PDA

View Full Version : Expected W-L


LITTLE NELL
07-29-2012, 08:57 AM
In ESPN's expanded standings they have the Sox expected W-L record at 57-43. Does anyone know how that is figured, maybe by blown saves or something like that.

DonnieDarko
07-29-2012, 09:09 AM
In ESPN's expanded standings they have the Sox expected W-L record at 57-43. Does anyone know how that is figured, maybe by blown saves or something like that.

Nope, not based on blown saves. Here's the formula:

RS^1.82/((RS^1.82)+(RA^1.82))

RS = Runs Scored
RA = Runs Allowed

SI1020
07-29-2012, 09:17 AM
So how come their Pythagorean W-L according to baseball-reference is 56-44? That looks like the same formula.

DonnieDarko
07-29-2012, 09:25 AM
So how come their Pythagorean W-L according to baseball-reference is 56-44? That looks like the same formula.

I think they are the exact same formula, just under a different name.

Tragg
07-29-2012, 09:37 AM
In ESPN's expanded standings they have the Sox expected W-L record at 57-43. Does anyone know how that is figured, maybe by blown saves or something like that.

First order wins - adjusted for margin of victory
Second order wins - adjusted for how many runs should have been scored based on underlying stats
Third order wins - second order wins plus adjustments for strength of schedule

Not sure what ESPN's using - probably first

SI1020
07-29-2012, 09:47 AM
I think judging a team based on run differential is stupid. Phythagorean W-L is one stat I try my best to ignore. The 2005 Cleveland Indians were world champions, right?

central44
07-29-2012, 12:25 PM
I think judging a team based on run differential is stupid. Phythagorean W-L is one stat I try my best to ignore. The 2005 Cleveland Indians were world champions, right?


I don't think it's perfect but it does tell a lot about a team--it's a more relevant stat than say, wins and losses for a pitcher. A team with a good run differential is more than likely good at scoring runs while preventing them for the other team, which is the mark of a talented club.

That said, there are always exceptions--like the 2005 Sox who made a habit of winning those 1-0 games. Again, not perfect, but not the worst stat we use.

johnny bench
07-29-2012, 12:35 PM
In ESPN's expanded standings they have the Sox expected W-L record at 57-43. Does anyone know how that is figured, maybe by blown saves or something like that.

You can find the ESPN definition on the Relative Power Index (http://espn.go.com/mlb/stats/rpi) page. A footnote to this page reads: "*ExW-L and ExWP are derived from Bill James' Pythagorean theorem of baseball: Runs scored [squared] / (Runs scored [squared] + runs allowed [squared]). This formula was designed to relate a team's runs scored and runs allowed to its won-lost record."

The result of this calculation is a number between 0 and 1, which is the expected winning percentage for a team. Multiply this percentage by the number of games played to-date in the season to yield the number of expected wins.

DSpivack
07-29-2012, 12:45 PM
I don't think it's perfect but it does tell a lot about a team--it's a more relevant stat than say, wins and losses for a pitcher. A team with a good run differential is more than likely good at scoring runs while preventing them for the other team, which is the mark of a talented club.

That said, there are always exceptions--like the 2005 Sox who made a habit of winning those 1-0 games. Again, not perfect, but not the worst stat we use.

You can find the ESPN definition on the Relative Power Index (http://espn.go.com/mlb/stats/rpi) page. A footnote to this page reads: "*ExW-L and ExWP are derived from Bill James' Pythagorean theorem of baseball: Runs scored [squared] / (Runs scored [squared] + runs allowed [squared]). This formula was designed to relate a team's runs scored and runs allowed to its won-lost record."

The result of this calculation is a number between 0 and 1, which is the expected winning percentage for a team. Multiply this percentage by the number of games played to-date in the season to yield the number of expected wins.

It seems to me it ignores variance in a team's run scoring. I remember Sox teams of the early 2000s who would have a high total run total and thus a good run differential, but they'd score 10 one game and 0 the next. One team might have a higher mean in average runs; another may have a lower mean but a higher median. I would rather have a consistent offense, even if meant fewer total runs.

SI1020
07-29-2012, 01:26 PM
It seems to me it ignores variance in a team's run scoring. I remember Sox teams of the early 2000s who would have a high total run total and thus a good run differential, but they'd score 10 one game and 0 the next. One team might have a higher mean in average runs; another may have a lower mean but a higher median. I would rather have a consistent offense, even if meant fewer total runs. Thank you.

doublem23
07-29-2012, 01:52 PM
It seems to me it ignores variance in a team's run scoring. I remember Sox teams of the early 2000s who would have a high total run total and thus a good run differential, but they'd score 10 one game and 0 the next. One team might have a higher mean in average runs; another may have a lower mean but a higher median. I would rather have a consistent offense, even if meant fewer total runs.

It does, but, in general, the greater the run differential between runs scored and runs allowed, the better the team. Of course there are exceptions to the rule, but in general it evens out.

Anyways, B-R's Pythagorean W-L uses 1.83 as the exponent in its equation. If ESPN's ExW-L is using 1.82 as their exponent, that probably accounts for the slight variance. Really not a big deal in the long run.

