Joe Posnanski writes about sports for a living, particularly baseball. Here, he writes about sports and also Springsteen, Hamilton, Harry Potter, iPads, infomercials, his idolization of Duane Kuiper, his family and magic. He lives in Charlotte with his wife Margo, two daughters Elizabeth and Katie, and their dog Westley. Joe is currently working on a book about Harry Houdini in today's world.

Wonder how it stacks up?
Apr 3 – Doolittle .258 (Pomeranz win)
Apr 9 – Chavez .262 (Otero with the blown save and win)
Apr 14 – Gregerson .185 (save)
Apr 20 – Chavez .238 (Chavez 1-0)
Apr 25 – Gregerson .104 (Gregerson win)
Apr 30 – Chavez .119 (Chavez 2-0)
May 6 – Chavez -0.195 (Chavez 2-1)
Johnson had a -0.147 WPA
May 12 Doolittle .295 (Chavez 3-1)
May 18 Chavez .089 (Chavez 4-1)
So by Tango, Chavez was 7-0, and both WPA and win he is 4-1 but in different games for 3 of the 4 wins.

I was about to say we should just use the Schmutter rule: In a win, give the pitcher with the best WPA the win. In a loss, give the pitcher with the worst WPA the loss. Period.

WPA for starters will depend on the game state – a great game where your team scores 4 in the first inning and then nothing else is worth much less by WPA than the same game where the runs came in the ninth (I’m assuming a starter throwing a CG SO).

Wins give the starter extra credit for their offense scoring runs. WPA for starters gives them extra credit for the offense not scoring runs. In both cases, those are nuisance factors when evaluating a start, no features. Game scores attempt to avoid both

Excellent work JOe (and Tom). Might I point out one error in your stats at the beginning of your article Joe. If you go back to 1900 I am pretty sure that teams who gave up 0 runs do not have a 1.000 winning percentage. They do have a 1.000 non-losing percentage, but those two things are not precisely the same.

Anticipating a post about how Joe counted Rapper’s Delight’s ha’s. (And probably the relative proportion of nonsense in songs with nonsense syllables).

I’m sure I’m not the first person to suggest this, but why not create a simulated, offense independent, park adjusted W-L based on the number of runs allowed and innings pitched. If a pitcher pitches a complete game one run game in average park conditions, that pitcher would get .889 wins and .111 losses (based on the table on Joe’s article). Lets say that’s one run over 6 innings, or 1.5 runs per 9 innings, which say for convenience’s sake, would yield a winning pct. of .800. The pitcher would get .53 wins (80 percent of 6/9, and .13 losses (20 percent of 6/9). You could adjust for park factors by dividing the number of runs allowed by the park factor. The pitcher would have exactly as many decisions as imnings pitched divided by 9. It’s fairly complicated to compute, but the end result would be familiar, a W-L record.

I use annual ERA+ and earned runs allowed to get at this by plugging into negative binomial distribution — which simulates the behavior of run scoring pretty well.

This was done in (I believe) Baseball Between the Numbers. They have a pretty good chapter on what’s wrong with ERA and W as pitcher evaluations. I don’t think they included the loss component—they just credited the pitcher in your example with .889 of a win for that game, although it’s of course been a while since I’ve thumbed through that book.

It will never make sense to me how a man alive in this century can care this much about the pitcher win and so little about what super heroes to draft in his fantasy league

Okay, it has to be conceded the win is a lousy stat: It doesn’t measure what it purports to measure; it gives credit (and blame) to the wrong people, sometimes comically so; and it creates the mistaken impression that pitchers do something that everyone knows they don’t (and can’t) do. And it’s especially terrible as a metric for determining the Cy Young or other awards, as run support and bullpen help are so crucial for determining single-season W totals. Whether the stat doesn’t tell you anything, or (which I think is more correct) it tells you far too many things all at once, it’s obviously a limited stat that can be replaced by a couple dozen other measures.

