Jump to content
North Side Baseball
Posted
It's not a random sample, for one.

Awesome

I don't care what they said. They're simply wrong.

 

now THIS is awesome.

 

Consider this. Let's say we take players the first three months of the season, minimum 200 PAs. We then split up their PA's into two equal sets, randomly. You can do this with a rand function very easily. If you don't want to do that, just alternate them and throw them into two sets that way. I chose the first method. Then in those two sets, calculate OPS. Keep in mind that these two sets have 100-175 PAs max. Then for the players calculate their true talent level the best you can. A simple way is to weight the three previous years, hence why we wanted to look at the first three months of the season. You can also use a projection system to come up with one. Then take the difference between their "true" OPS and their actual OPS's for each set.

 

If it were truly random like you all (and they) say, then we would expect this correlation to be zero, or very close to it. Since doing well (or poor) in one set of 125 PAs shouldn't have anything to do with the other set of 125 PAs. However, when you look at the data you'll find a correlation that's in the .30-.40 range between the differences. Meaning that the two are certainly related and that high fluctuations in short term true talent level exist. It's even more surprising considering that each set is around 150 PAs.

 

Thus, their hypothesis can be thrown out the window and they are simply wrong.

How many tests did you perform? If it is a small number of tests, then a correlation of .3 - .4 is really not sufficient evidence to back your hypothesis.

 

over 1000 and I did it twice. the second time using the rand function again to split up the PAs differently. Got similar results.

  • Replies 244
  • Created
  • Last Reply

Top Posters In This Topic

Top Posters In This Topic

Posted (edited)
I thought Id post this due to the lack of recent Webb news.

 

Brandon Webb "remains excited" about the Nationals, and the two sides talked during the winter meetings.

 

Remember when every player wanted to play for the Cubs, which coincedently ended when Hendry lost his bottomless money pit?

 

Do we really want a player that "remains excited" about the Nationals?

 

I bet it's a negotiating ploy. :-k

Edited by haubrich91
Posted
It's not a random sample, for one.

Awesome

I don't care what they said. They're simply wrong.

 

now THIS is awesome.

 

Consider this. Let's say we take players the first three months of the season, minimum 200 PAs. We then split up their PA's into two equal sets, randomly. You can do this with a rand function very easily. If you don't want to do that, just alternate them and throw them into two sets that way. I chose the first method. Then in those two sets, calculate OPS. Keep in mind that these two sets have 100-175 PAs max. Then for the players calculate their true talent level the best you can. A simple way is to weight the three previous years, hence why we wanted to look at the first three months of the season. You can also use a projection system to come up with one. Then take the difference between their "true" OPS and their actual OPS's for each set.

 

If it were truly random like you all (and they) say, then we would expect this correlation to be zero, or very close to it. Since doing well (or poor) in one set of 125 PAs shouldn't have anything to do with the other set of 125 PAs. However, when you look at the data you'll find a correlation that's in the .30-.40 range between the differences. Meaning that the two are certainly related and that high fluctuations in short term true talent level exist. It's even more surprising considering that each set is around 150 PAs.

 

Thus, their hypothesis can be thrown out the window and they are simply wrong.

How many tests did you perform? If it is a small number of tests, then a correlation of .3 - .4 is really not sufficient evidence to back your hypothesis.

 

over 1000 and I did it twice. the second time using the rand function again to split up the PAs differently. Got similar results.

1) you clearly are willing to put a lot more into this argument than I will be

2) .3 is not a real high correlation, though it is likely to be statistically significant after 1000 trials if you were properly randomizing the samples between each trial

3) your methodology has a couple of flaws in how it relates to the current argument, though. The first of which is that you seem to be slicing at bats within the same portion of a season - potentially even within the same games. If so, you'll likely see a positive correlation between the samples due to having at bats from each sample occurring under the same conditions (same day, vs same pitcher, same weather conditions, etc.). However, the majority of Byrd's at bats in the #3 position happened after Lee had been traded and were largely consecutive. So there are a very large number of other potential causal factors other than batting third that you have been ignoring: he could have been battling a small injury, he could have simply been slumping, he could have had a run of difficult pitching matchups, etc. All of those are more likely causal factors than a change of approach when moving from #5 or #4 to #3 in the lineup.