TDog
07-29-2012, 02:05 PM
It seems to me it ignores variance in a team's run scoring. I remember Sox teams of the early 2000s who would have a high total run total and thus a good run differential, but they'd score 10 one game and 0 the next. One team might have a higher mean in average runs; another may have a lower mean but a higher median. I would rather have a consistent offense, even if meant fewer total runs.

You're right. That's why expected wins and Pythagorean records are meaningless stats. The best teams win a lot of one-run games. The worst teams lose a lot of one-run games. This isn't isn't a matter of luck. If a team is up by two in the ninth with a runner on third and one out, there isn't going to be a play at the plate on a fielded ground ball. You really don't know what run differential means unless you see how the runs are distributed.

For expected wins to mean something, the stat would have to look at things like blown saves, but wouldn't treat blown saves in the seventh the same as blown saves in the ninth, and would have to consider those rare games when teams gave up leads too big to qualify as blown saves for individual pitchers. And if you say this team whould have won because it couged up a 3-2 win in the ninth, you can't really say this team should have lost because it scored two in the ninth two win 4-3. There are a lot of games that both teams really should be expected win. There ware many extra-inning games both teams are just a productive out away from a win.

Expected wins is a stat that is meant to measure the quality of a team, but it treats all runs scored and given up by and against all teams as if they are equally meaningful. As a stat, it really doesn't tell you anything.

SBSoxFan
07-29-2012, 02:15 PM
Nope, not based on blown saves. Here's the formula:

RS^1.82/((RS^1.82)+(RA^1.82))

RS = Runs Scored
RA = Runs Allowed

I've seen this done for football as well. The Pythagorean theorem relates the sides of right triangles, e.g., a^2 + b^2 = c^2, where a and b are lengths of the two sides of a triangle which include the right angle and c is the length of the third side which is opposite the right angle (the hypotenuse). A common example of this is the 3, 4, 5 right triangle since 9 + 16 = 25.

The Pythagoreans did not believe in irrational numbers, like 1.82, for example. So, it's a misnomer to refer to such an equation as Pythagorean anything.

The Pythagoreans were also a bit radical and would have "removed" any followers who suggested such a number existed.

It seems to me it ignores variance in a team's run scoring. I remember Sox teams of the early 2000s who would have a high total run total and thus a good run differential, but they'd score 10 one game and 0 the next. One team might have a higher mean in average runs; another may have a lower mean but a higher median. I would rather have a consistent offense, even if meant fewer total runs.

I've always felt the same way. It's certainly off at an extreme case where a team wins nine games 1-0 and loses its tenth game 10-1. So, the actual record is 9-1 versus a predicted of 5-5.

The last time the Sox were sweeping Texas, I calculated the average and standard deviation of runs for and runs against for both teams in 2012 year to date. I found that the standard deviations of each team were so large that any analysis would likely not be statistically significant.

TDog
07-29-2012, 02:57 PM
I've seen this done for football as well. The Pythagorean theorem relates the sides of right triangles, e.g., a^2 + b^2 = c^2, where a and b are lengths of the two sides of a triangle which include the right angle and c is the length of the third side which is opposite the right angle (the hypotenuse). A common example of this is the 3, 4, 5 right triangle since 9 + 16 = 25. ...

This reminds me of a high school geometry class about 40 years ago when Mr. Wroblewski called on me to answer a homework question asking the distance between home plate and second base. I hadn't done my homework but I had spent plenty of time reading the Major League Baseball rulebook, which includes all the infield dimensions. I answered what I knew, that it was 127 feet, 3 3/8 inches. Of course, I was told I was wrong. The correct answer wasn't in feet and inches, but a square root. I didn't challenge Mr. Wroblewski to go out and measure it.

I was wrong because I knew the actual measurement. But if he had asked me how many wins the White Sox had that morning, he wouldn't have had me do some formula based on run differential. It seems sort of silly to look at how many runs teams score out of context and how many runs they give up out of context and come up with what is supposed to be some sort of standard quality for that team.

SBSoxFan
07-29-2012, 04:04 PM
This reminds me of a high school geometry class about 40 years ago when Mr. Wroblewski called on me to answer a homework question asking the distance between home plate and second base. I hadn't done my homework but I had spent plenty of time reading the Major League Baseball rulebook, which includes all the infield dimensions. I answered what I knew, that it was 127 feet, 3 3/8 inches. Of course, I was told I was wrong. The correct answer wasn't in feet and inches, but a square root. I didn't challenge Mr. Wroblewski to go out and measure it.

I was wrong because I knew the actual measurement. But if he had asked me how many wins the White Sox had that morning, he wouldn't have had me do some formula based on run differential. It seems sort of silly to look at how many runs teams score out of context and how many runs they give up out of context and come up with what is supposed to be some sort of standard quality for that team.

Pedagogically speaking, your teacher should not have told you that your answer was incorrect just because it was in a different form then he expected. However, I also think it's important to understand where this stuff comes from. I cringed when I read the phrase "Bill James' Pythagorean theorem" in a previous post.

On the baseball side, I've never understood the attraction since expected W/L is not a predictive tool. What's done is done; all it can lead to is frustration over the idea that a team lost games it "should not" have.

