Is Barry Zito among the most or least reliable starters?

We are now in the tenth decade of the live-ball era, generally acknowledged to have begun in 1920. Offensive and defensive periods have come and gone within this era, even to extremes like the 1960s, sometimes referred to as the second dead-ball era, and the homer-happy 1990s and 2000s.

With those changes, the use of starting pitchers has also changed. This post will explore those changes and look at the pitchers who have most frequently exceeded and most frequently fallen short of the changing performance standards expected of starting pitchers.

First, let’s look at length of start. At the beginning of the live-ball era, starters were expected to also be finishers or, at the least, pitch late into games. That expectation has changed to the point where pitching 8+ innings has happened in only 10% of starts since 2000, compared to more than 50% of starts prior to 1950.

Here’s a chart showing change in the length of starts.

Thus, the median length of start has grown consistently shorter, as shown in the table below.

Table 1: Cumulative Percentage of Starts by Innings Pitched per Start 
 Decade 0-4+ 5+ 6+ 7+ 8+ 9+
1920-29 19.5% 25.7% 33.4% 43.6% 57.9% 100%
1930-39 20.5% 27.2% 36.6% 48.3% 61.7% 100%
1940-49 21.1% 28.2% 37.6% 49.4% 62.6% 100%
1950-59 24.3% 32.8% 44.2% 57.7% 69.4% 100%
1960-69 23.4% 34.0% 48.1% 64.4% 75.4% 100%
1970-79 20.9% 32.0% 47.9% 64.1% 76.2% 100%
1980-89 20.9% 35.3% 54.5% 74.0% 85.5% 100%
1990-99 19.8% 37.3% 60.8% 82.6% 93.4% 100%
2000-12 17.5% 38.4% 67.7% 90.0% 97.1% 100%

The level of performance delivered by pitchers has also changed over time, as measured by Game Score. This metric is influenced by how long the pitcher pitches, how hard the pitcher is hit, and also by a pitcher’s strikeouts and walks. A higher game score indicates a game where a starting pitcher has pitched longer and better than in other games with inferior game scores.

The table below shows median game scores for each decade and length of start. The trend to (mostly) increasing game scores over time is a product of higher strikeout rates, but also of where starts of a certain length fall on the continuum of all starts in a particular decade. For the example cited earlier of starts of 8+ innings, over 35% of starts were longer (CGs or extra innings) prior to 1950, but less than 3% have been longer since 2000; thus, the game score for the post-2000 starts would be expected to be higher inasmuch as they more clearly represent the “cream” of the game starts of that period.

Table 2: Median Game Score by Length of Start
Decade 5+ 6+ 7+ 8+ 9+
1920-29 33 38 45 49 64
1930-39 34 39 46 50 65
1940-49 37 43 49 54 69
1950-59 39 45 52 56 70
1960-69 43 50 57 62 75
1970-79 41 48 55 61 72
1980-89 42 49 58 63 74
1990-99 41 50 59 67 77
2000-12 41 52 62 70 80

The above table is represented graphically below.

Putting together Table 1 and Table 2 gives us an approximate estimate of average performance level by starters in each decade, in terms of both innings pitched (quantity) and game score (quality). As an example, consider the 1950s. From Table 1, we see that the median start length was at 7+ innings and the corresponding median game score from Table 2 is 52. Thus, starts of 7+ innings with a game score of 52 can be a  proxy for average starter performance in the 1950s.

Following from the above example, I looked at the pitchers who most often turned in starts above and below median performance levels for both innings pitched and game score. These pitchers are identified in the following tables, showing the top 15 and ties in number of games, sorted by % of starts represented by those games.

First, the pitchers most often having games at or above the median innings pitched and game score thresholds. Best way to view the tables is to by decade, by typing the first year of the decade (ends in zero, e.g. 1950) in the Search box.

