Somebody please explain this to me. Offensive WAR. Josh Willingham's is higher than Ryan Braun's

[Funketown]

Thurston - I personally just regress the heck out of the defensive metrics and use a little bit of judgment on how to adjust a player's WAR based on their defense. Take Alcides Escobar, who is having a really nice season but has a UZR of ~-15, so only an fWAR of 1.0. I view Escobar as a +5 defensive guy per season, despite the oddball UZR number, so my back-of-the-envelope WAR number for him comes out to more like 2.5.

 

It is pretty tough to judge "replacement level", as in reality it probably varies from year to year and era to era, although I have no idea how much. I think that fangraphs estimates "Replacement Level" as something like 20 runs below average over the course of a season, so you could just subtract 20/600 PA's or whatever to compare to an average player.

 

I'm not sure how you would do the offensive metrics without Linear Weights. You could use something like WPA I suppose, but I don't really care for that approach.

[gregmag]

Thurston -- what metrics do you prefer? Also, what do you want to do about defense, what's your ground for claiming that you can evaluate average performance better than replacement-level performance, and what do you mean by the very loaded phrase "actual production"? For someone who claims to have thought a lot about this stuff, you haven't enhanced your dismissive assertions with many of your insights, and I'd like to know what you're thinking so that I can better evaluate your critique.

[Thurston Fluff]

Thurston - I personally just regress the heck out of the defensive metrics and use a little bit of judgment on how to adjust a player's WAR based on their defense. Take Alcides Escobar, who is having a really nice season but has a UZR of ~-15, so only an fWAR of 1.0. I view Escobar as a +5 defensive guy per season, despite the oddball UZR number, so my back-of-the-envelope WAR number for him comes out to more like 2.5.

 

That isn't a bad idea that has been expressed before. Take a correct number of years for defense and use it over the single season UZR. While I agree that helps I still am not sure how to correctly weight the value of defense to offense though. I used to think defense was worth a lot more than I do now so what I thought would be accurate a couple years ago would not be today. If there is some place that shows the relative value of offense to defense it would be helpful in making WAR useful. I'm not against trying to make a simple number accurately portray a player's value. I just don't think any do so far.

Thurston -- what metrics do you prefer?

 

Depends on what I want the metric for. If I want to see how well someone is doing this season I look at different things than I do if I'm looking at what he may do in the future. IF you give me a scenario of what I want done I can tell you what I would look at but no one metric fits every purpose.

Also, what do you want to do about defense,

 

As it pertains to WAR nothing until there is a way of evaluating it in equal values for the quantity as offense. I don't think one simple number has any value at all if it isn't accurate. So why try to use it just because it is one simple number? I'll just look at offense and defense separately until they come up with a way to properly weight them. Frankly I don't think defense is as valuable as offense to begin with. So when they get around to weighing the value of defense compared to offense then come up with a way to get an accurate measure of defense in a single season (or change it to the appropriate level of three years to every year of offensive stats) I'll take another look at WAR. Until then I'll do the traditional route of looking at more than one metric. Time consuming I know but at least more accurate.

what's your ground for claiming that you can evaluate average performance better than replacement-level performance

 

I can take all the stats for second basemen and average them. That is a hard and accurate stat. I can't take all the AAAA 2nd basemen, put them in the majors and see what their production is. So the idea that replacement level is what any old AAAA player can do is an estimate of what would really happen. While that can work as a baseline I don't think it is the best way to do it. Show me how a player does vs the average production at his position then I can see where his value is vs reality. While both work as a baseline I always prefer real baselines vs estimates based on a made up hypothetical player.

and what do you mean by the very loaded phrase "actual production"

 

It's not a loaded phrase. Replacement level is an amalgamation of what anyone would, in theory, be able to produce if they played in the majors. While that number is based off calculations over a period of time that calculation had to make some assumptions to arrive at that number. There will always be assumptions on what replacement level production really is. For example who is any old AAAA player?

Lets take the number we have today. If anyone can do it then anyone should be able to put up replacement numbers. If anyone could do it then why are there so many negative WAR's? That should, in theory, be an extreme anomaly if replacement level is really what anyone can do. Since it isn't all that rare I have to conclude not just any old AAAA can guy do it. If it isn't accurate it is arbitrary. While arbitrary can work as a baseline I think ones based on actual events are better than arbitrary ones. If we average actual stats from actual players we get an actual number of what an average player actually produced. Now we can compare any given player to a real baseline vs against a standardized hypothetical. One not based on assumptions of an abstract definition of a theoretical players but based on real players producing real numbers. Isn't that a better way of doing it?

For someone who claims to have thought a lot about this stuff, you haven't enhanced your dismissive assertions with many of your insights,

 

Frankly I've done so many times on this site that I'm kind of sick of repeating them again. I simply didn't feel like rehashing it all again which is why I stayed away until now. I only posted that one here because don't think people should be dismissed as uniformed for disagreeing with stat like WAR. Go back and look at some of the other WAR threads and you can find every case I've made several times over if what I posted above isn't enough.

[gregmag]

Sorry -- I missed your name change. I didn't get the "BUC" acronym until I saw this post and put two and two together. You're right, you've made these points a lot, and with all respect you've never moved me very far, although I think your inclination to look at a lot of metrics is wise and well-taken.