Frater Perdurabo
07-29-2012, 04:30 PM
On the baseball side, I've never understood the attraction since expected W/L is not a predictive tool. What's done is done; all it can lead to is frustration over the idea that a team lost games it "should not" have.

Agreed. But I'll go further and say we really don't need a formula to find instances where a team loses games they should have won.

SI1020
07-29-2012, 04:38 PM
The Pythagoreans were also a bit radical and would have "removed" any followers who suggested such a number existed.
They had much in common with some of the devotees of the Jamesian revolution. Of course as of now, they just try to remove you from the conversation. They haven't yet reached the numbers and the power to take the next step.

doublem23
07-29-2012, 04:44 PM
On the baseball side, I've never understood the attraction since expected W/L is not a predictive tool. What's done is done; all it can lead to is frustration over the idea that a team lost games it "should not" have.

Actually it is; at the very least it's no better or worse than traditional W-L for predictive purposes. But off people choose to not use all information available to them, that is their own, personal preference. But let's not lie and say it's useless.

Frater Perdurabo
07-29-2012, 05:18 PM
Actually it is; at the very least it's no better or worse than traditional W-L for predictive purposes. But off people choose to not use all information available to them, that is their own, personal preference. But let's not lie and say it's useless.

It's dubious even as a tool to predict the past.

doublem23
07-29-2012, 06:03 PM
It's dubious even as a tool to predict the past.

Probably because nobody in their right mind would want to do that

WhiteSox5187
07-29-2012, 06:08 PM
Probably because nobody in their right mind would want to do that

Isn't that what Pythagorean Win Theorem does? It looks at the runs scored v. runs saved or whatever and say "This is what a team's record SHOULD be"?

Frater Perdurabo
07-29-2012, 07:20 PM
Probably because nobody in their right mind would want to do that

Pythagorean W-L is the baseball equivalent of counterfactual history. It's like asking, "What if the South won the Civil War?" Interesting to discuss? Sure. But ultimately it is meaningless.

doublem23
07-30-2012, 12:08 AM
Pythagorean W-L is the baseball equivalent of counterfactual history. It's like asking, "What if the South won the Civil War?" Interesting to discuss? Sure. But ultimately it is meaningless.

Uh... no? Runs scored vs. runs allowed are very real, tangible stats. They're the most crucial in all of baseball.

You guys are correct in your critique of Pythagorean W-L that it doesn't reward teams for winning 1-0 games... And that's because generally speaking, the team that regularly wins games 6-2 or 7-3 is better than the team that wins games 1-0 or 2-1.

lpneck
07-30-2012, 12:10 AM
The Pythagoreans did not believe in irrational numbers, like 1.82, for example. So, it's a misnomer to refer to such an equation as Pythagorean anything.

1.82 is not an irrational number.

Frater Perdurabo
07-30-2012, 05:50 AM
Uh... no? Runs scored vs. runs allowed are very real, tangible stats. They're the most crucial in all of baseball.

You guys are correct in your critique of Pythagorean W-L that it doesn't reward teams for winning 1-0 games... And that's because generally speaking, the team that regularly wins games 6-2 or 7-3 is better than the team that wins games 1-0 or 2-1.

The real, actual Won-Loss record is the only thing that counts in determining if a team wins the division. I'm not trying to be a smart ass, but anything else is just for arguments, and arguments can be made with stats and/or with observations. I agree that the most compelling arguments are supported by both quantitative and qualitative evidence, but arguments don't make the playoffs.

Frater Perdurabo
07-30-2012, 05:56 AM
Division titles are based on REAL W-L records.

Teams don't win squat based on what their run differential, plugged into a formula, suggests their record SHOULD BE.

That's what I mean by "meaningless."

doublem23
07-30-2012, 07:52 AM
I don't even know who you're even trying to argue with, nobody disagrees with any of that. Just because Pythagorean W-L is a tool some people choose to use to help their analysis of the game doesn't mean they automatically think it's Gospel.

I honestly have no idea what point you're even trying to make.

asindc
07-30-2012, 09:13 AM
While FP can certainly speak for himself, the bottom line for me is that Pythagorean W-L is speculative in nature.

doublem23
07-30-2012, 09:25 AM
While FP can certainly speak for himself, the bottom line for me is that Pythagorean W-L is speculative in nature.

Of course it is, everything is speculative if you're trying to use it for predictive purposes, otherwise the team with the better W-L record would always beat the team with the worse W-L record every time, right?

You guys seem to be arguing against a non-existent group of people who take Pythagorean W-L as the only and end-all tool used to determine team strength. Absolutely nobody who understands what that stat measures thinks that way about it. You're basically just screaming at a brick wall.

SephClone89
07-30-2012, 09:30 AM
Of course it is, everything is speculative if you're trying to use it for predictive purposes, otherwise the team with the better W-L record would always beat the team with the worse W-L record every time, right?

You guys seem to be arguing against a non-existent group of people who take Pythagorean W-L as the only and end-all tool used to determine team strength. Absolutely nobody who understands what that stat measures thinks that way about it. You're basically just screaming at a brick wall.

:gulp:

asindc
07-30-2012, 09:58 AM
Of course it is, everything is speculative if you're trying to use it for predictive purposes, otherwise the team with the better W-L record would always beat the team with the worse W-L record every time, right?

