Batter's Box Interactive Magazine Batter's Box Interactive Magazine Batter's Box Interactive Magazine
Jordan Furlong postulated the following in the April 27 Game Thread:

Hey, here's a project for anyone interested in doing the legwork -- rank the Blue Jays' best and worst 5th starters in franchise history...I'll bet you'll find Josh Towers has been one of the best the Jays have sent out there.

Legwork and pitchers in the same sentence? A rapier-like wit, this one. Let’s get right to it.

There are some rules we need to establish. In order,

  1. A Fifth Starter must make at least 15 starts.
  2. A Fifth Starter must pitch at least 90 innings.
  3. A Fifth Starter must not be in the top four in Games Started on any given team.
  4. If more than one pitcher qualifies in any given year for the same team, the one with fewer relief appearances is designated the fifth starter.

We started with 507 pitcher-seasons, and the 14 remaining pitchers are shown below, ranked by ERA+.


Fifth Starter   Year   GS   ERA    K/9   WHIP   ERA+  IP
Halladay	2001   16   3.16   8.2	 1.17   150   105
Candiotti	1991   19   2.98   5.6	 1.26   142   130
Bush	        2004   16   3.69   5.9	 1.29   132    98
Stottlemyre	1994   19   4.22   6.7	 1.45   114   141
Alexander	1983   15   3.93   3.5	 1.31   110   117
Guzman	        1998   22   4.41   7.0	 1.41   106   145

Stewart	        1993   26   4.44   5.3	 1.37    99   162
Stottlemyre	1989   18   3.88   4.4	 1.46    97   128
Alexander	1986   17   4.46   5.3	 1.30    95   111
Moore	        1979   16   4.84   3.3	 1.59    90   139
Cerutti	        1990   23   4.76   3.2	 1.54    83   140
Moore	        1978   18   4.93   4.7	 1.57    80   144
Hamilton	1999   18   6.52   5.1	 1.63    75    98
Stottlemyre	1988   16   4.34   6.2	 1.62    69    98

Thanks to Baseball-Reference.com, possibly the greatest site in the world, for the historical statistics.

The best fifth starter in Blue Jays history? None other than Roy Halladay in 2001, one year removed from his terrible season and two years before his Cy Young season. (Aside: Is anyone surprised that Joey Hamilton is at the bottom of this list?) Dave Bush is the top-rated Rookie Fifth Starter (now there's a worthy title) and I was surprised to see so much of Todd Stottlemyre, since he shows up three times out of 14 pitchers.

What about Jordan’s Josh Towers query?

Well, Towers’ 2003 season was right in the middle, with a 105 ERA+ but the second-lowest WHIP, 1.21. However, that wasn’t a Fifth Starter year because he didn’t pitch 90 innings. He barely cleared 60. If you want to consider him a fifth starter in 2003, then he was, at best, in the middle of the pack.

In 2004, Josh Towers was not very good. He wasn’t considered the fifth starter because only three pitchers, Roy Halladay, Miguel Batista and Ted Lilly, "out-started" him and the Bush-Hentgen-Miller mix behind him all took starts away from each other in one way or another. Anyway, his 2004 would have been 10th out of 15, so...no, not good.

Jordan said: I'll bet you'll find Josh Towers has been one of the best [fifth starters] the Jays have sent out there.
Rob says: Sorry, Jordan, it looks like Towers wasn’t one of the best. Now, as for that bet...oh, I'm just kidding. Maybe 2005 will be the year for Control Towers.

Then I decided to check out every team in 2004, to determine the best fifth starter in the American League. Using the same qualifications, we end up with 10 pitchers from 14 teams:


Fifth Starter	 Team  GS   ERA    K/9   WHIP   ERA+  IP
Tim Hudson	 OAK   27   3.53   4.9   1.26   133   189
David Bush	 TOR   16   3.69   5.9   1.23   132    98
Bronson Arroyo	 BOS   29   4.03   7.2   1.22   121   179

Aaron Sele	 ANA   24   5.05   3.5   1.62    91   132
*T Mulholland	 MIN   15   5.00   4.4   1.59    91   123
Joaquin Benoit	 TEX   15   5.68   8.3   1.40    89   103
Gary Knotts	 DET   19   5.25   5.4   1.48    86   135
Jason Davis	 CLE   19   5.51   5.7   1.74    82   114
Jose Contreras	 NYY   18   5.64   7.7   1.41    80    96
Mike Wood	 KCR   17   5.94   4.9   1.40    75   100

Dave Bush shows up again. He was beaten by just one pitcher for the 2004 American League Todd Stottlemyre Award, and that man was not your typical Fifth Starter. Tim Hudson missed a few starts, so Zito-Redman-Mulder-Harden all passed him in Games Started. Otherwise, he wouldn't show up here. Very close, Mr. Bush, but not quite.

