The Trap: Generalizing and Aggregate Statistics

[Catch 8]

Consider two scenarios in a close game:

1) Nobody on, nobody out.

2) Runner(s) in scoring position with two out.

I like patience and high OBP in Scenario 1, while I like aggression and Adam Jones in Scenario 2.

The OBP on a player's baseball card--or mentioned in a thread on the OH--aggregates the outcomes from the above as well as other scenarios.  These scenarios are COMPLETELY different and require COMPLETELY different approaches!

Statistics such as OBP are aggregated because it is simplified and intuitive to the human mind.  Humans do not like complexity, so we present things conveniently.  In this process, crucial information gets lost.

I haven't done the research myself, but I will push myself to one day--and I won't limit myself to OBP.

Do you have a problem with a pitcher throwing fewer strikes and perhaps walking a hitter with an open base?  It is the same concept.  

A goal of analytics is to offer a balance between complexity and meaningfulness.  

Sure, having the worst OBP in the world probably tells us something.  But we should be interested in statistics such as OBP on a much more meaningful and efficient level.

In this manner of thinking, a lead off hitter's OBP is the most significant--because he begins every game in Scenario 1. 

I'm too tired to continue, but I have a lot to say here.

 

[Luke-OH]

7 minutes ago, Catch 8 said:

Consider two scenarios in a close game:

1) Nobody on, nobody out.

2) Runner(s) in scoring position with two out.

I like patience and high OBP in Scenario 1, while I like aggression and Adam Jones in Scenario 2.

The OBP on a player's baseball card--or mentioned in a thread on the OH--aggregates the outcomes from the above as well as other scenarios.  These scenarios are COMPLETELY different and require COMPLETELY different approaches!

Statistics such as OBP are aggregated because it is simplified and intuitive to the human mind.  Humans do not like complexity, so we present things conveniently.  In this process, crucial information gets lost.

I haven't done the research myself, but I will push myself to one day--and I won't limit myself to OBP.

Do you have a problem with a pitcher throwing fewer strikes and perhaps walking a hitter with an open base?  It is the same concept.  

A goal of analytics is to offer a balance between complexity and meaningfulness.  

Sure, having the worst OBP in the world probably tells us something.  But we should be interested in statistics such as OBP on a much more meaningful and efficient level.

In this manner of thinking, a lead off hitter's OBP is the most significant--because he begins every game in Scenario 1. 

I'm too tired to continue, but I have a lot to say here.

These are good lines of thought.

Here are some additional things to consider...

1. Is situational hitting a true talent to the point where situational hitting results can predict future situational performance with more accuracy than looking at the aggregate of the player's stats?

2. Are the datasets of situational hitting/pitching large enough to provide any meaningful data?

These ideas of yours aren't completely novel, I've read discussions of similar things by sabermetricians, but generally the answer is that the specific splits are a worse predictor of future performance than the aggregate. If you can't get to some reasonable estimation of true talent, what's the point, because you won't be able to make any forward-looking predictions. 

[Jurczak]

Getting on base is always better than not getting on base. Having a high OBP certainly does not preclude you from being able to swing the bat. Most high OBP players have respectable averages as well. Very few players have high walk rates without the ability to make a pitcher pay for throwing strikes.

[Catch 8]

7 hours ago, Luke-OH said:

These are good lines of thought.

Here are some additional things to consider...

1. Is situational hitting a true talent to the point where situational hitting results can predict future situational performance with more accuracy than looking at the aggregate of the player's stats?

2. Are the datasets of situational hitting/pitching large enough to provide any meaningful data?

These ideas of yours aren't completely novel, I've read discussions of similar things by sabermetricians, but generally the answer is that the specific splits are a worse predictor of future performance than the aggregate. If you can't get to some reasonable estimation of true talent, what's the point, because you won't be able to make any forward-looking predictions. 

Luke, I appreciate you challenging me!!  I love it!

Bear with me--I'm overworked and my thoughts are all over the place.

1. The desired results (at least to me) are different in the two situations I mentioned.  I want the run in (via a hit) in Scenario 2.  A walk isn't bad, but it isn't the objective--it is punting the responsibility to the next hitter(s).  The mindset should be hit the ball hard.  I love the big inning--but I don't play for them with two out.   I could define different desired outcome variables for the two situations, meaning we need to look at the situations differently.

2. Balls in play data? Perhaps?  Perhaps not?  What if we considered taking a 2-1 FB with a man on second and two out (close game, etc) to be a "bad" outcome?  I'm thinking out loud.  Hitting performance is so variable--be it within a series, an at-bat, a career, several seasons, a postseason, etc. Taking a borderline 2-1 pitch for a strike is not a bad thing for a lead off hitter.