You guys seem to be arguing against a non-existent group of people who take Pythagorean W-L as the only and end-all tool used to determine team strength. Absolutely nobody who understands what that stat measures thinks that way about it. You're basically just screaming at a brick wall.

I disagree with that. There are those who argue that one team is better than another team based on this stat. You obviously don't, but there are others who do.

SI1020
07-30-2012, 10:28 AM
I disagree with that. There are those who argue that one team is better than another team based on this stat. You obviously don't, but there are others who do. Or that FIP is the be all and end all to evaluating pitching. Just like Billy Beane and his Scott Hatteberg OBP miracle were the second coming. One flavor of the month after another. It's become a giant cacophony of confusion.

doublem23
07-30-2012, 10:46 AM
I disagree with that. There are those who argue that one team is better than another team based on this stat. You obviously don't, but there are others who do.

Or that FIP is the be all and end all to evaluating pitching. Just like Billy Beane and his Scott Hatteberg OBP miracle were the second coming. One flavor of the month after another. It's become a giant cacophony of confusion.

I personally think you guys are either misinterpretting what you are reading out there, purposefully distorting what you are seeing to justify your own beliefs about so called "sabermetrics", or perhaps, just reading the work of dumb ****ing morons. Let's be clear, NOBODY with the slightest understanding of baseball stats thinks there is one, end-all, be-all God statistic that perfectly sums up everything. The sport is way too complicated for that. Even WAR, which aims to be the one and only stat you need to know, is constantly being tweaked or there are slight variances from site to site. Everyone who works with baseball stats, either professionally or as a hobby is aware of this. If you're reading a guy who is telling you that FIP is all you need to know about pitchers, or OBP is all you need to know about hitters, or Pythagorean W-L is all you need to know about relative team strength, STOP READING WHAT THAT IDIOT IS WRITING. Both of you are smart guys with a strong understanding of the sport. You don't need to spend time and energy getting all worked up over what some nonsensical doofus is barfing out of his computer. Nobody with a working brain would argue these points.

That doesn't mean some stats carry more weight than others. Is FIP better than ERA? Yes. Is OBP better than BA? Yes. Is Pythagorean W-L better than traditional W-L (for predictive purposes)? Yes. But they are all pieces of the puzzle. No one will tell you everything you need to know.

asindc
07-30-2012, 10:53 AM
I personally think you guys are either misinterpretting what you are reading out there, purposefully distorting what you are seeing to justify your own beliefs about so called "sabermetrics", or perhaps, just reading the work of dumb ****ing morons. Let's be clear, NOBODY with the slightest understanding of baseball stats thinks there is one, end-all, be-all God statistic that perfectly sums up everything. The sport is way too complicated for that. Even WAR, which aims to be the one and only stat you need to know, is constantly being tweaked or there are slight variances from site to site. Everyone who works with baseball stats, either professionally or as a hobby is aware of this. If you're reading a guy who is telling you that FIP is all you need to know about pitchers, or OBP is all you need to know about hitters, or Pythagorean W-L is all you need to know about relative team strength, STOP READING WHAT THAT IDIOT IS WRITING. Both of you are smart guys with a strong understanding of the sport. You don't need to spend time and energy getting all worked up over what some nonsensical doofus is barfing out of his computer. Nobody with a working brain would argue these points.

That doesn't mean some stats carry more weight than others. Is FIP better than ERA? Yes. Is OBP better than BA? Yes. Is Pythagorean W-L better than traditional W-L (for predictive purposes)? Yes. But they are all pieces of the puzzle. No one will tell you everything you need to know.

I wasn't saying there are those who say Pythagorean W-L is the ultimate predictor, I was and am saying that there are those ("idiots," as you call them) who proclaim that the stat, along with others, can tell you who is the better team in the present time. It is that line of reasoning that I disagree with. By the way, I don't spend nearly enough time needed to get worked up about stats, traditional or sabermetric.

doublem23
07-30-2012, 10:59 AM
I wasn't saying there are those who say Pythagorean W-L is the ultimate predictor, I was and am saying that there are those ("idiots," as you call them) who proclaim that the stat, along with others, can tell you who is the better team in the present time. It is that line of reasoning that I disagree with. By the way, I don't spend nearly enough time needed to get worked up about stats, traditional or sabermetric.

Anybody looking at Pythagorean W-L and Pythagorean W-L alone and making a sweeping statement like that is, yes, an idiot. I mean, seriously, if you're going to be that lazy in your analysis, at least use SRS.

amsteel
07-30-2012, 11:15 AM
I think Pyth W-L is a good place to start when evaluating a manager. If a team is performing below the expected W-L they're probably losing alot of close games.

It's not a perfect measurement, but it's a place to start.

In general, I wouldn't say it's an accurate predictor of future results, rather an evaluation of past performance.

doublem23
07-30-2012, 11:26 AM
In general, I wouldn't say it's an accurate predictor of future results, rather an evaluation of past performance.

It's not a great predictor of future results, but generally speaking, when compared to traditional W-L record, it is better.

TDog
07-30-2012, 11:54 AM
Uh... no? Runs scored vs. runs allowed are very real, tangible stats. They're the most crucial in all of baseball.

