Interesting article on Defensive WAR explanation

[weams]

To qualify them. I don't know what to do about the unreliability of WAR with the fielding component included. I guess it is still better than thinking Tony Bautista was good. 

[DrungoHazewood]

18 hours ago, Can_of_corn said:

Are you saying to use eyes to supplement the metrics or instead of the metrics?

I always wanted someone to do an experiement where they take a bunch of people and have them watch hundreds of games but give them zero information on the hitters.  No OPS, no batting averages, no RBI, no WAR, no doubles, no nothing.  You'd probably have to do this with college or the minors or a foreign league where the players wouldn't be previously known by the observers.  The people can take notes, but won't have access to any stats all year long.  Then at the end of the season they have to rank all the hitters they observed.  I think that at a high level the rankings would be pretty good, but as you drilled down into individual cases there would be no shortage of absurdity.

That's basically how we've judged defense for 150 years. 

Okay, so maybe people have paid attention to fielding percentage and errors, so perhaps we give our experimental subjects strikeouts and batting average.  But then you're not too far from how we judged hitting from the beginning of time through, say, 1980.

[DrungoHazewood]

Going back to the original post, this has always been my criticism of dWAR and oWAR.  There is no such thing as defensive Wins Above Replacement or Offensive Wins Above Replacement.  There are only total wins above/below replacement.  A player's performances in total combine to define their contributions above or below replacement level, or average or whatever.  You can't take that out of context and then partially reassemble it and have that mean something. 

I mean, sure, if you contort and wrap your mind around several esoteric concepts I suppose it can work.  But in reality people gravitate towards oWAR and dWAR as distinct things that add up to WAR, when they're not that at all.  They both contain positional adjustment, which means they're not additive.  On their own oWAR and dWAR don't really mean anything.  

The much more straightforward thing to do is to just look at runs above or below average for each component (fielding, hitting, baserunning, pitching) and if you want to convert to WAR then add all those up, divide by 10 runs/win, and add in the replacement level and positional adjustments.  I understand what Sean Foreman was trying to do, but what's happened is far more confusing for the vast majority of people.

Just don't use dWAR and oWAR.  Everyone already knows Keith Hernandez wasn't as good with the glove as Ozzie Smith, that's why he was at first base (ignoring that he's left-handed).

[LookinUp]

On 12/5/2018 at 2:24 PM, weams said:

I still think the defensive component if far from reliable. 

It's probably less bad than you're acknowledging. Statistics are all about the components and the context. If you don't understand that part, you get back to the lies, damn lies and statistics line that people love to throw out.

I think the data and components necessary to discreetly judge fielding almost certainly exist. I think they also can judge park factors, shifts, etc. separate and apart from an individual fielder's talent. It's not THAT complicated, but it gets convoluted to us somewhat casual observers when the creators of these statistics mislabel what they're doing, and make no mistake, that's all they did here. They took raw data on runs, applied adjustments and mis-named oWAR and dWAR.

Great article and thread.