But its failings are of a sort that disguise the trees behind the forest, or if you prefer, it’s a stat for people who like Impressionist paintings. Given a broad enough timeline—and here I’m thinking of the length of a typical long career—it’s actually a medium adequate stat. If you look at the lifetime records of an assortment of unnamed pitchers, you can actually get a pretty good idea of who the guys are. Just compare the guy who’s 324-292 to the guy who’s 251-174—you know exactly who’s better, and why. Or a guy who goes 101-60—you can’t tell exactly what story is behind that record, but you can narrow it down to maybe three stories; either he pitched well but got hurt young and hung ’em up, or maybe he moved to the ‘pen and played mostly for pretty good teams. A guy who’s 224-175, you not only know his career—great pitcher who managed a fine career but played for his share of mediocre teams—but you also know he’s a lefty.

Again, within a single season it’s a lousy stat, and within a single game it’s absurd. But if all you get to sum up a career on plaque is W-L and ERA… hey, it’s not as revelatory as WAR and ERA+. But it’s good enough to talk about what’s after all just a game.

Actually, the “Win” isn’t a bad overall stat for starting pitchers. When you use statistical methodology (actual mathematical analysis, and not just creating a formula based on “stats” that gives you a result that you like), you will find that the Win correlates very strongly with other measures of a pitcher’s effectiveness like WHIP and ERA+

I admit my quick and dirty analysis could have problems (I probably should have gone with Won-Loss percentage rather than just Wins, for example). If someone better than me at statistical analysis would like to have a go at it, I’d welcome it.

I just love that we may end up using “Tango” as a statistic, for example: “Who got the Tango last night?” or “The pitcher with the best Tango record won the Cy Young once again”

I wonder if fantasy leagues recognize it as a statistic. I am emailing my league to switch to the Tango for next season.

Must be going around again… was reading this earlier: http://www.beyondtheboxscore.com/2014/5/19/5723968/chicago-white-sox-chris-sale-wpa-major-league-baseball-pitcher-effectiveness

Wonder how it stacks up?

Apr 3 – Doolittle .258 (Pomeranz win)

Apr 9 – Chavez .262 (Otero with the blown save and win)

Apr 14 – Gregerson .185 (save)

Apr 20 – Chavez .238 (Chavez 1-0)

Apr 25 – Gregerson .104 (Gregerson win)

Apr 30 – Chavez .119 (Chavez 2-0)

May 6 – Chavez -0.195 (Chavez 2-1)

Johnson had a -0.147 WPA

May 12 Doolittle .295 (Chavez 3-1)

May 18 Chavez .089 (Chavez 4-1)

So by Tango, Chavez was 7-0, and both WPA and win he is 4-1 but in different games for 3 of the 4 wins.

The article I link provided wpa for starters over their careers. So was just curious and joe provided the framework. I knew the limits, but curious.

I think I’m ok with the pitcher win provided we use the context. I know it is not saying much about pitcher value and treat it as such.

WPA is not all that useful for evaluating starting pitchers. It’s better than W-L record, but it has a lot of flaws.

Here’s Tom Tango at his best, explaining why:

http://tangotiger.com/index.php/site/article/blast-from-the-past-dont-use-wpa-for-starting-pitchers

I was about to say we should just use the Schmutter rule: In a win, give the pitcher with the best WPA the win. In a loss, give the pitcher with the worst WPA the loss. Period.

Why Tom Tango refuses to reveal his true identity? Does he works for the CIA on the side or something?

He’s really Joe Biden. He has a carefully cultivated buffoonish reputation to uphold.

WPA for starters will depend on the game state – a great game where your team scores 4 in the first inning and then nothing else is worth much less by WPA than the same game where the runs came in the ninth (I’m assuming a starter throwing a CG SO).

Wins give the starter extra credit for their offense scoring runs. WPA for starters gives them extra credit for the offense not scoring runs. In both cases, those are nuisance factors when evaluating a start, no features. Game scores attempt to avoid both

Excellent work JOe (and Tom). Might I point out one error in your stats at the beginning of your article Joe. If you go back to 1900 I am pretty sure that teams who gave up 0 runs do not have a 1.000 winning percentage. They do have a 1.000 non-losing percentage, but those two things are not precisely the same.