You guys are correct in your critique of Pythagorean W-L that it doesn't reward teams for winning 1-0 games... And that's because generally speaking, the team that regularly wins games 6-2 or 7-3 is better than the team that wins games 1-0 or 2-1.

But what if the team that wins a lot of 1-0 and 2-1 games wins 10 more games than the team that wins a lot of 6-2 and 7-3 games while losing a lot of 1-0 and 2-1 games, or even 6-5 games? Even if the teams haf identical records, in major league baseball, expected wins doesn't really tell you anything. There are teams who have gone on to win championships who could be outscored in a three-game series and still win two of three games.

The White Sox this year lost four runs in their run differential because of four runs allowed in the ninth of what was a 7-0 game by a pitcher who was saving the bullpen and is no longer on the team. Does the fact that Brent Morel had only five hits and five RBIs in 27 at bats with runners in scoring position, negatively affecting White Sox run production, make them a weaker team even though the pitcher who gave up the four meaningless runs in the ninth was traded for the a different third baseman who has more than four times the RBIs as Morel in seven fewer games?

Every team changes during the season, good teams as well as bad, for better as well as worse. Look at the White Sox opening day lineup and the starting pitcher. Look at who closed the 1-0 win on opening day in April 2005 and who closed the 1-0 win that ended the 2005 season on October 26.

Runs scored and runs against are interesting stats because they tell you about a team's success on offense and defense. But using those numbers to come up with a should-be win-loss record is out of context. The runs the White Sox scored against the Rangers in the first game of their lost homestand are not as important as the runs they didn't score Sunday night.

doublem23
07-30-2012, 12:16 PM
But what if the team that wins a lot of 1-0 and 2-1 games wins 10 more games than the team that wins a lot of 6-2 and 7-3 games while losing a lot of 1-0 and 2-1 games, or even 6-5 games? Even if the teams haf identical records, in major league baseball, expected wins doesn't really tell you anything. There are teams who have gone on to win championships who could be outscored in a three-game series and still win two of three games.

:facepalm:

Yes, you are correct, it is possible to win 2 of 3 games in a series and be outscored. The majority of the time, that doesn't happen. Exceptions to the rule doesn't invalidate the rule.

The White Sox this year lost four runs in their run differential because of four runs allowed in the ninth of what was a 7-0 game by a pitcher who was saving the bullpen and is no longer on the team. Does the fact that Brent Morel had only five hits and five RBIs in 27 at bats with runners in scoring position, negatively affecting White Sox run production, make them a weaker team even though the pitcher who gave up the four meaningless runs in the ninth was traded for the a different third baseman who has more than four times the RBIs as Morel in seven fewer games?

Every team changes during the season, good teams as well as bad, for better as well as worse. Look at the White Sox opening day lineup and the starting pitcher. Look at who closed the 1-0 win on opening day in April 2005 and who closed the 1-0 win that ended the 2005 season on October 26.


And this is exactly why Pythagorean W-L is just one of the many tools guys look at. This is exactly why nobody takes it as the one and only, all-telling statistic. But, FWIW, even through Brent Morel's Garbage-A-Thon, the Sox's Pythagorean W-L record was the best in the division (through his last game, May 17), which has accurately forecasted them to lead the division for most of the season from that date.

How many ****ing times do I have to reiterate that this is not some kind of all-powerful, magical, tell-all statistic? Just because a stat doesn't tell you EVERYTHING doesn't mean it says nothing. Just like the team with the best team batting average isn't always the best offensive team in the league. Or just like the team with the best ERA isn't necessarily the best pitching team in the league. What the hell isn't clear about this?

Runs scored and runs against are interesting stats because they tell you about a team's success on offense and defense. But using those numbers to come up with a should-be win-loss record is out of context. The runs the White Sox scored against the Rangers in the first game of their lost homestand are not as important as the runs they didn't score Sunday night.

Jesus Tap Dancing ****ing Christ

Look, teams that score more runs and allow fewer runs over the course of many games are generally better than teams that score less runs and allow more. That's really all there is to it.

NardiWasHere
07-30-2012, 12:58 PM
Doublem, I usually agree with you about 99% of the time... But I don't see much value in looking at run differential to either look at the future or learn about the past.

I've also seen writers heavily rely on it as well... Especially in preseason predictions.

You say it's better than regular win-loss to use to predict the future. That may be true, but I don't think either are very good.

If you use it just to point out a team out scored or got out scored and make no further judgments without looking at the specifics, then it is ok, I guess. But how useful is that?

kittle42
07-30-2012, 02:34 PM
Look, teams that score more runs and allow fewer runs over the course of many games are generally better than teams that score less runs and allow more. That's really all there is to it.

But, but, but....I can name an annual exception! Thus, your rule is completely off base!

Fired up, frustrated doubs rules!

kittle42
07-30-2012, 02:35 PM
If you use it just to point out a team out scored or got out scored and make no further judgments without looking at the specifics, then it is ok, I guess. But how useful is that?

Thanks to this discussion, I looked into the methodology some more, and while I also question its usefulness (in sort of a "who really cares after the fact?" way), TDog's reasons are certainly not invalidators.

