Baseball Articles | Index

The Quality Start is a Useful Statistic

By David W. Smith
Spring, 1992

The quality start is a relatively new statistic devised by sportswriter John Lowe in an attempt to evaluate the performance of starting pitchers in terms other than the traditional values of ERA and wins and losses. A starting pitcher is credited with a quality start if he pitches at least six innings and allows three or fewer earned runs. The original motivation for the present study was to pursue a statement written by Bill James in the 1987 Baseball Abstract. While discussing some of the pros and cons of quality starts as a statistic, he noted that there had been criticism (by Moss Klein, writing in The Sporting News) of the quality start in that it would be possible in principle for a pitcher to go exactly six innings in every start and allow exactly three earned runs each time, compiling an ERA of 4.50, although each start would be categorized as a quality start. Moss that this possibility invalidated the entire concept. Bill thought Klein's criticism was an unreasonable extreme example and he ventured that "...I doubt that any pitcher had an ERA higher than 3.20 in his quality starts." My intuition on this point agreed with James', so I decided to pursue the question by using the Project Scoresheet data base, which covers all Major League games played from 1984 through 1991. Bill was right about the ERA of pitchers in Quality Starts, but there are in fact other interesting conclusions as well.

First, let's address the initial question: what is the ERA of pitchers in their quality starts? Table 1 is a summary of the results from all games from 1984-1991, 16831 games total, of which about 52% were quality starts. The information in this table is divided two ways. First, each league is presented individually, and second, the games are separated by quality starts and non-quality starts.


Table 1. Summary of Quality Start data from 1984-1991



          %QS  Starts     W     L   Pct       IP   ERA  IP/Start

AL QS    49.3    8935  6184  2749  .692  67370.1  1.93      7.54

AL NQS           9195  2878  6313  .313  44578.1  7.65      4.85



NL QS    54.9    8522  5590  2927  .656  62767.1  1.89      7.37

NL NQS           7010  2170  4833  .309  33171.2  7.28      4.73



ML QS    51.9   17457 11774  5676  .674 130137.2  1.91      7.45

ML NQS          16205  5048 11146  .311  77750.0  7.50      4.80



Totals          33662 16822 16822  .500 207887.2  4.00      6.18

    Note 1:

    The 33662 Starts came in 16831 games, of which 9 were ties.

    Note 2:

    There were 989 quality starts in which the pitcher had exactly six innings pitched

    and allowed 3 earned runs, for a "Klein percentage" of 5.7%.

There are three main points to make from this table.

1. The ERA of starting pitchers when they have a quality start is over five runs per game better than in games in which they do not have a quality start (1.91 vs 7.50 overall).

2. The winning percentage of the starting pitchers when they have a quality start is more than twice what it is when they don't have one (.674 vs .311 overall).

3. The innings pitched per start is also substantially different in quality start and non-quality start games (7.45 vs 4.80).

The conclusion seems clear that quality starts, taken in the aggregate, reflect much better than average performances, and that, in the aggregate, quality starts correlate with team winning percentage. Furthermore, these good performances are also of longer duration, meaning that the bullpen is given some rest when a quality start is taking place. What about Moss Klein's concern about the "minimum" quality start, with its ERA of 4.50? Over the last eight seasons, there have been 17457 quality starts, and 989 of them have been exactly six innings with three earned runs. On a percentage basis, which I call the "Klein percentage", this is a value of 5.7%. As shown in Table 2, there are five categories of quality starts which occurred more often, all with substantially better ERA. It's pretty clear that Moss has missed the significance of quality starts by his concentration on the extreme case.


Table 2. Major Categories of Quality starts from 1984 to 1991



 IP   ER    ERA   Number  % of all QS

7.0    2   2.57     1348      7.7

9.0    0   0.00     1343      7.7

7.0    1   1.29     1254      7.2

9.0    1   1.00     1163      6.7

6.0    2   3.00     1000      5.7

6.0    3   4.50      988      5.7

Another approach to this question is to look at the performances of individual pitchers. Table 3 is a summary of the 18 pitchers with the most quality starts since 1984. The results are presented in two ways, first by total number of quality starts, and second by quality start percentage.


