This article is so 1991. You know what killed the complete game? Billy Martin's handling of the 1980-81 A's killed the complete game.

The article is full of overgeneralizations and strawman arguments, as well as complete ignorance of medical and statistical studies that have shown that pitching a lot of innings is a bad idea for pitchers for the most part.

I do have to agree with his assessment of a quality start. It is not really indicative of anything other than consistency - certainly not value. Halladay only has 5 QS to McGowan's 7 and Burnett's 6, yet he has a substantially better ERA than both of them and is leading the majors in innings. I think pitch counts are useful, but they aren't the be-all and end-all.

Maybe QS does tell you how many times you gave your team a chance to win.

Does anybody really pay attention to QS anymore? I can't recall ever looking this stat up on anybody. Or holds.

I think QS is useful in that it begins to separate, for the less rabid fan, the pitcher's win total from the pitcher's actual performance. Many, many fans closely correlate a pitcher's win total with how good the pitcher is. QS starts the separation and allows for better insight into the actual performance.

QS is definitely rough but it can provide a glimpse into a pitcher's run distribution. Assuming that we are following that a QS = IP>=6 and R=<3 and given these two hypothetical pitchers...

Pitcher A

Start 1: 6 IP, 3 ER
Start 2: 6 IP, 3 ER
Start 3: 9 IP, 0 ER
Start 4: 3 IP, 6 ER

Pitcher B

Start 1: 9 IP, 0 ER
Start 2: 9 IP, 0 ER
Start 3: 3 IP, 6 ER
Start 4: 3 IP, 6 ER

They'll have the same ERA but Pitcher A gave his team more "chances to win" since he had three quality starts as opposed to Pitcher B's two. Of course, it helps in this hypothetical to set the pitchers' ERA's equal, but it's an example of a rough use of the QS statistic. There are quite a few other metrics I'd use before QS, though.

I don't understand why they don't just say a QS is either two runs in six innings, or three runs in seven or more innings. That would take away the easy "but you could have a 4.50 ERA!" criticism, and get people to concentrate on the stat itself, which is a decent one, as far as it goes.

That said, as Rany recently wrote better than I could, the stat has the obvious problem that it treats the game where you give up two or three runs as equivalent to a complete-game shutout, and Support-Neutral W-L is the better version of it.

When I was a kid, Dave Dravecky pitched until his arm fell off. Then he came back and pitched the second game with his other arm.

They'll have the same ERA but Pitcher A gave his team more "chances to win" since he had three quality starts as opposed to Pitcher B's two. Of course, it helps in this hypothetical to set the pitchers' ERA's equal, but it's an example of a rough use of the QS statistic. There are quite a few other metrics I'd use before QS, though.

I've never understood this argument. Let's assume a 4.5 league ERA, and that both teams score 4.5 runs a game.
A
Start 1 - 4.5 vs 4.5 - T
Start 2 - 4.5 vs 4.5 - T
Start 3 - 0 vs 4.5 - W
Start 4 - 9.0 vs 4.5 - L

B
Start 1 - 0 vs 4.5 - W
Start 2 - 0 vs 4.5 - W
Start 3 - 9.0 vs 4.5 L
Start 4 - 9.0 vs 4.5 L

So one finishes 2-0-2 and the other 1-2-1. No great difference. It's just as likely pitcher A's team loses the two games as wins them, and, assuming a static 4.50 ERA, you can posit just as many situations whereby Pitcher B is better as you can where Pitcher A is better, at least in terms of wins. As the run environment gets higher, it becomes more likely that Pitcher A gets wins in 1 and 2, but so it becomes more likely that pitcher B wins.

I've never seen any analysis to suggest the the cumulative WPA is better for pitcher A than pitcher B, but I'd be interested if it existed. If I had to guess one way, I'd guess that Pitcher B was slightly more probable to win more games over a season of x length simply because a start like 3 or 4 implies offence-friendly conditions, which in turn implies his team may score more than league-average runs.

Most importantly, it's very unlikely that pitcher B finishes under .500. Not that this means anything, but if we're talking about 'chances to win', it's worth saying that you be'd happy if a 4.50 ERA pitcher walked away with 18 wins, right?