 

You're also overlooking something in your argument: the Kosuke factor. Kosuke has 230 April at bats now in his career - he also has a .326/.438/.535/.973 in those at bats. He now has 223 June at bats - with a .211/.314/.300/.614 in those AB's. I think individual players can show VERY sigificant swings in performance over the course of 200-300 at bat samples, don't you?

Posted
i think you're trying to agree with me... im the one saying that streaks exist and that performing above or below true talent level in the short term isn't completely random. that's all i am saying. my argument has little to do with him being "better" batting third. my argument is that short term fluctuations exist, that they're not entirely random, and that's partly why his numbers differ in different spots, because those ABs in each spot tend to come in groups. thats been my argument the entire time and why i said it wasnt a "random sample" of 323 ABs.
Posted
i think you're trying to agree with me... im the one saying that streaks exist and that performing above or below true talent level in the short term isn't completely random. that's all i am saying. my argument has little to do with him being "better" batting third. my argument is that short term fluctuations exist, that they're not entirely random, and that's partly why his numbers differ in different spots, because those ABs in each spot tend to come in groups. thats been my argument the entire time and why i said it wasnt a "random sample" of 323 ABs.

 

I don't think the original argument was that it was a random sample, but that it wasn't meaningful going forward. It was said Byrd shouldn't bat third because he's not a #3 hitter and his numbers in that spot were cited as an example. Others came back and argued that those numbers didn't mean much because they weren't predictive and didn't necessarily show what he would always do in the third spot.

Posted
323 PAs is not a small sample.

 

Yes, it is. In that number of PA's roughly 25% of players can be +/- 100 points of their actual production. The other 75% of players will be +/- 50 points. That's purely through luck.

 

You need At Least 1000 PAs to start getting a good handle on their real ability.

 

As was alluded to earlier, what if he goes off for a .326/.395/.515/.910 line over his next 323 PA's? Does that mean he's a #3 hitter? Does that mean he was unlucky those first 323 PAs, or was he really lucky the second 323 PA's. You're basing your proclamation on the equivalent of a half season of production. That's nowhere near enough of a sample.

 

You just made that number up. My point isn't that it's predicative. My point is that there is a reason. Randomness isn't it. His true talent level during those 323 PAs was not his career average.

Posted
It's not a random sample, for one.

Awesome

I don't care what they said. They're simply wrong.

 

now THIS is awesome.

 

Consider this. Let's say we take players the first three months of the season, minimum 200 PAs. We then split up their PA's into two equal sets, randomly. You can do this with a rand function very easily. If you don't want to do that, just alternate them and throw them into two sets that way. I chose the first method. Then in those two sets, calculate OPS. Keep in mind that these two sets have 100-175 PAs max. Then for the players calculate their true talent level the best you can. A simple way is to weight the three previous years, hence why we wanted to look at the first three months of the season. You can also use a projection system to come up with one. Then take the difference between their "true" OPS and their actual OPS's for each set.

 

If it were truly random like you all (and they) say, then we would expect this correlation to be zero, or very close to it. Since doing well (or poor) in one set of 125 PAs shouldn't have anything to do with the other set of 125 PAs. However, when you look at the data you'll find a correlation that's in the .30-.40 range between the differences. Meaning that the two are certainly related and that high fluctuations in short term true talent level exist. It's even more surprising considering that each set is around 150 PAs.

 

Thus, their hypothesis can be thrown out the window and they are simply wrong.

 

Wouldn't the correlation depend upon how comparable the ~300 PA "sample" OPS is to the "true" talent value? If the difference is great between the "true" value and the overall "sample" value, I would be shocked if there wasn't some correlation between the randomized subsets. Unless the sample is VERY small or the sampling VERY weird, two random subsets of ~300 PA of say, .650 OPS can be expected to be close to .650 as well. If the true OPS is .800, then the difference will be close to .150 for each data set. You should reasonably expect a positive correlation, especially over many iterations, which would statistically suppress outliers. Maybe I'm missing something very important in the analysis, but it seems that the only place a near zero correlation might crop up would be a stretch of PAs where the player produced very near his "true" talent, in which case, what are we all arguing about, again?

Posted
323 PAs is not a small sample.

 

Yes, it is. In that number of PA's roughly 25% of players can be +/- 100 points of their actual production. The other 75% of players will be +/- 50 points. That's purely through luck.

