5/30 Cobs (Wood) @ Brewers (Estrada) 7:10 CT WGN

[Hairyducked Idiot]

That doesn't mean I was wrong.

It means you thought a stat was more meaningful tha. It actually was. You were and are wrong.

 

It's -17 now, but you have to take out some of the negative blowouts and you'll see that it basically hasn't changed.

[MWV]

Gooney and Mojo you are making me sympathize with Kyle and I hate you both for it.

[Sammy Sofa]

At the types of things that happen to real baseball teams, this principle still works.

 

Honestly, I can't wrap my head around that concept; it seems so simplistic to be based on runs scored and runs given up totals. I don't get how it's...nuanced (if that's the right word) like that.

[Sammy Sofa]

Gooney and Mojo you are making me sympathize with Kyle and I hate you both for it.

 

Hey, I have no problem copping to what I don't understand about the game. Gooney is just gooney.

[MWV]

Fair distinction. Hatred withdrawn.

[Guest]

You guys would really like hockey.

[Guest]

I've stated numerous times I have zero clue how most baseball statistics "work" and generally just rely on being able to figure out what is good and bad in the individual categories (and a lot I don't even bother with because I can't grasp what the hell they're even supposed to mean no matter how many times I read it). With run differential I always just assumed that there was some kind of flaw in it because of outlier games and that the pythag projectons somehow adjusted for it, but I guess not.

The adjustment for outliers is built into the gigantic sample size. That is, it needs no adjustment.

[Guest]

I've stated numerous times I have zero clue how most baseball statistics "work" and generally just rely on being able to figure out what is good and bad in the individual categories (and a lot I don't even bother with because I can't grasp what the hell they're even supposed to mean no matter how many times I read it). With run differential I always just assumed that there was some kind of flaw in it because of outlier games and that the pythag projectons somehow adjusted for it, but I guess not.

The adjustment for outliers is built into the gigantic sample size. That is, it needs no adjustment.

 

However, because it's built on sample size, it's hard to evaluate until the end of the season.

[Guest]

ok, think about it this way...

 

everyone agrees that the larger the sample, the more likely the statistic is going to reflect the true process, right?

 

after 50 games, you have 50 data points as far as w/l goes.

 

After 50 games, you have around 400 data points as far as runs scored / runs allowed goes.

[Sammy Sofa]

ok, think about it this way...

 

everyone agrees that the larger the sample, the more likely the statistic is going to reflect the true process, right?

 

after 50 games, you have 50 data points as far as w/l goes.

 

After 50 games, you have around 400 data points as far as runs scored / runs allowed goes.

 

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

[Sammy Sofa]

You guys would really like hockey.

 

I finally found out what "17 Seconds" means tonight!

[Hairyducked Idiot]

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

 

 

After 40 games, run differential is more accurate at predicting a team's win percentage for the rest of the season than their actual record or previous year's record.

[Sammy Sofa]

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

 

 

After 40 games, run differential is more accurate at predicting a team's win percentage for the rest of the season than their actual record or previous year's record.

 

Once run through the pythag formula, right?

[Guest]

Then explain xFIP and BAbip.

 

BABIP: http://www.fangraphs.com/library/offense/babip/

 

xFIP: http://www.fangraphs.com/library/pitching/xfip/

[Guest]

ok, think about it this way...

 

everyone agrees that the larger the sample, the more likely the statistic is going to reflect the true process, right?

 

after 50 games, you have 50 data points as far as w/l goes.

 

After 50 games, you have around 400 data points as far as runs scored / runs allowed goes.

 

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

 

it's already more concrete than their W/L record

 

sequencing of the results of PA is completely random, so the larger sample of results of PAs both for and against your team say more about its ability to win or lose games than the actual w/l results

 

also i can't believe goony is this wrong

[Guest]

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

 

 

After 40 games, run differential is more accurate at predicting a team's win percentage for the rest of the season than their actual record or previous year's record.

 

honestly, i don't see why it wouldn't pretty much immediately be more predictive than their actual record.

[Guest]

So is there a point during the season when it becomes more "concrete?" Or is it, like sulley said, that really doesn't hold much weight until the end of the season? I've been taking it as more of a predictive stat.

 

 

After 40 games, run differential is more accurate at predicting a team's win percentage for the rest of the season than their actual record or previous year's record.

 

Once run through the pythag formula, right?

 

the pythag formula is basically just a translation of run differential into a w-l form to tell you what record that differential should produce over the given amount of games.

[Sammy Sofa]

Then explain xFIP and BAbip.

 

BABIP: http://www.fangraphs.com/library/offense/babip/

 

That's what I get for always going to BR and never using FG; BR's little hover definition thing doesn't have this key bit:

 

The average BABIP for hitters is around .290 to .310.

 

So I was thinking it was supposed to be relative to the player's BA.

 

Similar thing with FG's handy little xFIP chart.

[SouthSideRyan]

It definitely is.

 

Actually, I think Brett showed that over the last 20 years we've been pretty amazingly bad at keeping up with Pythagorean.

 

Yeah and over 25 years it wasn't that bad it was typical bad stats use from him

[SouthSideRyan]

The whole "it's the third year and they're going to have their worst year yet" part that really bothers me. I doubt even the most deluded "we gotta suck before can get better"-types thought that would have happened.

 

FWIW, I honestly think this is not only the bottoming out point, but that next year is going to be the best since 2009.

 

You mean we're going to win 72 games? Huzzah!

[Guest]

The whole "it's the third year and they're going to have their worst year yet" part that really bothers me. I doubt even the most deluded "we gotta suck before can get better"-types thought that would have happened.

 

FWIW, I honestly think this is not only the bottoming out point, but that next year is going to be the best since 2009.

 

You mean we're going to win 72 games? Huzzah!

 

We won 75 in 2010

[SouthSideRyan]

The whole "it's the third year and they're going to have their worst year yet" part that really bothers me. I doubt even the most deluded "we gotta suck before can get better"-types thought that would have happened.

 

FWIW, I honestly think this is not only the bottoming out point, but that next year is going to be the best since 2009.

 

You mean we're going to win 72 games? Huzzah!

 

We won 75 in 2010

 

Whoops was thinking this turd sandwich PLAN started in '11 post withdrawn

[Guest]

ok, think about it this way...

 

everyone agrees that the larger the sample, the more likely the statistic is going to reflect the true process, right?

 

after 50 games, you have 50 data points as far as w/l goes.

 

After 50 games, you have around 400 data points as far as runs scored / runs allowed goes.

Outcome of A is not independent of B. But when B says they should have around and a .500 winning percentage and A says something far, far different and worse, I tend to have more faith in A in the case of team like the 2014 Cubs (bad OPS and good pitching). I'd believe that run differential was more predictive for the 2014 Cubs if they had close to a .500 record and a negative run differential.

 

Agreed that copious amounts of data are useful because they tend to smooth out variations/outliers. The study posted a while back says 50 games is the point at which B is more predictive than A towards final record because of copious amounts of data. That DOES NOT NECESSARILY mean that B is more predictive than A for any individual case. The outliers may be the most important data (as Kyle sort of alluded to in a backwards fashion), because they show where the model breaks down. I know people like to attribute a lot to luck, but all luck is is another word for saying I don't know.

[Hairyducked Idiot]

"this is a special case because I feel like it is" is the ultimate dumb.

[Guest]

"this is a special case because I feel like it is" is the ultimate dumb.

You should know.