Anticipating a post about how Joe counted Rapper’s Delight’s ha’s. (And probably the relative proportion of nonsense in songs with nonsense syllables).

I’m sure I’m not the first person to suggest this, but why not create a simulated, offense independent, park adjusted W-L based on the number of runs allowed and innings pitched. If a pitcher pitches a complete game one run game in average park conditions, that pitcher would get .889 wins and .111 losses (based on the table on Joe’s article). Lets say that’s one run over 6 innings, or 1.5 runs per 9 innings, which say for convenience’s sake, would yield a winning pct. of .800. The pitcher would get .53 wins (80 percent of 6/9, and .13 losses (20 percent of 6/9). You could adjust for park factors by dividing the number of runs allowed by the park factor. The pitcher would have exactly as many decisions as imnings pitched divided by 9. It’s fairly complicated to compute, but the end result would be familiar, a W-L record.

I use annual ERA+ and earned runs allowed to get at this by plugging into negative binomial distribution — which simulates the behavior of run scoring pretty well.

This was done in (I believe)

Baseball Between the Numbers. They have a pretty good chapter on what’s wrong with ERA and W as pitcher evaluations. I don’t think they included the loss component—they just credited the pitcher in your example with .889 of a win for that game, although it’s of course been a while since I’ve thumbed through that book.It will never make sense to me how a man alive in this century can care this much about the pitcher win and so little about what super heroes to draft in his fantasy league

Okay, it has to be conceded the win is a lousy stat: It doesn’t measure what it purports to measure; it gives credit (and blame) to the wrong people, sometimes comically so; and it creates the mistaken impression that pitchers do something that everyone knows they don’t (and can’t) do. And it’s especially terrible as a metric for determining the Cy Young or other awards, as run support and bullpen help are so crucial for determining single-season W totals. Whether the stat doesn’t tell you anything, or (which I think is more correct) it tells you far too many things all at once, it’s obviously a limited stat that can be replaced by a couple dozen other measures.

But its failings are of a sort that disguise the trees behind the forest, or if you prefer, it’s a stat for people who like Impressionist paintings. Given a broad enough timeline—and here I’m thinking of the length of a typical long career—it’s actually a medium adequate stat. If you look at the lifetime records of an assortment of unnamed pitchers, you can actually get a pretty good idea of who the guys are. Just compare the guy who’s 324-292 to the guy who’s 251-174—you know exactly who’s better, and why. Or a guy who goes 101-60—you can’t tell exactly what story is behind that record, but you can narrow it down to maybe three stories; either he pitched well but got hurt young and hung ’em up, or maybe he moved to the ‘pen and played mostly for pretty good teams. A guy who’s 224-175, you not only know his career—great pitcher who managed a fine career but played for his share of mediocre teams—but you also know he’s a lefty.

Again, within a single season it’s a lousy stat, and within a single game it’s absurd. But if all you get to sum up a career on plaque is W-L and ERA… hey, it’s not as revelatory as WAR and ERA+. But it’s good enough to talk about what’s after all just a game.

Actually, the “Win” isn’t a bad overall stat for starting pitchers. When you use statistical methodology (actual mathematical analysis, and not just creating a formula based on “stats” that gives you a result that you like), you will find that the Win correlates very strongly with other measures of a pitcher’s effectiveness like WHIP and ERA+

http://pureblather.com/2014/05/20/on-pitchers-wins/

I admit my quick and dirty analysis could have problems (I probably should have gone with Won-Loss percentage rather than just Wins, for example). If someone better than me at statistical analysis would like to have a go at it, I’d welcome it.

I just love that we may end up using “Tango” as a statistic, for example: “Who got the Tango last night?” or “The pitcher with the best Tango record won the Cy Young once again”

I wonder if fantasy leagues recognize it as a statistic. I am emailing my league to switch to the Tango for next season.