(Note: A new rule came into play here -- if a pitcher shows up as a Fifth Starter for two different teams, he is removed from the study. Freddy Garcia made 15 starts for Seattle and 16 starts for Chicago after being traded for Miguel Olivo and Jeremy Reed on June 27. I took Garcia out because the very idea of a fifth starter means you're the fifth-best pitcher on your team, not on two different teams at once.)

I'll do the same for the 2005 American League once the statistics are available in the off-season.

What do the Bauxites notice in these stats? Anything stand out at you? Wondering why a certain pitcher isn't listed? I ignored the National League because I generally ignore the National League. I hate watching pitchers hit.

Oh, and it's perfectly possible that I missed a qualified Fifth Starter somewhere. Feel free to point out any errors and ask any questions.

I'll start this off by asking everyone who they think will be Toronto's Fifth Starter this year, and what his ERA will be. Remember, you can answer "nobody," if you think whoever makes the fifth-most starts will pitch less than 90 innings.

A Quick-and-Dirty Fifth Starter Study | 16 comments | Create New Account
The following comments are owned by whomever posted them. This site is not responsible for what they say.
Thomas - Saturday, April 30 2005 @ 05:33 PM EDT (#114590) #
Nice work, Rob.

One quibble is that your "new rule" should eliminate Candiotti's 1991 from consideration on the Jays list, as he had the fifth-most starts for the Indians that year.

I think, by any subjective measure, it's pretty obvious that Arroyo was really the best 5th starter in the AL, and likely baseball, last year. Those sort of numbers from the fifth spot in the rotation are exceedingly rare and his number of starts pushes him over Bush as quantity has to be a factor, too.

No study of "numbered" starters is ever going to rock solid, because the whole system is somewhat problematic in the first place, but as brief look at the situation, it was a fine job.
Mick Doherty - Saturday, April 30 2005 @ 06:02 PM EDT (#114592) #
Contreras is a dubious listing, too, since he made the sixth-most starts for CHW, with 13, after leaving NYY in the ill-fated Loaiza deal. (Now there's a phrase more than one club is familiar with!)
Rob - Saturday, April 30 2005 @ 06:51 PM EDT (#114594) #
Yes, when introducing a new rule, it would make sense to apply said rule retroactively. I didn't even look at Candiotti's Cleveland stats for 1991 -- which, incidentally, were better than his Toronto numbers.

Contreras may be Jason Dubious, but he slipped underneath with 13 GS and 75 IP. Should I change the cutoff point, or is 15/90 good enough?
Mick Doherty - Saturday, April 30 2005 @ 06:53 PM EDT (#114595) #
Nah, don't change anything. If you went to 12/75, I'd just find someone who was at 10/68. You know.

Incidentally, don't you mean "Blanche" Dubious?
Rob - Saturday, April 30 2005 @ 07:05 PM EDT (#114598) #
That's two TV-related memories you've inspired in the last three minutes, Mick -- Ned Flanders screaming "Stella!" and, from your Hall of Names piece, the telephone call for "Cartwright" on Seinfeld.

Here are the guys who were on the bubble: Chan Ho Park, Orlando Hernandez, Justin Miller, Pat Hentgen, Eric DuBose, Matt Riley. Not much there -- two pitchers from Toronto, two from Baltimore -- but El Duque was five innings away from inclusion, and he was better than Contreras. The very fact that my 15/90 limitation did not include Chan Ho Park can only be a good thing, though.

I'd just find someone who was at 10/68

(Cue Henny Youngman) "Take John Wasdin...please." 10/65.

StephenT - Sunday, May 01 2005 @ 01:29 AM EDT (#114611) #
The "5th starter" is the pitcher rated #5 by management, not necessarily whoever ends up 5th in starts.