I am certain my ideas are not novel, nor did I mean to come off that way   

I believe hitting with a man in scoring position and two out and leading off an inning are completely different tasks, but require similar (difficult) skills.  This is not home vs. road or day vs. night.  These situations require two different approaches and different definitions of success/failure.  A first-pitch hard ground-out to third with a man in scoring position is a job well-done.  Play the percentages.  I'd prefer the lead-off guy see a few pitches--and not be as aggressive early-in-the-count--in general.  I don't think I'm giving a good example.  I have to think of a better one.

[Catch 8]

7 hours ago, Jurczak said:

Getting on base is always better than not getting on base. Having a high OBP certainly does not preclude you from being able to swing the bat. Most high OBP players have respectable averages as well. Very few players have high walk rates without the ability to make a pitcher pay for throwing strikes.

Yes it is.  But bad hitters seek walks with a man on second, two out, normal game situation.

I should have differentiated between OBP and drawing walks.

[Frobby]

I do think most hitters probably have a different mindset in a two out RISP situation than in a leadoff situation.   But there’s a danger in getting into too aggressive of a mindset with two outs and RISP.   Simply put, most hitters aren’t going to hit well when swinging at pitches that are not strikes.    They’d be better off forcing the pitcher to throw strikes and passing the baton if he doesn’t.  

FYI, here are some league-wide stats from last year:

Overall: .249/.318/.416

Leadoff: .247/.307/.415

Bases empty: .242/.308/.408

Runners on: .259/.331/.426 (.323 OBP w/o intentional walks)

RISP: .256/.337/.421 (.323 OBP without intentional walks)

RISP, 2 outs: .232/.330/.387 (.313 OBP without intentional walks)

What I take away from that is that pitchers are very careful with 2 outs and RISP.    Both the intentional walk rate and the unintentional walk rate are higher than in any other situation.    And when the hitters don’t walk, their BA is lower than in any other situation.   That’s because some of them are swinging at pitches that aren’t strikes.     

Adam Jones, by the way, has been (for him) a patient hitter in 2 out RISP situations — .281/.334/.433 compared to .278/.318/.456 overall.    That’s a bit contrary to what one might have guessed.

[now]

9 minutes ago, Frobby said:

I do think most hitters probably have a different mindset in a two out RISP situation than in a leadoff situation.   But there’s a danger in getting into too aggressive of a mindset with two outs and RISP.   Simply put, most hitters aren’t going to hit well when swinging at pitches that are not strikes.    They’d be better off forcing the pitcher to throw strikes and passing the baton if he doesn’t.  

FYI, here are some league-wide stats from last year:

Overall: .249/.318/.416

Leadoff: .247/.307/.415

Bases empty: .242/.308/.408

Runners on: .259/.331/.426 (.323 OBP w/o intentional walks)

RISP: .256/.337/.421 (.323 OBP without intentional walks)

RISP, 2 outs: .232/.330/.387 (.313 OBP without intentional walks)

What I take away from that is that pitchers are very careful with 2 outs and RISP.    Both the intentional walk rate and the unintentional walk rate are higher than in any other situation.    And when the hitters don’t walk, their BA is lower than in any other situation.   That’s because some of them are swinging at pitches that aren’t strikes.     

Adam Jones, by the way, has been (for him) a patient hitter in 2 out RISP situations — .281/.334/.433 compared to .278/.318/.456 overall.    That’s a bit contrary to what one might have guessed.

Here's another factor to consider, along the lines of your mention of pitcher focus...

The better offensive numbers with runners on are maybe not due to batting or even pitching approach, but rather a reflection of, if there are runners on, the pitcher's just not doing so well already, that inning or that day.

- 

[Frobby]

9 minutes ago, now said:

Here's another factor to consider, along the lines of your mention of pitcher focus...

The better offensive numbers with runners on are maybe not due to batting or even pitching approach, but rather a reflection of, if there are runners on, the pitcher's just not doing so well already, that inning or that day.

- 

It’s harder to pitch with runners on.   You have to pitch from the stretch, worry about holding the runners, etc.

[now]

44 minutes ago, Frobby said:

It’s harder to pitch with runners on.   You have to pitch from the stretch, worry about holding the runners, etc.

Right, that's a more direct effect of pitching with runners on. The factor I was considering was more indirect, or retrospective - supposing that both the runners already on and the next hitters could be benefitting from the same cause, like poorer stuff that day.

[Frobby]

3 minutes ago, now said:

Right, that's a more direct effect of pitching with runners on. The factor I was considering was more indirect, or retrospective - supposing that both the runners already on and the next hitters could be benefitting from the same cause, like poorer stuff that day.

That makes sense as well.