Comments on: B-R Blog: About oWAR and dWAR and responding in a timely manner This and that about baseball stats. Tue, 16 Jul 2013 17:01:55 +0000 hourly 1 By: kds Fri, 28 Jan 2011 19:35:48 +0000 # 31, 32. One area in which the concept of replacement level is vital is player transactions. Since we always should be able to get a replacement level player at league minimum salary we never should pay anything for a worse player, as we would never knowingly play him, or even put him on the roster. For free agent signings, trades, etc., we figure how much they are worth by how far above replacement level the player is.

A replacement player does not have zero baseball value. There are enough of them that you should never pay for less than that value. Suppose the average team/league scores 4.5 rpg; pretty close to where we are now. A player with average fielding value who hit in the middle of the lineup and never missed a game would be worth about 81 runs on offense for the season. We think a replacement player is about 2 wins or 20 runs worse than average. So he would be worth about 60 runs if he played the whole season. Not worthless, but easy to get those 60 runs, so you would never settle for a player worth 50 runs since you would have to pay him the same ML minimum wage. So, in effect we ignore those 60 runs and only start counting above that level. For comparisons for the HOF and MVP's we could well use WAA, (wins above average), or an even higher standard; WAG, (wins above good.) (But now we are far out on the right tail of the value distribution curve.

By: Tangotiger Thu, 27 Jan 2011 18:51:58 +0000 "So, while this may be a stupid question, I have to ask: Why not both?"

You in fact could do both... as long as it's clear that you cannot do WAR = oWAR + dWAR.

You have 4 components:
1- hitting above average
2- fielding above positional average
3- positional value relative to neutral position
4- playing time value above replacement

These are the questions you have to ask yourself as a user:
A. Would I prefer seeing these as 4 separate columns? 3 columns? 2 columns? 1 (i.e., WAR)?

B. Do I need to have these be non-overlapping? Or can I overlap and have for example Item 3 be in multiple columns?

C. Once I have decided on these constraints, how can I best combine these 4 items?

So, ask yourself these questions, then show us your answers, so we can see how users would prefer to use this data.

By: Dvd Avins Wed, 26 Jan 2011 21:10:53 +0000 "In fact, a dWAR that combined rfield and rpos would really be defensive wins above average."

That's only true if you set zero equal to a DH's rPos.

By: Michael E Sullivan Wed, 26 Jan 2011 20:25:34 +0000 "It is true that dWAR here is really dWAA and should probably have separate leaderboards for each position like TZR does."

In fact, a dWAR that combined rfield and rpos would really be defensive wins above average. rPos looks at what each position does relative to average. You have to add it to rRep to get a number for offensive replacement value at a position. (this assumes that offensive difference between the positions is on average about the same in runs at replacement level versus average level.

dWAR is just rfield converted to wins, in the same way that WAR is all the run numbers converted to wins.

The problem with dWAR and oWAR is that they are misnamed. The names imply that they are measuring defensive value and offensive value respectively. But I would agree with twisto and tango that the positional adjustment is more properly thought of as defensive value, even though B-R uses the difference between average offense at a position to *measure* this value. The method of measurement means that rPos is essentially using the market consensus of GMs/Managers to determine the defensive value of playing a given position at an average level.

I understand what Sean was trying to do, which is to split out the one number about which there is a lot of legitimate skepticism from the rest of the WAR components. But the names are misleading. And worse, they suggest categories that I'd love to see on the site! Which I think is why so many people thing dWAR *should* mean total defensive value, because that's what the name implies.

Why not call what is now called dWAR TZWAR? how about oWAR, pWAR and tzWAR, with a dWAR which is pWAR+tzWAR.

You can't realistically compare players with just the current oWAR anyway, you have to add something for their defensive ability or lack thereof, it's just a question of whether you agree with TotalZone, or want to use some other metric, or just an eyeball test in its place. Personally I think totalzone is way better than an eyeball test, I just have to assign a lower level of confidence in it than I do other numbers.

By: Johnny Twisto Wed, 26 Jan 2011 06:05:11 +0000 Self-correction: Darrell Evans was really not close to being a career 3Bman.

By: Johnny Twisto Wed, 26 Jan 2011 05:59:57 +0000 I think you have it John. It's really just a matter of what you want to know about the player.

If we're just comparing two players on the offensive side of the ball, we don't necessarily care what position they play. Who's a better hitter, Joe Mauer or Ichiro Suzuki? In that case you just want to look at batting + baserunning.