Table 3. Individual Quality Start Leaders, 1984-1991

Quality Start Leaders in order of Total Quality Starts



Name              QS  NQS  QS Pct  Years

Frank Viola      177  108   62.1   84-91

Dwight Gooden    167   69   70.8   84-91

Ron Darling      164   87   65.3   84-91

Roger Clemens    161   79   67.1   84-91

Bob Welch        160   99   61.8   84-91

Nolan Ryan       158   93   62.9   84-91

Charlie Hough    154  114   57.5   84-91

Orel Hershiser   154   62   71.3   84-91

Mike Scott       153   80   65.7   84-90

Dave Stieb       150   92   62.0   84-91

Frank Tanana     149  111   57.3   84-91

Tom Browning     146  109   57.3   84-91

Bret Saberhagen  145   81   64.2   84-91

Mike Boddicker   144  120   54.5   84-91

Dave Stewart     135   95   58.7   84-91

Jimmy Key        132   85   60.8   85-91

Bert Blyleven    131  100   56.7   84-90

Doug Drabek      118   65   64.5   86-91



Quality Start Leaders in order of Quality Start Percentage

Name              QS  NQS  QS Pct  Years

Orel Hershiser   154   62   71.3   84-91

Dwight Gooden    167   69   70.8   84-91

Roger Clemens    161   79   67.1   84-91

Mike Scott       153   80   65.7   84-90

Ron Darling      164   87   65.3   84-91

Doug Drabek      118   65   64.5   86-91

Bret Saberhagen  145   81   64.2   84-91

Nolan Ryan       158   93   62.9   84-91

Frank Viola      177  108   62.1   84-91

Dave Stieb       150   92   62.0   84-91

Bob Welch        160   99   61.8   84-91

Jimmy Key        132   85   60.8   85-91

Dave Stewart     135   95   58.7   84-91

Charlie Hough    154  114   57.5   84-91

Frank Tanana     149  111   57.3   84-91

Tom Browning     146  109   57.3   84-91

Bert Blyleven    131  100   56.7   84-90

Mike Boddicker   144  120   54.5   84-91

The first list rewards longevity as well as good performances, which must be kept in mind when analyzing it. The second list removes this bias for longevity and we see some clear changes between the two. Frank Viola's total of 177 is impressive, but his 62.1% puts him in the middle of this select pack. Power pitchers, such as Gooden and Clemens are near the top, while the finesse artists like Tanana, Browning, Blyleven and Boddicker are at the bottom. I note with some surprise that fire-balling Nolan Ryan is only marginally ahead of Viola on the percentage list and that Orel Hershiser is overall percentage leader, even though he is not a true flame-thrower. Therefore quality start performance is not simply a reflection of power pitching.

The last thing to look at in examining the individual pitchers is exceptional single seasons. Table 4 gives two glimpses at some great performances. In the top part, we see that only three pitchers have had as many as 30 quality starts in a single season from 1984-1991, led by Dwight Gooden's unbelievable 33 in 1985, in which 97.1% of his games (all but two) were quality starts. In the bottom of the table are presented the best ERA seasons for those with at least 20 quality starts in a season. Four pitchers had a quality start ERA of less than 1.00, led by Gooden at 0.95, which wasn't even the year that he had the 33! (By the way his quality start ERA was 1.34 in 1985, when he had the 33 quality starts). Note, however, that even in these excellent seasons, these pitchers had non- quality start ERA above 6.00, with Gooden up to 8.56. At a minimum, this table tells us that a quality start performance really is different from an average game, and that a non-quality start, even for the top stars, is a far inferior performance.


Table 4. Exceptional Individual Seasons for Quality Starts



Top Individual Seasons for Quality Starts

Name              QS  NQS  QS Pct  Years

Dwight Gooden     33    2   97.1    1985

Mike Scott        32    5   86.5    1986

Bret Saberhagen   30    5   85.7    1989 



Best Season ERA in Quality Starts

Name              QS  QSERA  NQS  NQSERA  QS Pct  Year

Dwight Gooden     21   0.95   10    8.56    68    1984

Mike Moore        22   0.95   13    6.53    63    1989

John Tudor        27   0.98    9    6.60    75    1985

Jack Morris       20   0.99   15    7.39    57    1986

Orel Hershiser    25   1.01    9    7.38    74    1985

Dennis Martinez   21   1.07   10    6.54    68    1991

John Tudor        21   1.16    9    6.54    70    1988

Mike Boddicker    20   1.18   14    6.07    59    1984

Bob Ojeda         21   1.20    9    7.52    70    1986 

So what does all this mean about the meaningfulness of the quality start statistic? That question translates to the effect on winning the game, which is, after all, the purpose of the competition. The answer is complicated somewhat by the fact that both pitchers could very well have quality starts, but only one of them will win. From 1984-1991 there were 4793 games in which both starters had quality starts, a significant 54.9% of all quality starts (since both starters had quality starts in these games, they represent 9586 quality starts).

To sort this out, one might suppose that the thing to do is only look at those games when one pitcher as a quality start, to see if his performance was valuable. Well, I did that, but I'm not going to present the numbers, because it's really not a very meaningful analysis. Think about what those games represent: one starter does very well and the other one does not. It is hardly surprising, or meaningful, that, in those games, the pitcher with the quality start rarely loses. The other variable to take into account is that pitching changes are done rather differently in the two leagues, due to the Designated Hitter. We are all familiar with the argument that National League pitchers run the risk of being pinch-hit for in close games, even if they are pitching well, a risk that American League pitchers do not have to fear.

The other potentially confounding effect, which I have not explored here, is the changing use of relief pitchers, with middle men, setup men, and closers. Nonetheless, it is clear from table 1 that in both leagues, teams whose pitchers have a quality start have outstanding records in those games. Remember that most of the quality start losses were in games where the winning pitchers also had a quality start.However, I feel that there is a potential for misapplying the notion of quality starts by incorrectly concluding that every time a starter has a quality start he "should" win the game. The flaw here is that the quality start is really meaningful only in the aggregate, over the whole season.

Attempting to apply it to a single game is as meaningless as calculating a single game batting average or a single game ERA. Taking this longer term view into account and discarding the notion of predicting the outcome of a single game, we can see that there is a significant value to the quality start related to a starting pitcher's season-long durability and consistency. These features will benefit the team by keeping them close in a larger number of games and in allowing the manager to regulate his bullpen usage with more freedom.

All categories of baseball analysis have limitations on their usefulness, most of which are apparent from a common sense inspection. As one example, many announcers are fond of reporting how well certain batters do in bases loaded situations in a given season, with a batting average of .333 (or some other nice number). What they usually forget to emphasize, of course, is that the average probably has come from a very small number of opportunities, such as 3 hits in 9 at bats. Common sense tells us that a selected group of 9 at bats is very unlikely to be an accurate above, the quality start statistic certainly can be misused, but that is a weakness of the person doing the misapplication, not of the basic category itself. I conclude that the quality start statistic has substantial value in the appropriate context and that it should be used more widely within that context.