TDog
07-30-2012, 02:44 PM
...
Look, teams that score more runs and allow fewer runs over the course of many games are generally better than teams that score less runs and allow more. That's really all there is to it.

Of course, but the idea that you can take runs scored against one team and runs another team scores against that team and combine them for some sort of expected win total is looking for a truth in stats that isn't there.

The idea that teams that win a lot of one-run games are lucky ignores the fact that good teams have shut-down late-inning bullpens or that losing teams are losing because they don't. It ignores the realities of major league baseball. I am only surpised that someone hasn't come up with an expected win stat based on earned-run differentials that might get the 1970 White Sox up to 72 wins.

Expected wins don't work as a predictor over the course of a season. They don't approximate how well teams have actually done over the course of a season as well as actual wins and losses do. Run differential is irrelevant in a short series. The 2005 White Sox outscored the Astros by six runs in the World Series. The 2011 Cardinals outscored the Rangers by eight in the World Series, and they lost three games. In the 1960 World Series, the Yankees outscored the Pirates 55-27, but the Pirates won series.

Expected wins is simply an irrelevant stat. I am surprised that anyone takes places any value in it.

Frater Perdurabo
07-30-2012, 02:58 PM
I agree that Pythagorean W-L is one of many useful statistical tools that together somewhat have enhanced our understanding of the infinite complexities of the game.

I just wouldn't put much stock in it alone, if divorced both from other traditional stats and new SABR metrics, as well as from traditional qualitative scouting evaluations.

doublem23
07-30-2012, 03:10 PM
The idea that teams that win a lot of one-run games are lucky ignores the fact that good teams have shut-down late-inning bullpens or that losing teams are losing because they don't. It ignores the realities of major league baseball.

Yeah, but that's only one in what could be a million reasons why a team might win or lose more 1-run games than average. Maybe they just have a bad offense. Maybe they have bad pitching. Maybe they play in a pitcher's park. Maybe runs are just down across the league. Maybe they just are ****ing lucky. There's literally a zillion reasons that could have nothing do to with a team's bullpen to explain this phenomenon. You're missing forest for a ****ing leaf.

And, added to that, whose to say having a lights out, dynamic bullpen ISN'T a stroke of luck? The 2005 White Sox had an exceptional bullpen. The 2006 team didn't, even though many of the faces remained unchanged. Ditto the 2008 and 2009 Rays. Of all the elements that go into making a winning (or losing) baseball team, a good or bad bullpen probably is more reliant on luck than any other. Very few relievers have long, dominant careers; they either shine brightly and then burn out or they have flukey career years for no real explainable reason whatsoever.

I am only surpised that someone hasn't come up with an expected win stat based on earned-run differentials that might get the 1970 White Sox up to 72 wins.

The reason nobody has come up with that stat because it would be a wholly stupid stat.


Expected wins don't work as a predictor over the course of a season. They don't approximate how well teams have actually done over the course of a season as well as actual wins and losses do. Run differential is irrelevant in a short series. The 2005 White Sox outscored the Astros by six runs in the World Series. The 2011 Cardinals outscored the Rangers by eight in the World Series, and they lost three games. In the 1960 World Series, the Yankees outscored the Pirates 55-27, but the Pirates won series.

FOR THE UMPTEENTH MILLION TIME, NOBODY IS SAYING THAT PYTHAGOREAN W-L IS A FOOLPROOF PREDICTOR OF FUTURE RESULTS. WHO THE MOTHER****ING ****ING **** ARE YOU ARGUING THIS POINT WITH?!!?!?!?!

Every stat is basically meaningless in a short series. Just last weekend the A's swept the Yankees and there's literally no possible way anyone could conclude the A's are a better team than New York. Your point is news to exactly zero people.


Expected wins is simply an irrelevant stat. I am surprised that anyone takes places any value in it.

Expected wins is a relevant stat as long as you know what it's use is. It is not intended to predict future results, nor is it intended to rewrite historical ones. It's use is simply to define the relationship between runs scored and runs allowed in a more elegant and useful manner so the relationship between such crucial stats is easier to understand.

Regardless, I am not surprised there are people who put no value in it, since there are still people who ****ing care what a pitcher's W-L record is. You can take a horse to water but you can't make him drink.

Daver
07-30-2012, 03:15 PM
I agree that Pythagorean W-L is one of many useful statistical tools that together somewhat have enhanced our understanding of the infinite complexities of the game.

I just wouldn't put much stock in it alone, if divorced both from other traditional stats and new SABR metrics, as well as from traditional qualitative scouting evaluations.It's a simple game, some days you win, some days you lose, and some days it rains.

Frater Perdurabo
07-30-2012, 03:24 PM
It's a simple game, some days you win, some days you lose, and some days it rains.

That, too.

doublem23
07-30-2012, 03:30 PM
I agree that Pythagorean W-L is one of many useful statistical tools that together somewhat have enhanced our understanding of the infinite complexities of the game.

I just wouldn't put much stock in it alone, if divorced both from other traditional stats and new SABR metrics, as well as from traditional qualitative scouting evaluations.