The best automatic definition I can think of is to use the 5th different pitcher to start a game for the team during the season, though no automatic definition will be foolproof because of injuries, etc. If one was going to do a serious study, one should manually review each season to be sure of picking the right guy. (This might be hard for someone who didn't follow the team that year.)

Tim Hudson plainly was not a 5th-starter last year.

I think Towers was the Jays' official "5th starter" last year. Justin Miller might have been in that role later on (I forget now) when #1 was injured. I don't think Bush was ever thought of as a 5th starter last year.

A lot of the listed 5th-starters for the Jays above don't look right. e.g. in 2001, the 5th-starter was Michalak, not Halladay.

In 1993, I think the 5th-starter was Hentgen; certainly not Stewart. Hentgen apparently was rated #3 by the playoffs, but he was the 5th to start during the regular season. One might argue Hentgen really was rated #6 to start the season because Stewart was (I think) injured, but I'm comfortable with calling Hentgen a 5th-starter to start the 1993 season.

Hentgen went 19-9 in 1993. That might be the best for someone rated #5 to start the season.
Michael - Sunday, May 01 2005 @ 06:23 AM EDT (#114612) #
There are big problems with counting a #5 starter that way. By my calculation that set of 5 starters put up, on average, a 98 ERA+ with 125 IP. Most teams would take that from their #5 starter in a heartbeat. My guess is that if you used a similar "most games started" cutoff for Jays #4 and #3 you would see similar numbers put up (in terms of quality, obviously they'd each average more IP as you are choosing the guys with most starts). Even the #1 and #2 guys are probably not going to be that much better, although the higher you go you drop probably drop the really bad sub 80 ERA+ guys so the average may be a little higher.
JackFoley - Sunday, May 01 2005 @ 06:24 AM EDT (#114613) #
I believe a fifth starter is best described as the starter most likely to be bumped from the rotation should an upgrade present itself. I appreciare that that is not a particularly tangible definition, nor does it lend itself to easy statistical manipulation, but I think it's arguably the correct one.
CeeBee - Sunday, May 01 2005 @ 07:07 AM EDT (#114614) #
Slotting starters by games started, innings pitched or some other statistic doesn't really define the #5 guy. As has been mentioned the number 5 guy is usually the one who gets bumped out by a trade, a callup or an offday. He's almost always the first starter on the bubble and as such also subject to stints in long relief or mop up duty. One other factor that I don't think has been mentioned is injury. Many of the #5 starters chosen by GS /IP are only in the 5 slot because they spent time on the DL. Still makes interesting reading but its kind of like comparing apples to oranges. :)
Magpie - Sunday, May 01 2005 @ 10:27 AM EDT (#114619) #
I basically agree with Stephen T's point - teams begin a year thinking so-and-so if the 5th starter. And then stuff happens.

Last year in Toronto, that was definitely Towers. He wasn't just bumped - he was DFA'd. But then Doc got hurt, and Towers' replacement got hurt, and Hentgen failed utterly - and not only was Towers back, he was suddenly like the 3rd starter.

Actually examining this, for each team, would be a very large task indeed and well beyond the scope of any Quick and Dirty look.

And Hentgen did begin 1993 as the 5th starter, having nosed out Leiter for the job. But Stewart was indeed hurt, so Leiter made a couple of early season starts. Leiter later replaced Stottlemyre when Todd was out for a month, and Morris when he was shut down.

For our many readers who weren't there.

Rob - Sunday, May 01 2005 @ 10:41 AM EDT (#114622) #
The best automatic definition I can think of is to use the 5th different pitcher to start a game for the team during the season

That is better, and I'll most likely use that definition for the 2005 mini-study -- you can forgive me for not scouring the Retrosheet logs for 27 seasons. However, Ted Lilly would end up as the fifth starter for Toronto this year. I can fix that by replacing him with Towers for the study, but what if Towers goes down mid-season and Downs/Rosario/Gaudin pulls a Bush and pitches the last three months?

Actually examining this, for each team, would be a very large task indeed and well beyond the scope of any Quick and Dirty look.

Not as large as you think. They are: Chen (BAL), Schilling (BOS), Davis (CLE), Ledezma (DET), Anderson (KCR), Gregg (LAA), Mays (MIN), Brown (NYY), Harden (OAK), Franklin (SEA), Nomo (TBD), Astacio (TEX), Lilly (TOR).

