MLB Top 32

[lcjjdnh]

I'll try to help you. Let's say I go out and conduct a very simple survey -- maybe 3 questions -- of 100 people. That would probably cross the threshold of significance.

If I repeat the same survey with 1,000 people, my results are going to be a lot more reliable. Both have significance, but one is way more significant.

Sort of. It depends on the size of the population that you are sampling is. 1,000 will frequently be only marginally better than 100.

You should bow out of the statistical theory discussion, as it's clear you really don't have much to offer.

First you say difference in sample sizes doesn't matter. Then you say something else. Then you say something different from the first two.

This isn't difficult.

1) There is a threshold of significance for any sample (well, several, that you can choose from). If your sample is statistically significant, there's nothing wrong with comparing it to another, much larger sample that is also significant.

2) There is a diminishing return on increased accuracy that comes with sample size. You keep implying that the relationship is linear, and it's not.

Where are you getting this diminished return stuff? It's the opposite of what the link says. Did you bother to look at it?

He means a diminished rate of return. The marginal benefit of each additional member of the sample (accuracy of ((x+1)-(x)) decreases as you add members to the sample. That is, it still increases accuracy, but not by the same increment as adding the last person did. Just because moving from 50 people to 100 people might increase accuracy by X does not mean moving from 50 to 500 people will increase accuracy by 10X.

 

[MartinonMTV2]

I'll try to help you. Let's say I go out and conduct a very simple survey -- maybe 3 questions -- of 100 people. That would probably cross the threshold of significance.

If I repeat the same survey with 1,000 people, my results are going to be a lot more reliable. Both have significance, but one is way more significant.

Sort of. It depends on the size of the population that you are sampling is. 1,000 will frequently be only marginally better than 100.

You should bow out of the statistical theory discussion, as it's clear you really don't have much to offer.

First you say difference in sample sizes doesn't matter. Then you say something else. Then you say something different from the first two.

This isn't difficult.

1) There is a threshold of significance for any sample (well, several, that you can choose from). If your sample is statistically significant, there's nothing wrong with comparing it to another, much larger sample that is also significant.

2) There is a diminishing return on increased accuracy that comes with sample size. You keep implying that the relationship is linear, and it's not.

Where are you getting this diminished return stuff? It's the opposite of what the link says. Did you bother to look at it?

He means a diminished rate of return. The marginal benefit of each additional member of the sample (accuracy of ((x+1)-(x)) decreases as you add members to the sample. That is, it still increases accuracy, but not by the same increment as adding the last person did. Just because moving from 50 people to 100 people might increase accuracy by X does not mean moving from 50 to 500 people will increase accuracy by 10X.

I never said it would. I just want comparable pools.

But as I said, there's nothing more to be gained with the statistical theory discussion. I found the link and posted the relevant part. As they say, you can lead a horse to water, but you can't make it take a leak in the water.

 

[Azrael]

At a seance in Transylvania, the ghost of Mother Theresa petitioned the Székely National Council that the ghost of Vlad the Impaler be allowed to take corporeal form again in order to drive a stake, as well as a very small sample size, through the heart of this thread.

 

[LongTimeListener]

Az, you are clearly the only one bothered by this thread, and yet you can't just keep yourself away?

 

[Mark McGwire]

Az, you are clearly the only one bothered by this thread, and yet you can't just keep yourself away?

Oh, no, he's not.

It went full stupid long, long ago, and now features Drip and Boom performance art. Any semblance of a discussion is lost to the mists of... Well, this board.

 

[outofplace]

Ichiro is the best hitter if your definition is "person who gets hits."

When most people discuss "hitters," they mean "player who creates offensive value for his baseball team," and under that definition, Ichiro doesn't belong in the discussion with Pujols.

THANK YOU RICK!!!!! That's exactly what I've been saying.

So, now you are just saying you have an odd, narrow-minded view of what a hitter is. Got it.

Because that really is what you have if you are judging a hitter based only upon total hits. By your way of thinking, Juan Pierre was a better hitter than Joey Votto last season. And not only would Suzuki have been better than Pujols, but Adrian Beltre, Billy Butler, Martin Prado and Michael Young were, too.

I've covered enough MLB games and been around the game long enough to know if a guy is a pure hitter. Ichiro is a pure hitter. You're acting like an ass-clown to compare him to Juan Pierre just for the sake of keeping a one-sided debate going. You're better than that outofplace.

Well, you are right about one thing. It is one-sided. Your side lost long ago.

And I compared him to Juan Pierre to show just how far off base you truly are. Of course, this assumes McGwire is wrong and this isn't just you being an obstinant ass for the sake of entertaining yourself.

 

[RickStain]

As sample size increases, we become more confident about our estimate, and our intervals become smaller. For instance, with a million women, we could say we are 98 percent confident that the true, average weight of women is between 115 and 117 pounds. In other words, as sample size increases, our confidence in our measurements increases and the size of our confidence intervals decreases.

If you're going to keep discounting statistical theory and trying to say things that sound good to you, then there's no purpose in continuing here. You're simply wrong about the way the process works.

You don't actually know what "diminishing return" means. That explains why you think what I said is opposing the quoted selection.

