PDA

View Full Version : Luckiest pitcher?


brewcrew/chisox
07-01-2005, 10:52 AM
Maybe one of you can help me with this. Baseball Prospectus's stat of the day is a list of the top 5 "Luckiest" AL starters by LUCK.


Player Team W L E(W) E(L) LUCK
Chan Ho Park TEX 7 2 4.4 6. 6.58
Jon Garland CHA 12 3 7.0 4.6 6.57
Orlando Hernandez CHA 7 2 3.4 4.3 5.87
Matt Clement BOS 9 1 6.9 4.6 5.74
Gil Meche SEA 8 4 4.4 6.0 5.65



They define LUCK as "Luck, as measured by the number of extra wins, and short losses the pitcher actually got, versus his expected record. LUCK = (W-E(W))+(E(L)-L)"

:?:

Ok, I'm a bit confused. Is this a completely subjective stat here? What does "his expected record" mean and who is doing the predicitions? How are they getting that last stat at the end, and why are ElDuque and JG being considered "lucky"? Can one of you explain these numbers so that even an English teacher can understand?


link (http://www.baseballprospectus.com/)

edit: sorry I couldn't get the table to even up there but I think you should be able to read it.

Flight #24
07-01-2005, 10:57 AM
Maybe one of you can help me with this. Baseball Prospectus's stat of the day is a list of the top 5 "Luckiest" AL starters by LUCK.


Player Team W L E(W) E(L) LUCK
Chan Ho Park TEX 7 2 4.4 6. 6.58
Jon Garland CHA 12 3 7.0 4.6 6.57
Orlando Hernandez CHA 7 2 3.4 4.3 5.87
Matt Clement BOS 9 1 6.9 4.6 5.74
Gil Meche SEA 8 4 4.4 6.0 5.65



They define LUCK as "Luck, as measured by the number of extra wins, and short losses the pitcher actually got, versus his expected record. LUCK = (W-E(W))+(E(L)-L)"

:?:

Ok, I'm a bit confused. Is this a completely subjective stat here? What does "his expected record" mean and who is doing the predicitions? How are they getting that last stat at the end, and why are ElDuque and JG being considered "lucky"? Can one of you explain these numbers so that even an English teacher can understand?


link (http://www.baseballprospectus.com/)

edit: sorry I couldn't get the table to even up there but I think you should be able to read it.

There are a lot of possible answers, but the most accurate one is that BP has a model. When actual results diverge from their model, where most intelligent, objective people might say "hmmm....maybe maybe our model isn't as good as we thought", BP says "hmmmm....any divergence must be luck".

rdivaldi
07-01-2005, 10:59 AM
There are a lot of possible answers, but the most accurate one is that BP has a model. When actual results diverge from their model, where most intelligent, objective people might say "hmmm....maybe maybe our model isn't as good as we thought", BP says "hmmmm....any divergence must be luck".

Damn straight. BP has more "cover my ass" formulas than the federal government.

Ol' No. 2
07-01-2005, 11:03 AM
There are a lot of possible answers, but the most accurate one is that BP has a model. When actual results diverge from their model, where most intelligent, objective people might say "hmmm....maybe maybe our model isn't as good as we thought", BP says "hmmmm....any divergence must be luck".Exactly. It's based on comparison with other pitchers with similar numbers of IP and runs allowed. If you do better than that, it has to be luck because there can't possibly be any factors not accounted for.

scottjanssens
07-01-2005, 11:08 AM
It's a method of accounting for variences in performance such as number of innings pitched, run support received, good defence vs bad defense, etc. It doesn't mean that their formula says those pitchers should have those actual wins. It's a method of adjusting the numbers to make fair comparisons between players on different teams. Pitchers on the Sox benefit from the Sox stellar defense. Pitchers on other teams don't. Pitchers on Boston receive an average of 5.6 runs of support per game while pitchers on the Dodgers receive 4.3.

The formula tries to account for the variences so that proper comparisons can be made. I think it would be obvious that Garland wouldn't be 12-3 on a lot of other teams. On a few teams he might actually have a better record.

