The Pure Quality Start Thread

[calsnowskier]

I basically agree about saber being more predictive than historic.

Specifically, BABIP is not really a skill stat at the MLB level (of course Bonds would always have a higher BABIP than you or I if we were somehow able to get 600 MLB PA). With MLB hitters, it is more of a representation of luck. K pitchers are able to avoid this luck "noise" completely. This one of the reasons K pitchers TEND to be better and they move through the minors faster.

 

[tzill]

Thanks for the heads up on this thread, Cal.

Sabermetricians have long established that skills-based measures are a better gauge of pitching performance than ERA. Strikeout rate, strikeout-to-walk ratio and ground-ball rate are far better descriptors than many traditional gauges.

However, individual statistics can sometimes be too granular. One pitcher might go out and hurl a complete-game, three-hit shutout with seven strikeouts. Another might go seven innings and allow three runs but strike out 14. While the effect on ERA and peripherals will be different, both outings could be considered dominating performances.

By the same token, a pitcher could allow five runs in two innings or eight runs in four innings but both could be considered disasters.
PQS examines five measurables to determine whether a start was "dominating" or a "disaster."
Here's how it works: We score each outing on a 0-to-5 scale. The pitcher gets one point for each of the following:
•The pitcher must have gone a minimum of six innings. This measures endurance. Less than five innings is automatically a PQS of 0.
•He must have allowed no more than an equal number of hits to the number of innings pitched. This measures an ability to disrupt timing.
•His number of strikeouts must be no fewer than two less than his innings pitched. This measures an ability to miss bats.
•He must have struck out at least twice as many batters as he walked. This measures control.
•He must have allowed no more than one home run. This measures an ability to keep the ball out of the happy zones.
Outings that score a 4 or 5 are considered dominating. Outings that score a 0 or 1 are considered disasters.
If we look at the ratio of dominating performances to disasters, we can often uncover pitchers with hidden upside.

This is the Quality Start with a sabermetric DIPS twist, and it gets really easy to calculate once you get used to it. I don't think it's the end all or be all, but then nothing really is that. It is, as I like to say, another piece of the puzzle. A dominating start is scored a 4 or 5 and a disaster start is scored a 0 or 1. DOM% is the percentage of starts that are dominating, DIS% is the percentage of starts that are disasters (any start under 5.0 IP is automatically a 0, or disaster).

From my observations, a DOM at or above the 40% mark is indicative of good pitching; above 50% is great; above 70% is elite. A low DIS is also indicative of good pitching. Basically, you want to see a pitcher's DOM% to be over 40% and ideally over 50%, and you want their DIS to be under 20% and ideally under 10%. If you had to chose a high DOM% or a low DIS%, pitchers tend to have a lower ERA when you have a low DIS% vs. a high DOM% (obviously if you combine both, you have a much better chance of having an elite pitcher). But I think when the DOM% is high enough, you win more by choosing a high DOM% over a low DIS%, as there are more high quality games pitched overall.

No single metric is perfect, but this one is fairly useful. It's easy to compute and I keep records of the Giants PQS scores, and a pitcher's average PQS score for the year. Once we get into June (about 10 starts) it becomes a fairly useful HISTORICAL tool, and to a certain extent, predictive as well.

Hope that helps.

 

[MilkSpiller22]

i see where i was confused about the consistency... I was thrown off by the numbers, i thought the ratio would be score/(5* GS), but i see now that 4 and 5= yes 0,1=no and you do %yes and %no... or disaster and dominant...

I have one question with the strikeouts... What if you strike out only one batter but don't walk anyone, does that count or not... if not, then it is kind of silly that you can pitch a perfect game and only get a 3...

I still don't love the stat, but i think i am understanding it now... I just don't think i will ever like a stat like this since i don't agree with the weights and measures of this... that is also what i don't like about many sabremetric stats...

 

[calsnowskier]

i see where i was confused about the consistency... I was thrown off by the numbers, i thought the ratio would be score/(5* GS), but i see now that 4 and 5= yes 0,1=no and you do %yes and %no... or disaster and dominant...

I have one question with the strikeouts... What if you strike out only one batter but don't walk anyone, does that count or not... if not, then it is kind of silly that you can pitch a perfect game and only get a 3...

I still don't love the stat, but i think i am understanding it now... I just don't think i will ever like a stat like this since i don't agree with the weights and measures of this... that is also what i don't like about many sabremetric stats...

The perfect game you describe would be a 4. No "divide by zero errors" here.

 

[tzill]

i see where i was confused about the consistency... I was thrown off by the numbers, i thought the ratio would be score/(5* GS), but i see now that 4 and 5= yes 0,1=no and you do %yes and %no... or disaster and dominant...

I have one question with the strikeouts... What if you strike out only one batter but don't walk anyone, does that count or not... if not, then it is kind of silly that you can pitch a perfect game and only get a 3...