Looking over that, Schilling can be replaced with Arroyo, Harden with Saarloos and Lilly with Towers. The rest all seem like fifth starters to me.

NDG - Sunday, May 01 2005 @ 11:23 AM EDT (#114629) #
Except for Gregg in Anaheim. He spot started until Escobar came off the DL. Paul Byrd is really the fifth starter there.
NDG - Sunday, May 01 2005 @ 11:29 AM EDT (#114631) #
The problem with this method is that half of these guys are going to lose their jobs by midseason ... then what? It might be better to determine the top 4 for each team, then lump every other pitcher into #5. Using this method, both Josh Towers and David Bush would have qualified as #5 starters last year. I don't see the problem with this as they'll all get ranked, and it can still be determined which team got the best out of their fifth starter.
NDG - Sunday, May 01 2005 @ 11:34 AM EDT (#114632) #
Further to my last comment (just tell me to shut up any time now). They real goal of this mini-study should be who get's the most out of spots in the rotation that weren't expected to be good. Hence if a team has three starters DL'd for extended periods, they may end up with multiple 'fifth starters', but each of those starters weren't expected to contribute much at the start of the season. Therefore including them in the study may make sense.

Okay, I'm done now.
John Northey - Sunday, May 01 2005 @ 12:07 PM EDT (#114634) #
Fun study, but as others have pointed out there is something wrong when a guy who is clearly not a #5 guy leads the list.

If I was to try this study (and I might sometime if I have time on my hands) I'd probably go for the guy with the 5th best ERA on a team who had 15 or more starts (a minimum for a guy who was regularly pitching imo). Might reduce it to 10 if I found it hard to get 5 starters in most cases. That way a guy who is clearly the #5 guy will not be listed as the #3 guy due to injuries to the top 2, pushing them to the bottom of the rotation. Being healthy is not a qualification for being a teams 'ace', it helps but doesn't make it true.

Ah, screw it, here is a listing (can't resist stats and I love using the Lahman database)...

2004 Jay Starters
#1: David Bush     3.69
#2: Ted Lilly      4.06
#3: Roy Halladay   4.20
#4: Miguel Batista 4.80
#5: Josh Towers    5.11
Not bad. Might need a condition on starts to prevent guys like Bush from making the #1 slot, but Towers comes in #5 where he should be (Miller, Hentgen are #6/7 respectively).

Checking Jays history we get from 4 to 7 pitchers qualifying in any one year (2004 the only year with 7, 6 occured 5 times)...
Year Pitcher           W  L GS ERA
2004 Josh Towers       9  9 21 5.11
2001 Esteban Loaiza   11 11 30 5.02
1999 Kelvim Escobar   14 11 30 5.69
1998 Pat Hentgen      12 11 29 5.17
1994 Dave Stewart      7  8 22 5.87
1993 Jack Morris       7 12 27 6.19
1991 Todd Stottlemyre 15  8 34 3.78
1990 John Cerutti      9  9 23 4.76
1989 Mike Flanagan     8 10 30 3.93
1988 Jim Clancy       11 13 31 4.49
1986 Doyle Alexander   5  4 17 4.46
1983 Jim Gott          9 14 30 4.74
1979 Phil Huffman      6 18 31 5.77
1978 Jerry Garvin      4 12 22 5.54
Lots of funny results vs perception.

Loaiza (started season as ace), Hentgen '98 (started as #2, Woody was thought of as #5 but came up as #4 with an ERA more than 1/2 a run better than Hentgen), Stewart '94 and Morris '93 (both were supposed to be either #1 or #2 guys), Flannagan '89 (#3, Cerutti viewed as #5 but comes up as the #1 guy), Alexander '86 (Stieb is #6!).

This method produces more of a depth chart for the rotation than anything really. When your #5 guy has an ERA below 4 ('89 and '91) it is a very good thing.

Hrm... had to check #1 guys too since I have the data now.