That last comment was meant to be something of a joke.

That would take away the easy "but you could have a 4.50 ERA!" criticism

I don't know why that's a criticism. That's going to be reasonably close to league average, which means it gives a league average offense an average chance to win the game.

When I was a kid, Dave Dravecky pitched until his arm fell off.

when I was a kid, we pitched uphill in both directions for 9 innings

(and we liked it)

Does anybody really pay attention to QS anymore?

Fantasy players -- everyone knows fantasy players have rended the fabric of the game.

you be'd happy if a 4.50 ERA pitcher walked away with 18 wins, right

Of course I would, if he's on my team.

Leaders in QS, since 1956:

Don Sutton 483
Nolan Ryan 481
Greg Maddux 468
Roger Clemens 465
Tom Seaver 454
Gaylord Perry 453
Steve Carlton 447
Phil Niekro 442
Tom Glavine 435
Tommy John 431
Bert Blyleven 429

Leaders in CG over the same period:

Gaylord Perry 303
Fergie Jenkins 267
Bob Gibson 255
Steve Carlton 254
Phil Niekro 245
Juan Marichal 244
Bert Blyleven 242
Tom Seaver 231
Nolan Ryan 222
Jim Palmer 211

you be'd happy if a 4.50 ERA pitcher walked away with 18 wins, right

but QS pitchers DON'T have an ERA of 4.50

I don't know if the stat's been updated, but the last time someone did an analysis (I believe it was from 1984 to 92 or something), picher's ERA in QS's was about 1.90

I'll certainly take that

Neyer wrote an article back in 2006 that looked at pitchers' performance in QS and it was much better than 4.5 iirc. Unfortunately it's behind the insider firewall now. Perhaps Rob would be willing to reiterate his findings for us.

but QS pitchers DON'T have an ERA of 4.50

I don't know if the stat's been updated, but the last time someone did an analysis (I believe it was from 1984 to 92 or something), picher's ERA in QS's was about 1.90

I'll certainly take that

Of course. But the argument is that should a 6 innings start with an ERA of 4.5 be included in the mix. I'm sure that if the definition of QS went down to 5 innings and 5 ER, the cumulative ERA of all QS would still be very good, but no one would argue that giving up a run an inning is anything close to a good start.

I've always felt it should be made into minimum 6 IP, allowing no more than a run per 2 innings pitched. That way a Halladay who goes 9 and gives up 4 runs gets a quality start, while a guy going 6 giving up 4 does not.

Of course, in the end there are far better ways to measure this stuff and fine tuning a minor stat is hardly going to help or hurt anything.

I did a quick bb-ref search and found that in this year's quality starts, the total ERA is 2.03.

There have also been three relief appearances that would have otherwise qualified as QS's. Those relievers gave up 3 ER in 18 innings.

I've always felt it should be made into minimum 6 IP, allowing no more than a run per 2 innings pitched. That way a Halladay who goes 9 and gives up 4 runs gets a quality start, while a guy going 6 giving up 4 does not.


Wait, 6 IP and 3 ER is a QS, but 8+ IP and 4 ER is not? That makes no sense.

There have been seven non-quality starts this year in which the starter had an ERA of 4.50 or under and pitched more than six innings. Halladay has two of them.

Teams are 512-247-2 this year when their starter has a QS. Pitchers are 406-129 with 226 no-decisions.

Wait, 6 IP and 3 ER is a QS, but 8+ IP and 4 ER is not? That makes no sense.

The idea is to have a quick-and-dirty stat. Of course Support Neutral Adjusted Wins for Handedeness and Groundball Tendencies of Umpires is more accurate, but who has the time to calculate that?

That said, I would have preferred the stat to really count quality STARTS--that is, once you get the 18th out wihout allowing 4 runs, you have the QS. Period. No chance to lose it in the 7th inning by allowing the 4th run. If the manager prefers the starter to the reliever there, that's not the starter's fault.

I did a quick bb-ref search and found that in this year's quality starts, the total ERA is 2.03.

How did you do that? Using P-I, I can see all 764 QS's this year so far (that's over 25 per team by the way, almost half of the total games played.), but I can't figure out how to total the columns.