I still don't love the stat, but i think i am understanding it now... I just don't think i will ever like a stat like this since i don't agree with the weights and measures of this... that is also what i don't like about many sabremetric stats...

I think you've hypothosized a game that has never happened. If it has happened, it's happened so rarely that it's not worth adjusting for.

If you don't care for the stat, so be it. It's pretty useful.

 

[tzill]

The perfect game you describe would be a 4. No "divide by zero errors" here.

Actually that's a good point. He'd have to K at LEAST twice as many as he BBs. In a perfect game with one K (which I doubt has ever happened), he'd get a point for 1 being greater than 2x0 = 0.

So, it'd be a 4, a dominant start.

Hypothetically.....:rollseyes:

 

[MilkSpiller22]

I think you've hypothosized a game that has never happened. If it has happened, it's happened so rarely that it's not worth adjusting for.

If you don't care for the stat, so be it. It's pretty useful.

What are you talking about?? Maybe the perfect game is rare. But it doesn't have to be a perfect game but all perfect games are rare!! How about 6 innings 0 walks 1 strikeout?? It happens a lot that there is a 1K 0BB.

 

[calsnowskier]

What are you talking about?? Maybe the perfect game is rare. But it doesn't have to be a perfect game but all perfect games are rare!! How about 6 innings 0 walks 1 strikeout?? It happens a lot that there is a 1K 0BB.

What about it? Are saying a guy goes only 6 perfect then leaves the game? That is is a 4.

 

[MilkSpiller22]

What about it? Are saying a guy goes only 6 perfect then leaves the game? That is is a 4.

Was just commenting on tzill saying that the perfect game with 1K, 0BB was so rare that it is not worth adjusting for...

 

[calsnowskier]

Ok

Not getting your point...

 

[tzill]

What are you talking about?? Maybe the perfect game is rare. But it doesn't have to be a perfect game but all perfect games are rare!! How about 6 innings 0 walks 1 strikeout?? It happens a lot that there is a 1K 0BB.

As mentioned upthread, that start would get a PQS of 4. Not sure what you're driving at here...

 

[Darkstone42]

This would be a great thing to use to build a few plots to illustrate, visually, how good/dominant/consistent a pitcher is and how he got there.

I'm imagining distribution plots for frequency of score occurrences, as well as a percentage of each point earned over starts. You could flesh out something like pitcher A overall had a league average ERA, but he got there with an inverted distribution, lots of 0-1 performances, lots of 4-5 performances, not much in between. So his team had a change to win only roughly half his starts, but in that half they had a chance to win, he basically guaranteed they would. He has value as a 3-4 type in a rotation, therefore.

You could further delve into why Pitcher A was like that, maybe noticed he struck out a lot of batters and didn't give up many hits, but he could concede runs in bunches because of HR and BB troubles.

Pitcher B, conversely, pretty much always scored 2-3. The team nearly always had a change to win when he pitched, but it wasn't a guarantee, either. It turns out he limited walks and home runs, but he also gave up a lot of hits and didn't miss many bats. Again, you're probably looking at a 3-4 type, but for different reasons than Pitcher A.

So how do you decide between these pitchers? Well, a team with a terrible defense might prefer Pitcher A because he basically takes the defense out of the equation for both good and bad, whereas Pitcher B would fit in great on a team with a good defense since guys put the ball in play against him, but he doesn't beat himself with walks and home runs.

So we could elucidate both style and consistency with these distribution plots and get a better idea about the best environments in which to put pitchers so they succeed.

It's a nice, granular view at what QS attempts to do in determining pitcher consistency.

 

[calsnowskier]

This would be a great thing to use to build a few plots to illustrate, visually, how good/dominant/consistent a pitcher is and how he got there.

I'm imagining distribution plots for frequency of score occurrences, as well as a percentage of each point earned over starts. You could flesh out something like pitcher A overall had a league average ERA, but he got there with an inverted distribution, lots of 0-1 performances, lots of 4-5 performances, not much in between. So his team had a change to win only roughly half his starts, but in that half they had a chance to win, he basically guaranteed they would. He has value as a 3-4 type in a rotation, therefore.

You could further delve into why Pitcher A was like that, maybe noticed he struck out a lot of batters and didn't give up many hits, but he could concede runs in bunches because of HR and BB troubles.

Pitcher B, conversely, pretty much always scored 2-3. The team nearly always had a change to win when he pitched, but it wasn't a guarantee, either. It turns out he limited walks and home runs, but he also gave up a lot of hits and didn't miss many bats. Again, you're probably looking at a 3-4 type, but for different reasons than Pitcher A.

So how do you decide between these pitchers? Well, a team with a terrible defense might prefer Pitcher A because he basically takes the defense out of the equation for both good and bad, whereas Pitcher B would fit in great on a team with a good defense since guys put the ball in play against him, but he doesn't beat himself with walks and home runs.