 

You need At Least 1000 PAs to start getting a good handle on their real ability.

 

As was alluded to earlier, what if he goes off for a .326/.395/.515/.910 line over his next 323 PA's? Does that mean he's a #3 hitter? Does that mean he was unlucky those first 323 PAs, or was he really lucky the second 323 PA's. You're basing your proclamation on the equivalent of a half season of production. That's nowhere near enough of a sample.

 

You just made that number up. My point isn't that it's predicative. My point is that there is a reason. Randomness isn't it. His true talent level during those 323 PAs was not his career average.

 

I took that argument to be that the 323 PAs are not enough to determine what he'll be going forward in the third spot.

Posted
323 PAs is not a small sample.

 

Yes, it is. In that number of PA's roughly 25% of players can be +/- 100 points of their actual production. The other 75% of players will be +/- 50 points. That's purely through luck.

 

You need At Least 1000 PAs to start getting a good handle on their real ability.

 

As was alluded to earlier, what if he goes off for a .326/.395/.515/.910 line over his next 323 PA's? Does that mean he's a #3 hitter? Does that mean he was unlucky those first 323 PAs, or was he really lucky the second 323 PA's. You're basing your proclamation on the equivalent of a half season of production. That's nowhere near enough of a sample.

 

You just made that number up. My point isn't that it's predicative. My point is that there is a reason. Randomness isn't it. His true talent level during those 323 PAs was not his career average.

 

I took that argument to be that the 323 PAs are not enough to determine what he'll be going forward in the third spot.

 

Exactly. I'm not arguing that he was good before or that was his actual talent level, just that you can't tell what he's going to do based on such a small sample. If he had those numbers in 1500 (example) relatively consecutive PA's, then yes, you could probably say he's not very good at it.

 

 

And.

 

Has there been any more news on Webb, or anyone for that matter.

Posted
Would they just hurry up and sign him so I can be a little optimistic about this season. Sorry to bring up Webb in the Webb thread, but I get a little nervous that the Yankees lost out on Lee and they might jump on the wagon. I know they said they are not going to focus on starting pitching but I wouldn't count them out.
Posted (edited)
Wouldn't the correlation depend upon how comparable the ~300 PA "sample" OPS is to the "true" talent value? If the difference is great between the "true" value and the overall "sample" value, I would be shocked if there wasn't some correlation between the randomized subsets. Unless the sample is VERY small or the sampling VERY weird, two random subsets of ~300 PA of say, .650 OPS can be expected to be close to .650 as well. If the true OPS is .800, then the difference will be close to .150 for each data set. You should reasonably expect a positive correlation, especially over many iterations, which would statistically suppress outliers. Maybe I'm missing something very important in the analysis, but it seems that the only place a near zero correlation might crop up would be a stretch of PAs where the player produced very near his "true" talent, in which case, what are we all arguing about, again?

 

Take half of the PAs in the sample. Find the OPS of that sample minus his true talent level. Take the other half of the sample. Find the OPS of that sample minus his true talent level. Then find the correlation between (OPS1 - true OPS) and (OPS2 - true OPS). In theory, if no streaks exist then for the 300 PAs we're going to get a normal(ish) distribution around the players true OPS. Thus when we take a sample of 150 PAs. We should get a normal(ish) distribution around the players true OPS. So if we take our OPS over the 150 PAs minus the true talent level OPS, we're going to get two numbers that are completely unrelated.

Edited by Northside Blues
Posted
Can a mod please split off this Marlon Byrd/whatever else talk into its own thread so we can get back to talking about Brandon Webb?

 

Please?

 

Yeah, Ive been skipping over everything non-Webb related which has been most of the last couple of pages.

Posted
Would they just hurry up and sign him so I can be a little optimistic about this season. Sorry to bring up Webb in the Webb thread, but I get a little nervous that the Yankees lost out on Lee and they might jump on the wagon. I know they said they are not going to focus on starting pitching but I wouldn't count them out.

 

I thought the same thing, and just saw this on MLBTR

 

So which starters might the Yankees pursue? In a separate piece, Sherman says the Yankees don't like Brandon Webb all that much. However, they will consider pursuing a veteran with "good stuff but questionable health."