All Pythagorean W-L does is redefine runs scored and runs allowed in a format that is both weighted against a standard scale (games played) and conceptualize it in a way that is easier to understand for someone more versed in tradition stats (like wins and losses). Does it tell you anything meaningful that Team A score 950 runs to 900 allowed versus Team B who scored 750 against 700? No. But when you reconceptualize those numbers to show you can expect Team A to be 85-77 and Team B to be 86-76, now you can begin to make more in depth analysis of each of their respective performances.

All Pythagorean W-L does is add weight and context to a pair of very important stats that otherwise lack them.

SI1020
07-30-2012, 04:28 PM
Regardless, I am not surprised there are people who put no value in it, since there are still people who ****ing care what a pitcher's W-L record is. You can take a horse to water but you can't make him drink. See this is a prime example of how these debates inevitably end up. You're either riding the new wave hip and with it, or a hopeless troglodyte. How many times have I posted that even in the Eisenhower era we knew there was more to evaluating a player than BA HR RBI W-L ERA, and all the other traditional numbers? That doesn't make Pythagorean W-L any more relevant to those of us who have actually examined it and come away less than impressed. In fact I spent more time on this than I should have today and remain that way. Old stuff doesn't always deserve to last forever and new stuff isn't always better than some of the old. Even when I was very young I never gave a **** about trends, and always wanted to find out for myself.

TDog
07-30-2012, 05:21 PM
...
FOR THE UMPTEENTH MILLION TIME, NOBODY IS SAYING THAT PYTHAGOREAN W-L IS A FOOLPROOF PREDICTOR OF FUTURE RESULTS. WHO THE MOTHER****ING ****ING **** ARE YOU ARGUING THIS POINT WITH?!!?!?!?! ...

I am not arguing that point.

I am arguing that expected wins is a useless, contrived statistic that has no value.

doublem23
07-30-2012, 05:38 PM
I am not arguing that point.

I am arguing that expected wins is a useless, contrived statistic that has no value.

Yeah, except you posted:

.
Expected wins don't work as a predictor over the course of a season. They don't approximate how well teams have actually done over the course of a season as well as actual wins and losses do.

So I don't know how you expect me to take that you aren't arguing that, because you just did. Look, I'm not denying that some people do use Pythagorean W-L to try and predict future results (and YES again, for the millionth time, while Pythagorean W-L is a poor predictor of future results, it is slightly more accurate than traditional W-L), but I'm telling you those people DO NOT UNDERSTAND what the formula is designed for. If you read any qualified baseball statistician, they will resoundingly reject the idea that Pythagorean W-L is any kind of prediction tool, especially on a game-to-game or week-to-week basis.

Again, Pythagorean W-L has value, you just have to understand what it's value is, which you clearly do not and have no interest in learning. Your arguments are the equivalent to stating a new hammer you bought sucks because it doesn't cut a piece of wood in half. If you don't know how to use a tool, then it's obviously not going to of any value to you, but don't go around ripping on hammers because the rest of us know what the **** we're doing.

doublem23
07-30-2012, 05:41 PM
See this is a prime example of how these debates inevitably end up. You're either riding the new wave hip and with it, or a hopeless troglodyte. How many times have I posted that even in the Eisenhower era we knew there was more to evaluating a player than BA HR RBI W-L ERA, and all the other traditional numbers? That doesn't make Pythagorean W-L any more relevant to those of us who have actually examined it and come away less than impressed. In fact I spent more time on this than I should have today and remain that way. Old stuff doesn't always deserve to last forever and new stuff isn't always better than some of the old. Even when I was very young I never gave a **** about trends, and always wanted to find out for myself.

A pitcher's W-L record is the most worthless stat in all of baseball, I can't state that any more bluntly. You don't have to buy into any other sabermetric analysis, you can still use traditional ERA, still use batting average, ****, still use fielding percentage for all I care, but if you still quote a pitcher's record when defending or detracting him, then we're just going to have to agree to disagree.

TDog
07-30-2012, 06:08 PM
Yeah, except you posted:



So I don't know how you expect me to take that you aren't arguing that, because you just did, and YES again, for the millionth time, while Pythagorean W-L is a poor predictor of future results, it is slightly more accurate than traditional W-L. You can deny it all you like, but you're basically arguing that the Earth is flat and the Sun revolves around us.

Again, Pythagorean W-L has value, you just have to understand what it's value is, which you clearly do not and have no interest in learning. Your arguments are the equivalent to stating a new hammer you bought sucks because it doesn't cut a piece of wood in half. If you don't know how to use a tool, then it's obviously not going to of any value to you, but don't go around ripping on hammers because the rest of us know what the **** we're doing.

If expected wins were predictive of the future, if they reflected the past, it would be a valid statistic. What it is supposed to measure, some sort of vague team quality, is meaningless because because players don't play for run differential and managers don't manage for run differential.

Managers will let other teams run up scores to save their bullpens and will take out starting pitchers after six or seven in lopsided games even if they are pitching shutouts. They don't put in their frontline pitchers when the games are not close.

In the end, no one in baseball cares about run differentials except for the people who care about expected wins and the fans who want to lament that their teams weren't that bad, although I doubt that there are many stat people on the national scene who actually believe the White Sox are 4.5 games better than the Tigers.

The battleground is for wins and losses, not for overall runs at the end of the season.