So we could elucidate both style and consistency with these distribution plots and get a better idea about the best environments in which to put pitchers so they succeed.

It's a nice, granular view at what QS attempts to do in determining pitcher consistency.

This is taking it further than I have ever thought of taking the stat, but the potential is certainly there for any number of analysis'.

As for your scenario, if you have player A, than you would tend to have a much quicker hook than you would if you had player B. Player A either has it or he doesn't while player B tends to be much more even-keel.

 

[Darkstone42]

This is taking it further than I have ever thought of taking the stat, but the potential is certainly there for any number of analysis'.

As for your scenario, if you have player A, than you would tend to have a much quicker hook than you would if you had player B. Player A either has it or he doesn't while player B tends to be much more even-keel.

I thought about that, too, but that's probably a bigger thought in the post-season. Maybe you have both guys, but you're hoping A will get hot for the playoffs or have a dominating start while your ace is resting in a must-win game. It's a worthwhile risk to take on him being on, and you could have B on standby and a short hook for A if he doesn't have it.

 

[calsnowskier]

I thought about that, too, but that's probably a bigger thought in the post-season. Maybe you have both guys, but you're hoping A will get hot for the playoffs or have a dominating start while your ace is resting in a must-win game. It's a worthwhile risk to take on him being on, and you could have B on standby and a short hook for A if he doesn't have it.

Player A is Tim Lincecum right now, or Barry Zito the last couple years and to a lessor extent, Jonathan Sanchez in '10 and '11 (to point to specific examples on the Giants the last few years). Player B is more so like Ryan Vogelsong.

 

[MilkSpiller22]

I don't see the similarity between QS and PQS though, the only thing they really have in common is that they are a per game statistic, and both measure consistency in their own way...

I still don't see the full story that this statistic is showing, but i am starting to see some value in it...

 

[Darkstone42]

I don't see the similarity between QS and PQS though, the only thing they really have in common is that they are a per game statistic, and both measure consistency in their own way...

I still don't see the full story that this statistic is showing, but i am starting to see some value in it...

It targets the same thing (how often a pitcher has a good game), while also providing information on stinkers (which QS doesn't do), but does so from the perspective largely of things pitchers have more control over (contact management in the hits and home runs, walks, and strikeouts). Now, I agree that sequencing is a skill (or the result of many skills) for a pitcher, but I think most of the factors which lead to it are reflected nicely in the other information.

And more often than not, if a pitcher has a PQS, he'll also have a QS. Sure, there are exceptions, but those will be averaged out in the bulk. Conversely, it's less likely that a QS will also be a PQS (still very likely, of course, but slightly less so than the opposite), so PQS is likely a higher standard than QS.

Neither obviously tells us everything about a pitcher, but PQS is a lot more granular, so while we can probably extract about the same overall information, we can extract more details, as well. And that's important.

Here's another way to view it. QS tells us how often a pitcher gets the overall result he wants. PQS tells us how often the pitcher does the things which most frequently lead to the results he wants. In general, the latter is more telling, I think, in an evaluative sense.

 

[Darkstone42]

An addendum: I see a place for QS, but I think it needs to be stricter. 6+ IP, 3.5 or less ERA would be my measure. Right now, the bare minimum of 6 IP, 3 or fewer RA, well, a pitcher who has exactly that start every start over the course of a season will have 34 QS and still not be a very good pitcher overall. (That's a 4.50 ERA, and how many of those guys stick in premium spots in the rotation?)

 

[MilkSpiller22]

It targets the same thing (how often a pitcher has a good game), while also providing information on stinkers (which QS doesn't do), but does so from the perspective largely of things pitchers have more control over (contact management in the hits and home runs, walks, and strikeouts). Now, I agree that sequencing is a skill (or the result of many skills) for a pitcher, but I think most of the factors which lead to it are reflected nicely in the other information.

And more often than not, if a pitcher has a PQS, he'll also have a QS. Sure, there are exceptions, but those will be averaged out in the bulk. Conversely, it's less likely that a QS will also be a PQS (still very likely, of course, but slightly less so than the opposite), so PQS is likely a higher standard than QS.

Neither obviously tells us everything about a pitcher, but PQS is a lot more granular, so while we can probably extract about the same overall information, we can extract more details, as well. And that's important.

Here's another way to view it. QS tells us how often a pitcher gets the overall result he wants. PQS tells us how often the pitcher does the things which most frequently lead to the results he wants. In general, the latter is more telling, I think, in an evaluative sense.

But QS is more than just a good game, it predicts whether a pitcher should get the win or not... PQS has no affect on a team winning or losing, it just shows an overall outing discounting outlier innings...

 

[calsnowskier]

If a study were done, I am 95% positive that teams have won a higher % of PQS 4 starts than QS starts. A PQS of 2 or 3 is probably just slightly better than a generic QS.

Supposition, of course, since to my knowledge, there are no large scale studies if PQS.