THE BOOK--Playing The Percentages In Baseball

Tango, is the problem here that Chris is looking at the positional +- as fielding competence rather rather than runs saved? Because that’s what it looks like to me.

I won’t speak for Chris.

Tango, fair enough. I did always question why UZR/150 was scaled to games rather than opportunities. Is there any particular reason for that? A SS will still get more opportunity per 150 than say a 3B.

I presume the /150 is because that’s how people scale things.

We know what +12 runs means.  We don’t know what +.020 runs per play means.

Tom, do you have a link to what Chris wrote?

JD,

UZR/150 is not scaled to games.  It’s scaled to “defensive games”, which is based on opportunities.

It was in the Primer forums.  But, this is a pretty standard position by Chris, as you can see by the comments in this blog on related threads.

http://www.insidethebook.com/ee/index.php/site/member_posts/?name=Chris%20Dial

David: I think he was talking about the fact that SS have many more opps than 3B, then the denominator should be something that can link the two.  A SS “defensive game” has more opps than a 3B “defensive game”.

That’s the way I read him anyway…

JD, if I told you that a particular SS was +10 runs per 500 opportunities last year or that his true UZR is estimated at +5 per 500 opps, would you have any idea what that means?  If you knew that SS had 3.8 opps per game on the average, then you would know what to do with those numbers, but most people don’t know that and even if they did, they would rather know how much a player is “worth” defensively per an average game at his position.

I do the same thing with offense.  I usually talk about a player’s lwyt per 500 games where a “game” is defined as X number of PA (usually 4.1) for all players.  Technically, like UZR, I should use a different number of PA per game, depending on a player’s typical lineup slot or the lineup slot that a player of his type usually occupies.

"lwyt per 500 games”

MGL meants per 150 games, not 500.  (and that’s lwts).

"I always report the number”
How many players have 1000 innings at SS and CF in your assessment?

That probably made him a true +10 at SS.  Put him at 3B, and he comes in at +15 if he gets the same number of opps.  Since 3B get about 80% the opps of SS, that would make him +12.

Now, I’ll ignore your joke about 2004, s I was talking about his overall play at the position.

According to Fangraphs, he was a +12 at SS and a -2 at 3b.

I just finished up v1 of this year’s Oliver projections. When I get to them shortly one of the things I want to look at is the correlation between fielding rates at dual positions.

I calculated yearly totals, then applied a standard Marcel plus regression.

Now I’ve got more than a sufficient sample size for Adam Dunn in LF, and know that he had poor range. Then he plays a few games in RF, too small of a sample in isolation to determine much, so I’d have to give him a ‘weak’ average rating in RF...except I already know quite abit about how he performed in a similar position, LF. Or somebody else who might play a lot of SS and a little 2b.

I want to try regressing to an estimate based on how the player peformed at other positions, in addition to rgressing to the league average.

In the course of doing the research I hope to learn a few things about these relationships.

Brian, for my defensive projections, I now neutralize everyone’s UZR at all positions, using Tango’s conversions, and then whatever position they end up playing comes off of that neutralized projection.  That would not be a good way to advise a team (for example, to tell a manager, hey, even though Teixera has never played third base, his projection at third is good so you can play him there), but we can assume that when a player plays a position at which he has not played much, the manager has a good reason for that, and Tango’s conversions should apply (since they are based on similar managerial decisions). 

I really don’t see any problem, for example, taking A-Rod’s numbers at SS and then projecting him at third base, even if he had never played there before.  I agree with Tango, that he moved to third when he was already not such a good SS anymore and then while at third, he has probably lost another 3-4 runs a year as most players do as they age.  It is fully expected that he would be a below average 3B by now, even if he was a GG SS 6 and 7 years ago.

I have also heard Yankee fans say that Jeter has moved more towards the middle (and that is why his UZRs are better now) and because of that, A-Rod has moved more towards second base, which makes his UZR look bad (he fields fewer balls down the line).

mgl, that sounds like a good idea, it might take a little time for me to wrap my head all the way around it, but it would give a personalized regression point for a secondary position. I am only projecting positions actually played, but the secondaries lack good sample sizes.

"How many players have 1000 innings at SS and CF in your assessment? “

I completely trust my process for CF, LF, RF moves.  There is substantial moves between those positions. 

I mostly trust my process for 2B, 3B, SS moves.  Sometimes, they don’t make sense, but mostly they do.  There is large moves, but the selection bias occurs more here.

And that’s it.  That’s where my trust level stops.

For IF (as a group) to OF (as a group), my trust level goes way down.  That’s because it’s much different skillsets in play.  Not to mention that most moves are unidirection, so we also have the “familiarity factor” to account for.  And, no LH OF will move to the IF.

For 1B: that’s pretty much a tough one, and basically is forced in to ensure that it goes below the corner outfielders.

If your position rests on the SS to CF move, then I’m not arguing that one.  There’s no need to bring that one up.

But, if you are trying to say that we cannot, GENERALLY, say what happens when SS move to 3B, then I disagree with you totally.

There are 5 position pools: P, C, IF, OF, 1B

If we want to treat each of those distinct pools, cool.  We can agree on something.  We don’t need to figure translations between the groups by using the “dual positions” process.

But within IF and within OF, we have some reasonable translations using the dual positions process.

Tango,
I have never seen it articulated like that.  (Yes, the positions, but not the way you have here).

People use this info to suggest Jeter (or any other SS) to move to CF.  That’s not a good idea. 

I think we can “generally” say, but even with that, I’m not sure the numbers you have have enough sample behind them for the proliferation with which they are used.  Particularly when you use a positional adjustment for value (I think).

If I’m reading your responses correctly these questions should make sense.

Dave/6, What is the amount of opportunities that qualify one game?

Tango/8, Right, but if Dave is scaling UZR like he suggests then it really is runs per opportunity times 150 games worth of opportunities.

MGL/9, If I got the metric per 500 opportunities I would look at the metric as a guide for defensive ability. Currently though, I look at UZR as runs saved with defensive ability, if that makes sense. Not everyone will get the same amount of opportunities to save runs though. I am having a slight problem following you, but it seems that you are validating my first point, correct?

"People use this info to suggest Jeter (or any other SS) to move to CF.  That’s not a good idea.”

In general, I think most good shortstops have the speed and athleticism to be good at center field, but why would they?  They are more valuable at shortstop, unless their skills are such that they would be a lot better defensively in center. 

I agree Jeter to center doesn’t make sense now given his age.  Might have made sense at one time, but I don’t think the advocates were using an average position adjustment to make the case.

More like, Jeter is (or was) terrible at fielding groundballs hit to his left, but he does have good running speed, and you can see his skill tracking balls on popups, so he might be a pretty good center fielder.

You have some shortstops, like J.J. Hardy now or Cal Ripken in his prime, who excel at shortstop despite average or below running speed.  Nobody thinks they’d be any good playing center field.

And the better way to describe the positional adjustments is this way:

How much better is [player X] than a young Willie Bloomquist at this position?  Now, it doesn’t have to be EXACTLY Willie Bloomquist.  Just someone, or a composite, like him.  And preferably, a composite who is 75% RHH, and 25% LHH.  Basically, take a composite league average player, put him at each of the positions, and ask: how much better is player X compared to this single composite player.

So, composite Bloomquist will be a below-average SS, and above-average 1B.  That’s all we’re doing.  Nothing more.

Tango,
you might be doing nothing more, but it is certainly used more broadly.  You can wash your hands of it, and say you aren’t responsible for the misuse of your work, but I think it’s possible your clarity around these positional adjustments isn’t as great as you think they are.

Rally,
my research indicates that as well.  Bill Hall - well, people did think he could make the transition. 

I use actual players, and I think if one did a better job of evaluating based on body type (or other scouting), that would work.  Or the weaknesses when saying a player would do okay switching.  That is, we are never in the position of “not knowing any more”, so saying “with no other knowledge” isn’t of value.  In my research the players Rally says “fast players” are ones who worked out in CF, and other SS that moved didn’t (there are just 6 over the last 20 years, but the players skills are clearly different).

So the positional adjustment doesn’t really work for me there. 

I don’t think Luis Castillo could be an effective 3B.  No arm.  It doesn’t matter what the purported tradeoffs would be.

Castillo: that’s why the Fans Scouting Report is valuable… it gives us the breadth of components.  So, we can see what happens when SS with weak arms like Eckstein do at 3B.  And what happens when tall guys play 2B, etc, etc.  Can the erratic Lugo go to 3B, or can he only move to 2B, etc, etc, etc.

***

If someone is misusing my work, then point them out, and I’ll create a thread here, and I’ll explain why they are wrong (or right).

For example, in one of the first articles I ever did on the subject:
http://www.tangotiger.net/UZR9903TT.html

I wrote this:

Think of the columns “3” through “9” as how you would expect that player (or more accurately a player of that talent level) to play in each of the 7 positions. Since each player has his own unique character traits, you can have 2 players of equal talent level, but they would not at all be able to adapt to the same positions, even though they have the same true talent fielding level at a “neutral” position. So, be careful not to take this too far.

I’d be happy to add more clarity, but it’s not like I’m washing my dirty hands clean.  I already washed them with Ivory, and I didn’t also wash them with Purell.

I’m not going to be responsible for the words and actions of others, given that I’ve offered a good faith effort here.