The expected W and L stats are simply to compare pitchers. It should not be used for any other purpose.

scottjanssens
07-01-2005, 11:12 AM
it has to be luck because there can't possibly be any factors not accounted for.

You're making their point for them. When they say luck they don't mean luck as in I found $20 on the street today, how lucky. They mean any influences outside the pitchers control. Things such as offensive support and defensive capability. The luck shows how strong a team the pitcher has behind him. If anything having two pitchers high in luck simply shows how good a team the Sox have.

Ol' No. 2
07-01-2005, 11:22 AM
You're making their point for them. When they say luck they don't mean luck as in I found $20 on the street today, how lucky. They mean any influences outside the pitchers control. Things such as offensive support and defensive capability. The luck shows how strong a team the pitcher has behind him. If anything having two pitchers high in luck simply shows how good a team the Sox have.My criticism of this is similar to my criticism of other stats: they're based on averages without accounting for variability. If the statistical variability is larger than the "luck" factor, how much significance can you put in luck? It's similar to comparing teams' W-L records with their Pythagorean expected W-L. The standard error in the Pythagorean W-L is +/- 4 games. Yet they'll seriously claim that one team is outperforming their expected W-L if they're 4 games above it. It's nonsense.

The other criticism is in the way they use run support. Using their formulae, a pitcher who wins a game 15-0 is not as good as one who wins 3-2. :bs:

Flight #24
07-01-2005, 11:30 AM
You're making their point for them. When they say luck they don't mean luck as in I found $20 on the street today, how lucky. They mean any influences outside the pitchers control. Things such as offensive support and defensive capability. The luck shows how strong a team the pitcher has behind him. If anything having two pitchers high in luck simply shows how good a team the Sox have.

You mean influences outside things they define as being within the pitchers control. That's the key arrogance here. Basically it says "Our model controls for everything the pitcher can control. Therefore anything else is by definition outside of his control."

Of course, they also like to think that a pitcher can't control what happens on a batted ball, i.e. he can't get "bad contact" to generate popups and/or easy grounders. They also like to think that a pitcher can't impact the quality of the defense behind him with his pace.

To that I say: :bs:

To BP, I say: :booty: :whatever:

scottjanssens
07-01-2005, 12:06 PM
Sheesh, so ignore the threads that mention BP stats then. You add nothing to the conversation.

The metric isn't meant to be the end all be all of measuring pitchers' wins. It's simply a way of getting a fairer comparison. No, it's not perfect. But it's better than simply saying I bet Garland wouldn't be doing so well on another team. Can anyone here honestly tell me they think Garland would be 12-3 on the Astros? Garland is having a great year. The Luck stat does not say otherwise. If you think it does, that's the interpretation you're putting on the stat. It's an incorrect interpretation.

As for the pythagorean win stat, that's more a curiosity than anything. Anyone who uses it to support an argument is an idiot. Anyone who uses stats while ignoring margin of error is an idiot, or at least ignorant.

Flight #24
07-01-2005, 01:00 PM
Sheesh, so ignore the threads that mention BP stats then. You add nothing to the conversation.

The metric isn't meant to be the end all be all of measuring pitchers' wins. It's simply a way of getting a fairer comparison. No, it's not perfect. But it's better than simply saying I bet Garland wouldn't be doing so well on another team. Can anyone here honestly tell me they think Garland would be 12-3 on the Astros? Garland is having a great year. The Luck stat does not say otherwise. If you think it does, that's the interpretation you're putting on the stat. It's an incorrect interpretation.

As for the pythagorean win stat, that's more a curiosity than anything. Anyone who uses it to support an argument is an idiot. Anyone who uses stats while ignoring margin of error is an idiot, or at least ignorant.

It's not the stats that are the problem, it's the usage of them. If you want to say Garland wouldn't have as many wins because he's a groundball pitcher and he'd have a worse D, that's one thing. But you should then at least acknowledge that Garland is specifically trying to let guys hit the ball and get ground balls because he's got a good D, so in another situation he might not pitch the same. You can make similar arguments about EVERY pitcher.