[table id=80 /]


And, the pitchers most often below the median innings pitched and game score thresholds. Note that this does NOT mean the “worst” pitchers. Indeed, some pitchers appear both in this table and the preceding one. Perhaps you could say these are the pitchers good enough to get the ball on a regular basis and for an extended period, but who also may cause the most anxiety for their managers when they’re on the mound.

[table id=79 /]


Lastly, the title question. I pose this question because Barry Zito appears in the top 4 of both above tables for the 2000 to 2012 period. Most often above median, and also most often below that level. In this case, I suspect we’re looking mainly at two half-decades, or perhaps the tale of of San Francisco Bay. East side Barry (2000-2006) has 125 ERA+, but west side Barry (2007-2012) is at just 91 ERA+.

Leave a Reply

15 Comments on "Is Barry Zito among the most or least reliable starters?"

Notify of
Sort by:   newest | oldest | most voted
Mike L

I don’t want to overgeneralize, but the seems to be a lot more “feel” pitchers in the second group, the type who didn’t have great enough stuff to survive when their pitch placement was somewhat off. Which also mirrors the early Zito and the later one. Interesting charts.

Check out Barry Zito’s Game Log from 2012. Out of 32 total starts, he pitched 17 Quality Starts and another 2 starts of 5+ IP in which he allowed 2 runs. In those 19 starts he had an ERA of 1.99 including 5 starts where he went 7+ IP while allowing 0 runs. In his other 13 starts, his ERA was just under 9.00. 2011 was lost to injury, but his 2010 campaign was very similar to 2012. Barry Zito is either very good or very bad in his starts. Yes, I am sure he was very good more often… Read more »
no statistician but
Doug: To me it would be helpful to see the total number of starts of each pitcher as a kind of control. Just doing some basic grunt work on a few I was interested in, Koufax had 70% of his starts in the sixties above median. Halliday had 64% of his starts 2000 onward above median. Bunning had 56% above in the sixties and 32% below, whereas poster child Zito had 56% above and 29% below. Ford in the fifties had 58% above, Wynn 54%, Spahn 67%, Antonelli 55%, Pierce 62%. Knowing how many times a pitcher was above the… Read more »
One interesting thing I saw looking at your first chart and table was that while managers are more likely to take their starter out in the 6th inning or later, they are less likely to take them out before that. I think this is largely due to construction of the pitching staff. In the 50’s almost everyone was thought of primarily as a starter, so if you took your starter out in the 1st, 2nd or 3rd inning you had someone ready and capable of pitching many innings in the bullpen. Today, with the disappearance of the spot starter, the… Read more »

Good observation, kds.

I was struck by how the percentage of starts of less than 5 innings has been more or less constant. Your explanation of a possible reason makes a lot of sense.

Also, that current CG line also rings true. Unless a pitcher is totally dominant, he’s unlikely to stay in today. In the past, if you were up 9-5 or something like that in the 7th or the 8th, the logic seemed to be something like “our guy isn’t great today, but as long as he’s throwing strikes and we’ve got the lead, we’ll leave him in there”.

Thanks kds. I was going to post how I was intrigued by the 0-4 innning start % peaked in the 50’s and 60’s, but your explanation sounds perfect. ” In the 50′s almost everyone was thought of primarily as a starter, so if you took your starter out in the 1st, 2nd or 3rd inning you had someone ready and capable of pitching many innings in the bullpen. Today, with the disappearance of the spot starter, the swing man, and even the long man, the manager is less likely to take out an ineffective starter early because of the cost… Read more »

Just to show one way the game has changed, there was a Zane Grey story in The Red-Headed Outfield that featured a pitcher who gave up 7 runs in the first inning and won the game.

Richard Chester

Doug: Just a suggestion. I think you are better off creating 2D charts rather than 3D. The 3D charts do not properly align with the grid lines. Looking at your first chart the back edges of the tops of the columns should be tangent to the 100% line but there is a slight gap. I created a 3D chart of my own and the same thing happened. Also the grid lines pertain to the back edges of the columns but your eyes want to align them with the front edges.