Year	Pitcher 	W	L	GS	ERA
2004	David Bush	5	4	16	3.69
2003	Roy Halladay	22	7	36	3.25
2002	Roy Halladay	19	7	34	2.93
2001	Roy Halladay	5	3	16	3.16
2000	Frank Castillo	10	5	24	3.59
1999	Roy Halladay	8	7	18	3.92
1998	Roger Clemens	20	6	33	2.65
1997	Roger Clemens	21	7	34	2.05
1996	Juan Guzman	11	8	27	2.93
1995	David Cone	9	6	17	3.38
1994	Pat Hentgen	13	8	24	3.40
1993	Pat Hentgen	19	9	32	3.87
1992	Juan Guzman	16	5	28	2.64
1991	Tom Candiotti	6	7	19	2.98
1990	Dave Stieb	18	6	33	2.93
1989	John Cerutti	11	11	31	3.07
1988	Dave Stieb	16	8	31	3.04
1987	Jimmy Key	17	8	36	2.76
1986	Jimmy Key	14	11	35	3.57
1985	Dave Stieb	14	13	36	2.48
1984	Dave Stieb	16	8	35	2.83
1983	Dave Stieb	17	12	36	3.04
1982	Dave Stieb	17	14	38	3.25
1981	Dave Stieb	11	10	25	3.19
1980	Jim Clancy	13	16	34	3.30
1979	Tom Underwood	9	16	32	3.69
1978	Jim Clancy	10	12	30	4.09
1977	Jerry Garvin	10	18	34	4.19
Btw, didn't feel like calculating out ERA+. Thus Clemens in '97 is tops (makes sense) while Jerry Garvin in '77 comes out worst (given the era's I suspect this would hold with ERA+). Stieb is #1 7 times (5 in a row) while second is Halladay at 4 and 3 guys at 2 times (Key, Clancy, Clemens, Guzman, Hentgen) and a 7 guys at 1 time each (from Bush to Garvin). This passes the smell test with a few funny guys on the list (Frank Castillo and Bush for example) but overall matches perception fairly well.

OK, one more before I stop. The all-time most often #1/#2/#3/#4/#5/#6/#7 guys for the Jays (ie: were this more than anyone else). The stats are averages for the years they were in that role. The ERA is just a straight average, not adjusted for innings in each season.

    Pitcher      How Often  Avg ERA Avg W-L Avg Starts
#1: Dave Stieb       7        2.97    16-10    33
#2: Jim Clancy       3        3.72    15-12    37
#3: Jimmy Key        4        3.62    14-10    29
#4: Todd Stottlemyre 4        4.39    11-12    27
#5: Mike Flanagan    1        3.93     8-10    30
#6: Joey Hamilton    2        6.20     6- 8    20
#7: Pat Hentgen      1        6.95     2- 9    16
Clancy and Carpenter tied for #2 but I put Clancy as the true #2 as he started 111 games over those 3 seasons vs Carpenter at 82 starts. #5 was a massive tie at 1 times each. Stottlemyre probably would win the #5 slot as well with his 34 starts and 3.78 ERA in 91, but since he is #4 I gave it to Mike Flanagan and his 30 starts and 3.93 ERA in '89 (Phil Huffman has 31 starts but a 5.77 ERA...ugh)

Fun to play with eh? Hentgen was #1/#2/#3 2 times each, Clancy was all of #1 through #5 at some point, Key always top 3. And I better stop now or there will go my whole day.
Matthew E - Sunday, May 01 2005 @ 01:27 PM EDT (#114642) #
I don't think there's any rule you can apply to identify the definite #5 pitcher in any year. For one thing, there may be more than one of them in a given year. The team could spend the whole year auditioning four guys for the position. What the hey, they may not even know who their #3 and #4 guys are.

Take 1986. Doyle Alexander wasn't the fifth starter that year; he was one of the front four. Then he got traded, and Joe Johnson took his spot. The real fifth starter was John Cerutti, who worked as a swingman that year and on and off for the next few years.

Similarly in 2002. I call Pete Walker the fifth starter for that year, no matter what the numbers say. He was clearly a guy for whom the team's attitude was 'we can use this guy for now, but we'll drop him as soon as we get something better'. Chris Michalak was the same way. And you can't just apply some mathematical cutoffs to detect it; you have to understand what the team was up to.
A Quick-and-Dirty Fifth Starter Study | 16 comments | Create New Account
The following comments are owned by whomever posted them. This site is not responsible for what they say.