal, looking vs. a league average makes a lot of sense. For my purposes I've started calculating VORP, FIP, and some other advanced metrics to try and gauge the value of FAs when I'm making budget and signing decisions in the offseason. For my purposes, I think it makes much more sense to look at a potential acquisition vs. a replacement level guy (who should be any guy who can be signed for the league minimum). I feel like looking at a league average player is far too variable, meaning you would then have to determine how to acquire a "league average" player and how much a league average player should be worth.

Regarding your points on pitching
Maybe I misunderstood, but I think you're missing something if you don't include some leverage factor in your calculations. Much like the exercise you went through with different positional values, there's a marked difference between a starter, reliever, and closer. Reading some of Tom Tango's work, but the explaination is here http://www.insidethebook.com/ee/index.php/site/comments/how_to_calculate_war/

jtrinsey - not sure FIP is the best metric to use for pitchers.  I am convinced that WIS pitchers have significant control over their BABIP.  It's the reason that striekout rate is so unimportant.