But BP likes to gloss over it and just call certain guys "lucky", because their model says that ther are only a certain set of factors which determine success. Which is frankly insulting to guys like Garland that don't fall within their parameters.

fquaye149
07-01-2005, 01:11 PM
Sheesh, so ignore the threads that mention BP stats then. You add nothing to the conversation.

The metric isn't meant to be the end all be all of measuring pitchers' wins. It's simply a way of getting a fairer comparison. No, it's not perfect. But it's better than simply saying I bet Garland wouldn't be doing so well on another team. Can anyone here honestly tell me they think Garland would be 12-3 on the Astros? Garland is having a great year. The Luck stat does not say otherwise. If you think it does, that's the interpretation you're putting on the stat. It's an incorrect interpretation.

As for the pythagorean win stat, that's more a curiosity than anything. Anyone who uses it to support an argument is an idiot. Anyone who uses stats while ignoring margin of error is an idiot, or at least ignorant.

Garland would probably still have 2 complete game shutouts and a middle 3 era on the Astros. That is to say, run support doesn't dictate that as much. So yes, he would be pitching as well.

Huisj
07-01-2005, 03:27 PM
Garland would probably still have 2 complete game shutouts and a middle 3 era on the Astros. That is to say, run support doesn't dictate that as much. So yes, he would be pitching as well.

Heck, he might have 3 shutouts and an ERA of about 3. He'd get to face a pitcher 2 or 3 times a game.

Those NL pitchers sure are lucky to get to face a pitcher each time through the order.

ondafarm
07-01-2005, 03:33 PM
Certain managers consistently beat their expected wins. Chuck Tanner and Sparky Anderson come to mind. I'm not putting Ozzie in that group but he does seem to understand Garland very well and know when he's starting to falter. As opposed to Buehrle. I think Garland is lucky to have a manager like Ozzie.

FarWestChicago
07-01-2005, 03:45 PM
But BP likes to gloss over it and just call certain guys "lucky", because their model says that ther are only a certain set of factors which determine success. Which is frankly insulting to guys like Garland that don't fall within their parameters.That's the problem with BP and FOBB's. They make up a model of reality and declare it "reality". Then they declare themselves experts at "reality", and therefore superior to everybody else. Then they use loaded terms like "luck" and act wronged when some member of the "inferior masses" takes them to task for it. It's really quite amusing if you look at it from a certain perspective. What a bunch of silliness. :D:

ondafarm
07-01-2005, 03:46 PM
But stats are life !!!!

PaleHoseGeorge
07-01-2005, 03:55 PM
The Baseball Prospectus is strictly for amateur statisticians and (worse) amateur statistician wannabes. Some of the people who have "writer" credentials there are certifiably stupid.

It's a giant waste of time for anybody with better things to do this side of a propellerheaded geek.

SOXfnNlansing
07-01-2005, 07:58 PM
Maybe one of you can help me with this. Baseball Prospectus's stat of the day is a list of the top 5 "Luckiest" AL starters by LUCK.


Player Team W L E(W) E(L) LUCK
Chan Ho Park TEX 7 2 4.4 6. 6.58
Jon Garland CHA 12 3 7.0 4.6 6.57
Orlando Hernandez CHA 7 2 3.4 4.3 5.87
Matt Clement BOS 9 1 6.9 4.6 5.74
Gil Meche SEA 8 4 4.4 6.0 5.65



They define LUCK as "Luck, as measured by the number of extra wins, and short losses the pitcher actually got, versus his expected record. LUCK = (W-E(W))+(E(L)-L)"

:?:

Ok, I'm a bit confused. Is this a completely subjective stat here? What does "his expected record" mean and who is doing the predicitions? How are they getting that last stat at the end, and why are ElDuque and JG being considered "lucky"? Can one of you explain these numbers so that even an English teacher can understand?


link (http://www.baseballprospectus.com/)

edit: sorry I couldn't get the table to even up there but I think you should be able to read it.Expected record goes hand in hand with Simulated games played. wood/prior are the best 'unluckiest' pitchers in the league

Ol' No. 2
07-02-2005, 08:30 AM
Sheesh, so ignore the threads that mention BP stats then. You add nothing to the conversation.