What you believe is a new, better way of looking at things ignores the realtiies of major league baseball.

Daver
07-30-2012, 06:27 PM
Again, Pythagorean W-L has value, you just have to understand what it's value is, which you clearly do not and have no interest in learning. Your arguments are the equivalent to stating a new hammer you bought sucks because it doesn't cut a piece of wood in half. If you don't know how to use a tool, then it's obviously not going to of any value to you, but don't go around ripping on hammers because the rest of us know what the **** we're doing.

First you have to believe it has value, much like you have to believe PECOTA and VORP have value, but when all is said and done it is a mathematical theory devised to deliver a specific flawed result. After that you form an opinion on what is more or less flawed.

NardiWasHere
07-30-2012, 07:04 PM
A pitcher's W-L record is the most worthless stat in all of baseball, I can't state that any more bluntly. You don't have to buy into any other sabermetric analysis, you can still use traditional ERA, still use batting average, ****, still use fielding percentage for all I care, but if you still quote a pitcher's record when defending or detracting him, then we're just going to have to agree to disagree.

I think a pitcher's record is much more useful than expected win-loss. The stat guys' disdain for the W/L Record makes no sense to me. Why can you look at expected win-loss an say, "it doesn't tell the whole story, but it still has value" and then look at pitchers win-loss and say, "it doesn't tell the whole story, it's worthless?"

If a guy wins 15 games a year, he's probably doing something right. As you said, there might be exceptions but an exception doesn't disprove a rule.

SI1020
07-30-2012, 07:17 PM
A pitcher's W-L record is the most worthless stat in all of baseball, I can't state that any more bluntly. You don't have to buy into any other sabermetric analysis, you can still use traditional ERA, still use batting average, ****, still use fielding percentage for all I care, but if you still quote a pitcher's record when defending or detracting him, then we're just going to have to agree to disagree. I don't believe I defended pitcher's W-L record, in fact I've posted about how misleading it can sometimes be. Still, I don't throw any of the traditional stats out, just try to fit them in the proper context. Cy Young won 511 games mainly because he was ****ing great. Bob Friend had a lifetime 197-230 record and he was a damn good pitcher. You need to look at the total picture and that's what I try to do.

doublem23
07-30-2012, 09:01 PM
I think a pitcher's record is much more useful than expected win-loss. The stat guys' disdain for the W/L Record makes no sense to me. Why can you look at expected win-loss an say, "it doesn't tell the whole story, but it still has value" and then look at pitchers win-loss and say, "it doesn't tell the whole story, it's worthless?"

If a guy wins 15 games a year, he's probably doing something right. As you said, there might be exceptions but an exception doesn't disprove a rule.

A pitcher's W-L record is completely dependent on 1/2 the game that they have absolutely no control over. A team ultimately controls how many runs they score or allow. Very elementary difference. It would be like giving guys credit or punishing their batting average based on what the guys ahead and behind of them did, too. That would be a ridiculous stat.

doublem23
07-30-2012, 09:06 PM
First you have to believe it has value, much like you have to believe PECOTA and VORP have value, but when all is said and done it is a mathematical theory devised to deliver a specific flawed result. After that you form an opinion on what is more or less flawed.

Of course, every stat is ultimately flawed, nothing can account for all the variance that occurs in baseball. That's what makes trying to perfect these stats so much fun, you're chasing a horizon then never gets any closer.

That said, the only people so far I have seen in this thread ripping on Pythagorean W-L have absolutely 0.00% understanding of it's purpose. So it's no surprise to see them not like it, its easy to find to find the flaws in designs when you don't get how they work.

DSpivack
07-30-2012, 09:17 PM
Of course, every stat is ultimately flawed, nothing can account for all the variance that occurs in baseball. That's what makes trying to perfect these stats so much fun, you're chasing a horizon then never gets any closer.

That said, the only people so far I have seen in this thread ripping on Pythagorean W-L have absolutely 0.00% understanding of it's purpose. So it's no surprise to see them not like it, its easy to find to find the flaws in designs when you don't get how they work.

I understand it's purpose. I just think it's silly and inaccurate to use it to mean a team "should" be better or worse than their actual record, which I think a better reflection of the quality of the team than in manipulating that record based upon how many runs they score and give up. Maybe there is some algorithm out there that can accurately reflect that, but I don't think it's as simple as Runs Scored vs. Runs Against.

doublem23
07-30-2012, 09:30 PM
I understand it's purpose. I just think it's silly and inaccurate to use it to mean a team "should" be better or worse than their actual record, which I think a better reflection of the quality of the team than in manipulating that record based upon how many runs they score and give up. Maybe there is some algorithm out there that can accurately reflect that, but I don't think it's as simple as Runs Scored vs. Runs Against.

But why would you add more data than is needed? What else do you use to get a team's W-L record other than runs scored and runs allowed? Why would you want to complicate such a perfectly simple and effective formula?

By the way, there are metrics out there that do use more data to try and predict a team's W-L record, but they're just too complicated and cumbersome to be effective. Less is sometimes more.

SOXSINCE'70
07-31-2012, 04:32 PM
It's a simple game, some days you win, some days you lose, and some days it rains.

Thank you, Nuke LaLoosh.:D: