something that i think people need to know but they don't.

[Mephistopheles]

this hasn't come up here as recently as much as it has in the past but its still something people aren't getting. if you get it, good for you and welcome to the club.

 

park adjustments

 

when a statistic is park adjusted it does NOT mean what the player would do in an average park. For instance if players X's ERA in a park is 4.50 and his ERA for that park adjusted is 4.80, it does NOT mean he's a 4.80 pitcher in a neutral field. really we dont want to scale for this.

 

Same goes for hitters. If he created 50 runs in park x, and the park adjusted total is 45, that doesnt mean hed create 45 runs in a neutral park. many people who arent familiar with advanced stats and most weak stat-heads dont quite grasp this idea. They think a park adjusted stat means what theyd do in an average park. its not.

 

in fact it's good that it isnt. When you park adjust things you can do it with two different goals. The future, or value-production. If you're doing the former you should adjust it into an average park, but EqA, VORP and the such are NOT those kinds of statistics. They're the ones with the value-production intent in mind. So when they adjust for something they compare the players production to what an average player would do in that park.

 

For instance, if a player creates 100 runs in his park. and an average player creates 100 runs overall, one would assume that means hes an average player. Say then that the park's run factor is +5%, then that means an average player in that park would create 105 runs. Thats how you do the scaling.

 

Its important to do it this way because some players' skill sets are better suited for certain parks. So the value they create depends on this. A great example of this is Biggio in the Astrodome. He was a guy whose game was NOT impacted by the Domes high HR factor. He was great because of his OBP skills at 2B.

 

Im sure some of you knew all of this, but I know a lot you did not.

 

And I am making this thread a nice meph thread. That means I am not going to be mean to anyone in this thread nor arrogant. So if you have any questions about baseball, baseball statistics, the minors, the draft or the such, might as well ask them in here -- or if you want to just discuss anything. My knowledge is your knowledge in this thread. I wont bite*

 

*if you snap on me, you suck

[snoodmonger]

I suck at math.

 

I like advanced stats, but I don't understand them, and I couldn't calculate them if the Baby Jesus's life was at stake. So permit me to ask a really dumb question for clarification.

 

 

In this scenario:

 

For instance, if player A creates 100 runs in his park. and an average player creates 100 runs overall, one would assume that means hes an average player. Say then that the park's run factor is +5%, then that means an average player in that park would create 105 runs. Thats how you do the scaling.

 

that means player A is actually slightly below average. Or at least, he creates runs at a slightly lower rate in that park than the average player would. Is that right?

 

 

What I would like to see you (or someone) do is create a thread that gives a layman's explanation for many of the advanced stats and how they are calculated. perhaps not the actual equation, as that would be likely to make my head explode, but just a dummies explanation of how it's created and how the stat can be used. (The future vs value-production thing is interesting to me, for example.)

 

I will readily admit that four years ago, I thought OPS was a Naughty By Nature song. And now that I understand that and some of the more common saber stats, I see people using different, seemingly more complex ones. The WARPs, EqA, etc. So I think I need a data update.

 

I'm sure I'm not the only one, and it could be stickied for easy reference.

[Mephistopheles]

that means player A is actually slightly below average. Or at least, he creates runs at a slightly lower rate in that park than the average player would. Is that right?

 

You have to be careful with your choice of wording. It does not mean that player A is slightly below average. It means when he's playing in that park he's slightly below average. For all we know he could hit 420 ft flyouts to CF routinely but plays in Enron so theyre outs. Everywhere else he'd be a masher. But yes, in that park he creates runs at a slightly lower rate than the average player. So his value/production is slightly below average for that team. Again it's paramount to realize it does not mean he's a below average player.

 

I could make a thread. Sometime after next week when I have time.

[haltz]

What I would like to see you (or someone) do is create a thread that gives a layman's explanation for many of the advanced stats and how they are calculated. perhaps not the actual equation, as that would be likely to make my head explode, but just a dummies explanation of how it's created and how the stat can be used.

Sons of Sam Horn has a Statistical Reference Page on their Wiki. Probably not exactly what you are looking for, and I haven't really looked at it in a while, but it can be helpful.

[seanimal]

I have a question,

 

What are the worthwhile defensive metrics worth paying attention to? I've noticed people mention that BP's calculations of defensive contribution are poor, but I've never really seen any explanation as to why. In particular, what are you looking for when analyzing one's defensive contributions?

[davearm]

If I'm understanding you correctly, the park adjustments you're describing do not lend themselves to the sort of apples-to-apples player comparison that you'd want and expect them to.

 

What you're telling us is that park-adjusting Travis Hafner's numbers tells us how much better he was at Jacob's Field than the average hitter would've been at Jacob's Field.

 

Well, how good would Derrek Lee have been at Jacob's Field? Or Ryan Howard, or Prince Fielder, or David Eckstein? Park adjusting each of those other guys' numbers in the manner you describe gives us no clue. All we know is how good each guy was, relative to the average hitter, at his own home park. We don't get any sense of how good a certain guy would've been at some specific other park.

 

In order to compare Hafner to Lee to Howard to Fielder to Eckstein while neutralizing park effects, we'd want to know how each guy's production would translate in a neutral/average hitting environment (the adjustment you state is *not* being made), no?

 

I guess the bottom line is this: who cares how Biggio produced at MMP, compared to the average hitter, if we have no sense of how the majority of players in MLB would've produced at MMP? What I really need to know is how good Biggio would've been if you took away the advantages and disadvantages particular to MMP. If I do the same for everyone else in baseball, only then will I have a proper basis for comparing Biggio to everyone/anyone else.

[davearm]

I have a question,

 

What are the worthwhile defensive metrics worth paying attention to? I've noticed people mention that BP's calculations of defensive contribution are poor, but I've never really seen any explanation as to why. In particular, what are you looking for when analyzing one's defensive contributions?

The current state of the art in defensive metrics revolves around a few competing methods for analyzing play-by-play data. This is guys actually sitting in front of monitors watching replays of every batted ball from every game for the entire MLB season, and, using their judgement coupled with standardized guidelines, scoring fielders based on whether a given type of batted ball could or could not have been successfully fielded by an average player at a given position, and whether the ball in question actually was or was not successfully fielded by the player being evaluated.

 

The end result of such an exhaustive endeavor can be quite detailed. A particular SS may be shown to be very good going into the hole, only average going up the middle, poor on slowly-hit barehanded chances, terrific on popups toward the OF, etc. etc.

 

Popular thinking these days is that basically any defensive metric that isn't rooted in PBP data is obsolete and marginally valuable.

 

And now I'll stand aside as these thoughts are mercilessly shredded by the resident expert(s).

[haltz]

What are the worthwhile defensive metrics worth paying attention to? I've noticed people mention that BP's calculations of defensive contribution are poor, but I've never really seen any explanation as to why. In particular, what are you looking for when analyzing one's defensive contributions?

 

THT: Evaluating the Evaluators

 

I guess the bottom line is this: who cares how Biggio produced at MMP, compared to the average hitter, if we have no sense of how the majority of players in MLB would've produced at MMP?

 

Well, we do. MMP affects different hitters differently. Park effects have to be much more granular than just looking at a run-scoring environment if you want to look at ability, i.e., what a player would do in a neutral context.

 

I think this was basically said, but if you want to look at how valuable Biggio has been, you have to see how much his offensive contribution is worth at Minute Maid Funhouse. If the ultimate and obvious goal in MLB is wins, then you need to look at how how much a run means to a win at a certain ballpark.

 

It depends whether you want to target a player's ability or how valuable he's been. This is the difference between a projection, and something like VORP. Do you want to look backwards, or forwards?

 

At least, that's how I understand it. And being a weak stat-head at best, I'm sure Meph will correct me if need be.

[Derwood]

are players' road statistics adjusted the specific parks they are producing those stats in?

[hardcorecubfan]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

But thank you for clarifying

[abuck1220]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

 

and stats clearly don't tell you any of that.

[jyoung55]

What a ridiculous title for a thread.

 

I was hoping it had something to do with a crystal ball and the Cubs winning the World Series in my lifetime.

[davearm]

What are the worthwhile defensive metrics worth paying attention to? I've noticed people mention that BP's calculations of defensive contribution are poor, but I've never really seen any explanation as to why. In particular, what are you looking for when analyzing one's defensive contributions?

 

THT: Evaluating the Evaluators

 

I guess the bottom line is this: who cares how Biggio produced at MMP, compared to the average hitter, if we have no sense of how the majority of players in MLB would've produced at MMP?

 

Well, we do. MMP affects different hitters differently. Park effects have to be much more granular than just looking at a run-scoring environment if you want to look at ability, i.e., what a player would do in a neutral context.

 

I think this was basically said, but if you want to look at how valuable Biggio has been, you have to see how much his offensive contribution is worth at Minute Maid Funhouse. If the ultimate and obvious goal in MLB is wins, then you need to look at how how much a run means to a win at a certain ballpark.

 

It depends whether you want to target a player's ability or how valuable he's been. This is the difference between a projection, and something like VORP. Do you want to look backwards, or forwards?

 

At least, that's how I understand it. And being a weak stat-head at best, I'm sure Meph will correct me if need be.

What I want to do is have a basis for an apples-to-apples, park-independent comparison of Player A to Player B.

 

So if I say, for example, "Todd Helton is a better hitter than Carlos Delgado, " and you respond by saying, "well yeah, look at the park Helton hits in," where do we go from there?

[hardcorecubfan]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

 

and stats clearly don't tell you any of that.

 

Yes, they do but the math in order to get to the final number is dizzying.

[Derwood]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

 

and stats clearly don't tell you any of that.

 

Yes, they do but the math in order to get to the final number is dizzying.

 

lucky for all of us, there are website that do all the math for us

[Bull]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

 

and stats clearly don't tell you any of that.

 

Yes, they do but the math in order to get to the final number is dizzying.

 

lucky for all of us, there are website that do all the math for us

 

there is no "final number". All stats have value in their own way, but there is no one magic stat.

 

That's why we are able to use ridiculous amounts of bandwidth arguing.

[hardcorecubfan]

Stats do fascinate me for a long road trip from Amarillo to Houston but does the guy hit the ball, does he catch the ball and can he throw the ball on game day is the most important stat.

 

 

and stats clearly don't tell you any of that.

 

Yes, they do but the math in order to get to the final number is dizzying.

 

lucky for all of us, there are website that do all the math for us

 

there is no "final number". All stats have value in their own way, but there is no one magic stat.

 

That's why we are able to use ridiculous amounts of bandwidth arguing.

 

statistically is that more or less bandwith than the Steve Fossett search on google earth right now??? :roll:

[Mephistopheles]

I have a question,

 

What are the worthwhile defensive metrics worth paying attention to? I've noticed people mention that BP's calculations of defensive contribution are poor, but I've never really seen any explanation as to why. In particular, what are you looking for when analyzing one's defensive contributions?

 

Personally, I like the framework behind FRAR, but I dislike Rate2 which is how they calculate how much over/below average a player is defensively. FRAR was developed so it can analyze a player defensively based on box scores that have been around since forever. Other new wave statistics rely on batted ball type and directional vector off the bat. They're also probability based in that respect. They use a lot more relevant data and that's why they're more accurate. The framework of FRAR is based upon the principle that each position has a defensive run value. This difference actually is equivalent if you calculate between offense between two positions or if you calculate it between defensive prowess of a player moving from one to the other. For example, FRAR base value for a SS is 33 runs. For a 3B it's 22. If a +0 SS moves to 3B, statistically they usually become a +11 defensive 3B. So it kind of works. Also, usually the difference between an average SS and 3B over the course of a season is pretty damn close to eight runs. In fact using EqA this season for those positions and the difference is 11 runs.

 

If I'm understanding you correctly, the park adjustments you're describing do not lend themselves to the sort of apples-to-apples player comparison that you'd want and expect them to.

 

What you're telling us is that park-adjusting Travis Hafner's numbers tells us how much better he was at Jacob's Field than the average hitter would've been at Jacob's Field.

 

To do this, you have to adjust everything to a neutral environment and then convert it to the park environment you want it in. It's not apples to apples in the future, but it's apples to apples to value created in the past. BP tries to do this, but honestly they fall on their face. They actually assume that a players all-time EqA would be their EqA overall. They then build a line that would equal that production. This is their attempt to remedy this situation, and it's what they use in PECOTA. Honestly, I don't agree with it one bit. Here's a player card for Hafner.

http://i47.photobucket.com/albums/f189/kctigers23/Hafner1.gif

There are three EqAs on there. The first one under advanced batting statistics is an average player in his league (AL or NL) for that season. The second one is an average player all time. Their attempt to build a line in a neutral context is listed under Translated Batting Statistics. They start building it assuming the eqa all time will still exist. Like I said this is stupid.

 

I guess the bottom line is this: who cares how Biggio produced at MMP, compared to the average hitter, if we have no sense of how the majority of players in MLB would've produced at MMP? What I really need to know is how good Biggio would've been if you took away the advantages and disadvantages particular to MMP. If I do the same for everyone else in baseball, only then will I have a proper basis for comparing Biggio to everyone/anyone else.

 

In the case of Biggio we're looking at the past, value created precisely over an average and replacement player. Historically speaking the value is more important. Biggio would have been hella good anywhere. His strengths arent affected by any park, but he would not have been as valuable in a place where HRs were significantly increased. He would have still been valuable. Hes a 20 HR guy too. When comparing production of two players this is the correct way to do it. In general this way is fine for the future. Most parks are interchangeable. It's only the extremes that matter, the coors, domes, petco etc.

 

are players' road statistics adjusted the specific parks they are producing those stats in?

 

Im not sure if they do. If they do, it doesn't matter. If they don't, they don't matter. All of your road stats are going to be produced, on average, at a league average park. It's going to cancel out. Given the limited sample sizes its probably better not to find runs produced over average for a three game set and compare it to an average player.

 

This brings up something. What does a teams winning percentage represent? It represents how often a team would be expected to win against an average team. What does a person's batting average represent? It represents how often a player gets a hit off of a league average pitcher. In general all statistics are already adjusted this way when there's a large sample size. We don't need to do anything.

 

there is no "final number". All stats have value in their own way, but there is no one magic stat.

 

Sure there is. Only its a probability range. Not a given number.

[Omar]

How much wood would Kerry Wood chuck if Wood could chuck wood? Edited September 15, 2007 by Omar

[Mephistopheles]

1 piece of wood. Hed be hurt afterwards