The metric isn't meant to be the end all be all of measuring pitchers' wins. It's simply a way of getting a fairer comparison. No, it's not perfect. But it's better than simply saying I bet Garland wouldn't be doing so well on another team. Can anyone here honestly tell me they think Garland would be 12-3 on the Astros? Garland is having a great year. The Luck stat does not say otherwise. If you think it does, that's the interpretation you're putting on the stat. It's an incorrect interpretation.

As for the pythagorean win stat, that's more a curiosity than anything. Anyone who uses it to support an argument is an idiot. Anyone who uses stats while ignoring margin of error is an idiot, or at least ignorant.I'm repeating myself here, but the problem with these stats is the implicit assumption that they've accounted for all the variables, so any differences must be due to X. That's false. Baseball is a game of almost infinite variability. All these other factors show up as variability around the predicted outcome. I brought up the Pythagorean W-L as an example, but the same problem plagues all statistical analysis (not just baseball). When the differences between player A and player B is smaller than the statistical variation, you can't conclude there's a real difference. And since the statistical variation is typically large, it's rare that you can find real differences, and these tend to be so obvious that it's not necessary to use a convoluted statistic to see them. But you rarely see anyone show the raw statistical variation.

downstairs
07-02-2005, 08:55 AM
I'm repeating myself here, but the problem with these stats is the implicit assumption that they've accounted for all the variables, so any differences must be due to X. That's false. Baseball is a game of almost infinite variability. All these other factors show up as variability around the predicted outcome. I brought up the Pythagorean W-L as an example, but the same problem plagues all statistical analysis (not just baseball). When the differences between player A and player B is smaller than the statistical variation, you can't conclude there's a real difference. And since the statistical variation is typically large, it's rare that you can find real differences, and these tend to be so obvious that it's not necessary to use a convoluted statistic to see them. But you rarely see anyone show the raw statistical variation.


I am not a math expert, but I do happen to love the more complex statistical side of baseball.

I agree with you to a certain point, but you when you say "it's rare that you can find real differences, and these tend to be so obvious that it's not necessary to use a convoluted statistic to see them".... I don't agree.

You have to admit that the standard stats we use for players don't come anywhere close to telling their story. Pitchers' W-L, ERA, SO, BB are a minor part of who they are and how good they are. Batters' HR, Avg., BB, SO, etc only tell us surface-level stuff.

I don't disagree that many stats are thrown together in a manner that is just plain silly. But they're no worse than saying "Hank Aaron is the best player of all time because he hit 755 home runs."

Ol' No. 2
07-02-2005, 09:09 AM
I am not a math expert, but I do happen to love the more complex statistical side of baseball.

I agree with you to a certain point, but you when you say "it's rare that you can find real differences, and these tend to be so obvious that it's not necessary to use a convoluted statistic to see them".... I don't agree.

You have to admit that the standard stats we use for players don't come anywhere close to telling their story. Pitchers' W-L, ERA, SO, BB are a minor part of who they are and how good they are. Batters' HR, Avg., BB, SO, etc only tell us surface-level stuff.

I don't disagree that many stats are thrown together in a manner that is just plain silly. But they're no worse than saying "Hank Aaron is the best player of all time because he hit 755 home runs."One thing you have to remember is that when you combine several raw numbers in one formula to produce a single result, the variability of the raw numbers ADD TOGETHER to produce a higher variability in the resulting number. The more numbers that go into the equation, the bigger the variability in the result. This is why these highly derivative statistics are often mostly junk. It's rare that you see anyone show the variability of the statistic, but when they do you quickly realize that the variability overwhelms the small differences observed. A good example was an article I saw on PECOTA projections of OPS. The variability was something like +/- 80 pts! What's THAT worth? I could do at least as well just taking the average of the last three years.