 

[MartinonMTV2]

As sample size increases, we become more confident about our estimate, and our intervals become smaller. For instance, with a million women, we could say we are 98 percent confident that the true, average weight of women is between 115 and 117 pounds. In other words, as sample size increases, our confidence in our measurements increases and the size of our confidence intervals decreases.

If you're going to keep discounting statistical theory and trying to say things that sound good to you, then there's no purpose in continuing here. You're simply wrong about the way the process works.

You don't actually know what "diminishing return" means. That explains why you think what I said is opposing the quoted selection.

I do. But you apparently don't understand this subject, so I tried to find the most basic link I could and to get the part that specifically discusses sample sizes.

Of course each new member of a large sample increases the accuracy at a smaller increment than the one before it. That's the point -- to avoid wild swings from one or two results.

Read the part about confidence intervals; that explains it. Oh wait, you refuse to do anything other than stick to your jargon. Sorry.

My suggestion: You should probably stop insisting that others don't know the subject when you clearly are limited in it. You keep trying that approach, and you keep failing.

 

[93Devil]

Rick, you actually got me thinking.

Go into your databases and see if you can pull out this stat...

Take the number of total times a player reaches base (walks, hits, fielder's choices, HBP, etc...) and what percentage of the time do they cross home plate after reaching base? I will even allow you to use home runs because they have reached base, they just don't have to stop.

Lets say Ichiro reaches base 260 times a season and scores 110 runs. How does that stack up with a great OBP guy like Youk from Boston or a Ted Simmons from way back in day?

I honestly don't know how this will turn out, but I am thinking we will see guys like Jeter, Ichiro and others who are great on the bases seem a little more valuable.

 

[RickStain]

I understand all of that perfectly well.

What you don't understand is that none of that justifies your statement that you have to have comparably sized samples to make it work.

If you had said that we had an insufficient sample to be confident, that'd be fine. We could easily look at more data, if you wanted. But what you said is that you can't compare samples that aren't of similar size, which is wrong. You can have a small but adequate sample that takes you to a conclusion with 95% confidence, and another, much larger, sample that takes you to 98% confidence, and both are large enough to be acceptable and comparable.

 

[MartinonMTV2]

I understand all of that perfectly well.

What you don't understand is that none of that justifies your statement that you have to have comparably sized samples to make it work.

If you had said that we had an insufficient sample to be confident, that'd be fine. We could easily look at more data, if you wanted. But what you said is that you can't compare samples that aren't of similar size, which is wrong. You can have a small but adequate sample that takes you to a conclusion with 95% confidence, and another, much larger, sample that takes you to 98% confidence, and both are large enough to be acceptable and comparable.

Well, you finally said something that at least fits into the theory.

But I doubt we have anything that's within the parameters you describe. That was my point from the very beginning before you went into your version of the four corners.

Do we have a sample with 95% or 98% confidence? I doubt it. Do we have two samples close enough to compare? I doubt it.

How many times were there 3rd-no outs in the eighth inning? I doubt there were very many. That is the point.

 

[93Devil]

Rick, if you have the ability to pull it up, I would appreciate it.

The phrase "creating runs" really struck me, because as a pitcher, I really could care less about the Youker at first, but an Ichiro or a Tim Raines type runner, that would put me on edge.

 

[LongTimeListener]

Can we move on to 93Devil's suggestion? I'd like to see that data as well, if possible.

I'd be interested too, but I don't see how that can be done without the extreme variable of who is batting behind each hitter -- aside from the differences in the quality of team (Ichiro vs. Jeter), you have their placement in the lineup, whereby Jeter is going to have Teixeira and A-Rod driving him in while A-Rod and Teixeira would be relying on the bottom of the order to get them home.

 

[MartinonMTV2]

Can we move on to 93Devil's suggestion? I'd like to see that data as well, if possible.

I'd be interested too, but I don't see how that can be done without the extreme variable of who is batting behind each hitter -- aside from the differences in the quality of team (Ichiro vs. Jeter), you have their placement in the lineup, whereby Jeter is going to have Teixeira and A-Rod driving him in while A-Rod and Teixeira would be relying on the bottom of the order to get them home.

In other words, situations. We have some who disregard those.

 

[MartinonMTV2]

http://www.baseball-reference.com/about/wpa.shtml

OK, I found this. My shift starts in about 5 min., so I honestly have no idea which side of the argument, if any, this lands on.

I did find one piece that should be stressed:

Note that in this case all of the credit goes to the batter and all of the blame goes to the pitcher.

 

[93Devil]

Percentage of times scoring after reaching base for a few players...


Ichiro 38.03%
Bagwell 39.47%
Jeter 41.89%
Ruth 41.99%
Foli 38.02%
Pujols 41.06%

 

[Michael_ Gee]

We see from this list that it helps a great deal to be on a team with other good hitters, which one would expect, that Pujols is a really good hitter, which we knew, and most of all, that the differences in that list are not large, like four percent. If you reach base 300 times a year, a very high figure, that would equal 12 runs difference between Ichiro and Ruth -- and Lou Gehrig likely a lot to do with that.
Wait. I forgot. Are we counting home runs as reaching base on that list? I'm starting to lose details on this thread. If so, we also learn that home runs remain the most efficient means of scoring. Another known fact.

 

[93Devil]

When I get a chance (tomorrow) I will do plate appearances and runs scored. That might be a little more telling.

Tim Foli really suprised me.