 

Looks like Webb isn't on their radar

Posted
Would they just hurry up and sign him so I can be a little optimistic about this season. Sorry to bring up Webb in the Webb thread, but I get a little nervous that the Yankees lost out on Lee and they might jump on the wagon. I know they said they are not going to focus on starting pitching but I wouldn't count them out.

 

I thought the same thing, and just saw this on MLBTR

 

So which starters might the Yankees pursue? In a separate piece, Sherman says the Yankees don't like Brandon Webb all that much. However, they will consider pursuing a veteran with "good stuff but questionable health."

 

Looks like Webb isn't on their radar

 

Probably because their 100m infield won't be able to absorb that many groundballs over the course of a season.

Posted
Would they just hurry up and sign him so I can be a little optimistic about this season. Sorry to bring up Webb in the Webb thread, but I get a little nervous that the Yankees lost out on Lee and they might jump on the wagon. I know they said they are not going to focus on starting pitching but I wouldn't count them out.

 

I thought the same thing, and just saw this on MLBTR

 

So which starters might the Yankees pursue? In a separate piece, Sherman says the Yankees don't like Brandon Webb all that much. However, they will consider pursuing a veteran with "good stuff but questionable health."

 

Looks like Webb isn't on their radar

 

That eases my worries a little bit

Posted
. If Webb signs with the Cubs, he's likely to be the 4th or 5th starter in the rotation.

http://mlb.fanhouse.com/2010/12/17/source-brandon-webb-nearing-decision/

 

I'm pretty sure I see Webb as either ace or 2 starter quality or broken down and a waste of money. I find it hard to see a scenario where he settles into a 4th starter spot and pitches like one. 2007 Webb wouldn't be a 4th starter on the Phillies.

Posted
. If Webb signs with the Cubs, he's likely to be the 4th or 5th starter in the rotation.

http://mlb.fanhouse.com/2010/12/17/source-brandon-webb-nearing-decision/

 

I'm pretty sure I see Webb as either ace or 2 starter quality or broken down and a waste of money. I find it hard to see a scenario where he settles into a 4th starter spot and pitches like one. 2007 Webb wouldn't be a 4th starter on the Phillies.

 

I have a feeling this is not going to happen, I believe Cash goes to the rotation or the Cubs make a trade. I see the Cubs getting nixed by someone willing to offer more guarranteed.

Posted
. If Webb signs with the Cubs, he's likely to be the 4th or 5th starter in the rotation.

http://mlb.fanhouse.com/2010/12/17/source-brandon-webb-nearing-decision/

 

I'm pretty sure I see Webb as either ace or 2 starter quality or broken down and a waste of money. I find it hard to see a scenario where he settles into a 4th starter spot and pitches like one. 2007 Webb wouldn't be a 4th starter on the Phillies.

 

I have a feeling this is not going to happen, I believe Cash goes to the rotation or the Cubs make a trade. I see the Cubs getting nixed by someone willing to offer more guarranteed.

This is my feeling as well.

Posted
. If Webb signs with the Cubs, he's likely to be the 4th or 5th starter in the rotation.

http://mlb.fanhouse.com/2010/12/17/source-brandon-webb-nearing-decision/

 

I'm pretty sure I see Webb as either ace or 2 starter quality or broken down and a waste of money. I find it hard to see a scenario where he settles into a 4th starter spot and pitches like one. 2007 Webb wouldn't be a 4th starter on the Phillies.

 

I have a feeling this is not going to happen, I believe Cash goes to the rotation or the Cubs make a trade. I see the Cubs getting nixed by someone willing to offer more guarranteed.

This is my feeling as well.

 

Loved Webb pre-injury, but missing two seasons is a gamble not really all that worth taking given the current conditions of the team. If he signs a 1 year deal with someone and he does well, offer him something good next year when there is money to spend. If Tampa wants too much, start the year off with what you have and consider making adjustments at the deadline if they actually find themselves within shouting distance of the playoffs.

 

It almost seems like Fukudome has to go in order to make improvements, which may not be making improvements once you remove Fukudome anyway.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
The North Side Baseball Caretaker Fund
The North Side Baseball Caretaker Fund

You all care about this site. The next step is caring for it. We’re asking you to caretake this site so it can remain the premier Cubs community on the internet. Included with caretaking is ad-free browsing of North Side Baseball.

×
×
  • Create New...