What if you buy into the general WAR framework but don't like Total Zone's results? Now I better realize what Sean's initial impulse was in making the oWAR/dWAR division. So in that case, take the player's oWAR (as currently presented), and make your own defensive adjustment, or use UZR or whatever.

What if you want to know which players have the most defensive value? Ernie Banks moved from SS to 1B during his career, did he provide more value between those two positions than a career 3Bman like Darrell Evans? Then you want the Rpos + Rfield (positional value plus defensive performance).

Or if you just want to know who are the best fielders, regardless of position? Does Albert Pujols have a better glove than Troy Tulowitzki? You just look at Rfield (Total Zone). (Well, preferably, you look at other defensive numbers too, over a couple years, plus subjective opinion....but I digress.)

All of these are valid questions and things that WAR can be used for, and it's great that all the components are provided so we can use the pieces we want. I still think that the positional value belongs with dWAR, because WAR is attempting to measure value. How much is this player worth on offense, how much is he worth on defense? Very recently we have had a commenter on this blog scoffing at Derek Jeter appearing to be the worst defensive player of all time, according to dWAR. Well, he had a misunderstanding of the stats which goes beyond simple WAR accounting, but Jeter's defensive value has obviously not been close to the worst ever, and this would be more apparent if his positional adjustment was included in dWAR.

I really think Sky's idea from #1 is the way to go. The replacement value doesn't really belong either in defense or offense. "Replacement players" don't all have average gloves. I'd classify Shelley Duncan as a replacement player; his bat is average but his glove is below. What makes someone a replacement player is based on the totality of their contribution. So show offensive wins above average (Rbat+Rbaser+Rroe+Rdp), show defensive wins above average (Rfield+Rpos), and then Rrep, the playing time count, stands on its own.

By: John Autin Wed, 26 Jan 2011 04:47:19 +0000 I don't totally grasp the debate, but it sounds like one question is whether the "positional adjustment" should be reflected in dWAR or oWAR.

So, while this may be a stupid question, I have to ask: Why not both?

If we want both oWAR and dWAR to work as stand-alone measures that express the win values of those two components of a player's performance, then don't they both need to be expressed in terms relative to the position played?

It seems to me that even if I want to measure the win value of Dan Uggla's offensive production alone, I still have to adjust for the fact that he plays 2B, because that gives the Marlins a better chance to have a "plus" hitter at whatever less-challenging defensive position Uggla would otherwise play.

The positional adjustment in dWAR should reflect the relative importance of the different positions, e.g., a SS whose defense is 20% above average keeps more runs off the board than that same caliber of defense at 1B.

Granted, making a positional adjustment to oWAR brings selection bias into play; for example, one could argue that Alex Rodriguez could have played a competent SS for a few more years than he did, so it's not fair to "punish" his oWAR for his manager's decision to put him at 3B. But to me, that's just part of the game, no different than the selection bias involved in who gets the chance to play, period.

If I've completely missed the boat, tell me. I'm a slow learner, but I'll try to understand.

By: Johnny Twisto Wed, 26 Jan 2011 04:14:17 +0000 Duke, yes, the traditional defensive stats are affected. But most of the new defensive systems are not based on those stats (which is why, in theory, they are an improvement). They count (or estimate) how many balls are hit into a particular area of the field, and how many of those are turned into outs.

I don't think a 2Bman (or anyone) should be negatively affected by a teammate poaching balls which either could field, but I don't know the systems well enough to guarantee that this is always the case. Even if it is impossible to completely erase the results of poaching from the advanced stats, I am sure the impact is far, far less than it would be for traditional defensive numbers.

By: John Bowen Wed, 26 Jan 2011 03:44:21 +0000 I like your way Sean.

Inherent in a player's offensive value is their position.

Plop Mike Piazza behind the plate and he doesn't instantly become a good defensive player because his Rpos+Rfield is positive.

Mike Piazza's positional value is that Eric Karros and John Olerud can be in his lineups with him instead of, for example, Ben Davis or Greg Zaun.

By: dukeofflatbush Wed, 26 Jan 2011 03:36:56 +0000 Johnny,

But lets say the out was made, are you saying the 2nd baseman does not get affected?
But doesn't it decrease his total chances and assists?
And lets say, as a 2nd baseman, I play next to a very aggressive 1st basemen, whom makes an unnecessary 'Cabrera-like' play once every other game.
Wouldn't, through no fault of my own, make it appear on paper, like I had less range?
Like I said, I am not sure on how dWAR is compiled.