Flashes of brilliance: Introducing True Quality Starts

Today, I’d like to introduce a new statistic that I’ll be using as a supplement in pitcher evaluations. I call it True Quality Starts, which looks to improve upon the already existing quality start statistic.

The concept behind the original “quality start” stat is pretty sound. It is meant to give a glimpse at how consistently good a pitcher is. A “quality start,” however, is determined by pretty arbitrary means. A pitcher needs to allow no more than three earned runs and throw at least six innings. But what if the pitcher throws nine innings and gives up four earned runs? His ERA would be lower, yet he wouldn’t be credited with a quality start. And that’s completely ignoring the fact that pitchers don’t have a ton of control over how many runs they allow.


Before I delve too deeply into True Quality Starts, I’d like to thank Fan Graphs, which displays game logs (including batted-ball information) for every pitcher; a fantastic resource, to say the least.

I think the best way to describe what I have in mind for True Quality Starts would be to start with how I actually developed the stats. If you’re not into the specific calculations, skip down a paragraph, and start reading from there.

That being said, here is the methodology behind True Quality Starts. I first normalize the batted-ball figures to account for disproportionate line drives. After that, I apply relative run values to strikeouts, walks, adjusted ground balls, adjusted fly balls, and adjusted line drives. I take this figure, multiply by nine, and divide by innings pitched to get a sort of makeshift, one-game LIPS ERA. I then apply a simple above replacement level measure, calculated as (6 – LIPS ERA)*IP/9. I call this the TQS score.

After I do all this, I take every TQS score for the entire year and find the standard deviation among them. These standard deviations are then used to classify each start as one of the following: great, good, above average, below Average, bad, or awful.

TQS classifications

Above average starts are those within one positive standard deviation of the mean. Below average starts are those within one negative standard deviation of the mean. Good starts are those between one and two positive standard deviations of the mean. Bad starts are those between one and two negative standard deviations of the mean. Great starts are those above two positive standards deviations of the mean. Awful starts are those below two negative standards deviations of the mean.

From there, we tally them up and get a complete picture of a pitcher’s start history for a given year. In the tables you’ll see, I’ve also added a few additional categories:

  • Good and Great Percentage (GG%) measures how often the pitcher had either a good or great start.
  • Bad and Awful Percentage (BA%) measures how often the pitcher had either a bad or awful start.
  • Average Percentage (AVG%) measures how often the pitcher had an above or below average start.
  • True Quality Start Percentage (TQS%) measures how often the pitcher had an above average, good, or great start.
  • Good & Great to Bad & Awful ratio (GG/BA) measures the two against each other.


Now that you know how I calculate True Quality Starts, let’s talk about why I think they can be meaningful. The title of this article is “Flashes of brilliance.” If you’ve ever heard this phrase before, you know that it is used to describe players who have shown some real talent, but who have been unable to display it on a consistent basis. These pitchers have some serious breakout potential and could prove to be steals in your fantasy draft. As I’m sure you’ve surmised by now, I believe that TQS can be used as a tool to identify some of these guys.

Let’s look at a hypothetical player. He has a 4.50 LIPS ERA and, on the surface, appears to be a pretty unspectacular pitcher. This pitcher, however, also has a 30 percent Good and Great Percentage and a 30 percent Bad and Awful Percentage. All of those bad starts are going to have a large, negative impact on his LIPS ERA. All of those good starts, though, show us that this pitcher has some serious skills that could turn him into a star if he ever learns to put it all together. These types of pitchers could be prime late-round targets in your fantasy drafts.

Past breakouts and declines

Before we look ahead using TQS, let’s look back at a few players who could have been predicted to improve based upon TQS.

Erik Bedard


2004	25	4.59	4.56		7.93	4.65	38%
2005	26	4.00	3.95		7.94	3.62	39%
2006	27	3.76	4.06		7.84	3.16	48%
2007	28	3.16	3.13		10.93	2.82	47%

True Quality Starts

2004	26	42%	0%	4%	69%	27%	0%	4%	27%	14%
2005	24	54%	4%	12%	67%	17%	0%	17%	17%	100%
2006	33	64%	6%	18%	61%	15%	0%	24%	15%	160%
2007	28	79%	25%	32%	39%	4%	0%	57%	4%	1600%

While Bedard’s peripheral stats didn’t improve a whole lot from 2004-2006, there was a noticeable improvement in his TQS stats. Nearly every single category improved. His positive starts became more frequent, and his negative starts became more scarce. Maybe we couldn’t have predicted such a drastic jump in 2007, but this information might have persuaded us to take Bedard as a high-upside pick relatively late in the draft.

Rich Hill

A Hardball Times Update
Goodbye for now.


2005	25	9.13	4.20		7.99	6.46	38%
2006	26	4.17	4.12		8.15	3.53	29%
2007	27	3.92	3.78		8.45	2.91	35%

True Quality Starts

2005	4	0%	0%	0%	50%	50%	0%	0%	50%	0%
2006	16	50%	6%	19%	50%	25%	0%	25%	25%	100%
2007	32	69%	3%	25%	69%	3%	0%	28%	3%	900%

Hill was a popular sleeper pick going into 2007 and rewarded his owners with a good year. Looking back, his TQS stats would have supported his inclusion on a top sleeper list.

He had shown pretty good—though not great—peripherals to begin with, and his TQS stats showed that he had further room to improve. He had turned in some very good starts, but he was hampered by an equal number of not-so-good starts. In 2007, he improved his peripheral stats across the board while cutting his BA% down to just 3 percent. His GG% didn’t improve a whole lot, though his Above Average percentage did. If that upward shift continues, Hill could be even better in 2008.

Noah Lowry


2004	23	3.82	3.85		7.04	2.74	39%
2005	24	3.78	3.33		7.56	3.30	40%
2006	25	4.74	5.12		4.74	3.16	35%
2007	26	3.92	5.36		5.02	5.02	43%

True Quality Starts

2004	14	50%	21%	14%	64%	7%	0%	36%	7%	500%
2005	33	55%	0%	15%	73%	12%	0%	15%	12%	125%
2006	27	33%	0%	7%	74%	15%	4%	7%	19%	40%
2007	26	31%	0%	4%	69%	23%	4%	4%	27%	14%

Noah Lowry is a really interesting case. He dominated in 2004 in terms of TQS (21% Great Game percentage!), paving the way for a 3.33 LIPS ERA 2005. In 2006 and 2007, though, his LIPS ERA sat above 5.00. Before TQS, I didn’t know how to explain this other than that his peripheral stats fell off. Why they fell off, I had no idea.

Looking at TQS, though, you could see Lowry’s dominance severely decrease in 2005. If we looked at TQS then, we would have seen his good starts decrease by more than 100% and his bad starts increase by nearly 100%. We also would have seen him regressing towards league average (nearly three quarters of his starts were average!). That would have led us to be wary going into 2006, and we probably would have let someone else take the risk … and the accompanying fall.

Going forward, Lowry looks like a poor investment. His good starts continue to trend down and his bad starts continue to trend up.

Concluding thoughts

We’ll look at some potential TQS breakout candidates for 2008 someday soon. I also need to go over my strategy with closers, which I’ve held off on because I first need to explain the overall valuation methods I use. Look for those three things in the coming days!

Comments are closed.