How did you do that? Using P-I, I can see all 764 QS's this year so far (that's over 25 per team by the way, over half of the total games played.), but I can't figure out how to total the columns.

I sorted by IP and then ER and counted the number of games with each total. Then I just multiplied and added. Three of the 764 were relief appearances.

There have only been 2 games this year in which both starters threw QS? Am I reading 22 correctly?

There have only been 2 games this year in which both starters threw QS? Am I reading 22 correctly?

No, both starters in the suspended White Sox-Orioles game from a month ago threw QS's, so neither team won or lost. The W-L total includes games where each starter threw a QS, but those are counted twice (one win and one loss) in the total.

Another relevant stat for QS would be the team ERA when a starter has a QS. If you throw 6 or more innings with 3 or fewer runs, the game is going to be close or a blowout in your favor, so usually the relievers are going to be your top guys, who will probably deliver a pretty good ERA.

Has anyone ever done something with WE that measures pitchers by "giving the team a chance to win?" Not by summing up the WE of every event, but simply looking at the basic stats and computing the team's chance to win based on that...or would it be the same thing?

E.G. Starter goes 7 innings and allows 5 runs. Teams that allow 5 runs in 7 innings win the game 22% of the time, so the starter is awarded 22 "points."

I see this as being a decent way to measure pitchers based on "giving his team a chance to win" which would be more specific than QS but wouldn't depend on his offense/bullpen.

Is there anything like this out there?

Is that what Support-neutral W-L measures?

No, both starters in the suspended White Sox-Orioles game from a month ago threw QS's, so neither team won or lost. The W-L total includes games where each starter threw a QS, but those are counted twice (one win and one loss) in the total.

Ah, got it. Thanks.

Is that what Support-neutral W-L measures?

Pretty much. IIRC the method is, it adds the percent chance a pitcher would historically get the win leaving the game in the state he did to a SNW number and the percent chance a pitcher would historically take a loss in that state and add it to a SNL number.

Personally, I think SNWL is the bee's knees. It smooths out horribly bad outings.

So it's basically how I described, except it also includes the base-out state that the pitcher left with?

So, "% chance of team winning after allowin 3 runs in 5 innings with the bases loaded and nobody out in the sixth." If so, I agree that's pretty cool. And damn them for stealing my idea before I had it.

I'm curious as to when the idea of the quality start was invented. It seems to me to be a product of an era of increased offense. I doubt anybody would have called 6IP/3ER a quality start in 1968. I just think the metric might need to be adjusted for era, and we seem to be entering an era with a lower run scoring environment..

Scoring was higher in '06 and '07 than it was in '02 and '03, for example. This is all just fluctuation, with the same basic offensive level, which is, however, definitely somewhat lower than the mid and late '90s, thank goodness.

Personally, I think SNWL is the bee's knees.

Concur. Whatever happened to Wolverton anyways?

I'm curious as to when the idea of the quality start was invented. It seems to me to be a product of an era of increased offense. I doubt anybody would have called 6IP/3ER a quality start in 1968. I just think the metric might need to be adjusted for era, and we seem to be entering an era with a lower run scoring environment..

It was born, lived, and pretty much died in the 1980's, which was a much lower scoring environment than today.

If you look at the leaders in quality starts every year, they are exactly who you'd think they would be - the best pitchers in the league by every other metric. The best criticism of quality starts is that they usually don't tell you anything you wouldn't otherwise know.

It was born, lived, and pretty much died in the 1980's, which was a much lower scoring environment than today.

Well, I thank you for this. And QS definitely does track with the best starters, which makes sense.

I'm curious as to when the idea of the quality start was invented.

According to the Wikipedia article on "quality start," it was developed in 1985. The article also links to the study mentioned by Pasta-diving Jeter (jmac66) in post 16.

I like Quality Starts, but I like almost any stat that is based upon real world results. I agree that it should be a stat that automatically gets counted once you reach that level. I also love Holds as a stat too, it's on par with saves as a stat and isn't overrated to the degree that saves is.

I usually push for modification of quality starts based upon era or at least use other definitions for degrees of quality (such as 3 runs over 7 innings is deep quality or something like that)