37 thoughts on “Jekyll and Hyde Pitchers”

1. Voomo Zanzibar May 10, 2012 at 10:03 pm

   Caraway ended 10-24, 6.22

   At the end of May he was 6-3, 2.21, Zero DS

   The rest of the way that is 4-21, 8.82
   16 out of 22 starts were Disasters.

   Reply ↓
2. e pluribus munu May 11, 2012 at 12:03 am

   I’m going off the rails early in this string (with apologies, Doug; this research is not only imaginative and impressive, it’s a lot of fun) because it relates to an issue that has bothered me for over fifty years: the degree to which an occasional disastrous outing (starting or, more often, relieving) can distort the value of ERA.

   For example, it’s easy to recall John Smoltz in 2002: he gave up 8 runs to the Mets in one early 0.2 IP adventure and his ERA never recovered. Actually, checking B-R, that season he pitched over 80 innings, yielding 29 ER, but almost half (14 ER) came in three outings of just 2.2 IP total. Over those three appearances, his ERA was a non-Smoltzian 57.27; over his remaining 72 appearances – virtually his entire season – his ERA was 1.74. His ERA of record in 2002 was 3.25 – mediocre for a closer. Of course, we all knew what was going on at the time – Smoltz came in third in the Cy Young voting – but the ERA stat is simply useless: it misrepresents his performance by treating average as normal.

   I wonder whether a stat generated by a regression analysis (my own math is at fifth-grade level, so I have no idea what I’m talking about) could be developed that would smooth these sorts of distortions and yield an ERA measure that indicated normal performance, rather than average.

   (PS: For another type of Jekyll-Hyde pitcher, check out Dick Hyde’s season-to-season ERA.)

   Reply ↓
   1. Doug Post authorMay 11, 2012 at 1:04 am

      One way to do this is throw out the best and worst two starts and recalculate.

      Reply ↓
      1. e pluribus munu May 11, 2012 at 10:24 am

         Thanks for following my side track, Doug, and with math even I can do. That is the sort of remedy I was thinking of. But the +/- extremes often have unequal impact, and different pitchers have different numbers of outliers (though beyond a point, they’re no longer outliers). So a real stat that could be used for both practical and historical purposes would probably need to involve two numbers, one indicating a pitcher’s season performance norm and one his degree of variabilty from the norm.

         Here’s an example, unsing game score (not the same as ERA per game, but available on B-R): In 1962, Koufax produced the following distribution of game scores: 95/86/83/82/81/81/81/80/80/76/76/75/74/74/59/59/56/56/
         49/47/44/43/43/40/33/31.
         If a “quality start” (bad stat, I know) translates to something like a 45-50 game score (with considerably more elasticity for rising than diminishing numbers – those bottom three, even the 40, are all “disaster starts”: just a quirk of the game score stat), then this sequence tells you something about Koufax’s very high reliability in ’62 reaching and exceeding that standard. But his best start – a no-hitter (with five BB) – is actually much less an outlier from his upper tier performances than his disaster starts: and, in fact, the bottom three starts are all the effects of the finger injury and long DL stint that closed his season, when his ERA went from 2.06 to 2.54 over 9.2 IP. I’m thinking of some stat that would reduce without entirely erasing the impact of a nearly 25% rise in ERA, while indicating that there was some unusual variation from the norm in his performance.

         By the way, the only game I saw Koufax pitch that year was his 44 game score day: his debut vs. the Mets, a 13-hit CG gem. Only the Mets could get 13 hits against Koufax and lose by seven runs. (In game 2, they executed a triple play. Lost that one too.)

         Reply ↓
         1. Doug Post authorMay 11, 2012 at 1:22 pm

            Not often I see 13 hits and pitching gem in the same sentence, but I will defer to the eyewitness 🙂 .

            Your point about performance consistency/variability is well taken. Couple of thoughts that pop into my head about expressing both the expected performance level and expected consistency in one number are these:

            – GS75 – bottom qusrtile of game scores: the minimum performance level expected 75% of the time

            – GSR – game score rating: mean (or median) game score divided by standard deviation of game scores. My guess is a really good GSR would be around 10, and an average one around 5.

            Reply ↓
            1. e pluribus munu May 11, 2012 at 4:16 pm

               Doug, I can understand GS75 well enough; GSR is a stretch for me, but I think it’s closer to what I was getting at. (Of course, I’d be interested in a formula that works for all outings, not just starts.) I’ll link to some linear regression tutorials and see whether I can hold onto enough to think through your idea. Thanks!

               “Gem” was ironic. Koufax seemed befuddled by the Mets, the heat, and the behavior of the huge crowd of about 55K. It was the first chance to see Koufax back in NY since he’d become Koufax – the crowd went delirious with joy when he was introduced and at every Met strikeout. But the Mets were had already established themselves as so amazing in their awfulness that whenever one got a hit off Koufax the crowd erupted equally in joyful shock and awe. It made no sense, and I remember during one explosion of cheers Koufax ambling towards first base, throwing his arms in the air. – By the way, B-R reminds me that the losing pitcher that day was Jay Hook: this was one of the many 1962 disaster starts that earned him an honored place on your second list above.
   2. RJ May 11, 2012 at 9:09 am

      I’ve mentioned this elsewhere before, but Madison Bumgarner’s ERA last year would have come down from a respectable 3.21 to a pretty darn good 2.81 were it not for his 8 earned runs in 0.1 of an inning mid-season misadventure.

      Reply ↓
3. Voomo Zanzibar May 11, 2012 at 12:07 am

   Peaches Davis went into the 9th inning with a 14 run lead.
   And a catcher who was 6 for 6.

   http://www.baseball-reference.com/boxes/PHI/PHI193705090.shtml?utm_source=direct&utm_medium=Share&utm_campaign=ShareTool

   Reply ↓
   1. Doug Post authorMay 11, 2012 at 1:02 am

      That’s what a lot of those complete game victories were like. The manager’s attitude seemed to be “as long as we’re winning the game, why take out the pitcher?”.

      Reply ↓
   2. John Autin May 11, 2012 at 11:44 am

      The 14-run lead is nice, but I’m sure Peaches felt the game was in the bag as soon as he saw his opponent would be Losing Pitcher Mulcahy.

      Reply ↓
      1. Doug Post authorMay 11, 2012 at 1:38 pm

         Mulcahy just missed the list. He had 9 disaster starts in 3 straight seasons, 1937 to 1939. That’s 27 DS out of 92 starts, 29.3%.

         The rest of his career, he had only 8 other disasters, but in just 53 starts, 15.1%.

         Reply ↓
4. Voomo Zanzibar May 11, 2012 at 12:17 am

   Sammy Ellis, 22-10 while leading the league in earned runs.
   His Jekyll self had back-to-back outings of 11 and 14 innings each giving up one run.

   http://www.baseball-reference.com/players/gl.cgi?id=ellissa01&t=p&year=1965

   Reply ↓
   1. Doug Post authorMay 11, 2012 at 1:23 am

      What I wonder is whether Ellis’s J&H character had to do with his inexperience. In the previous season (his rookie year), Ellis was used mostly in relief, pitching 128 innings with a 146 ERA+, placing 16th in MVP voting! Then, he goes to 39 starts and 263 innings – a pretty drastic jump.

      The ERA the year before (and even the 99 ERA+ with 11 disaster starts) suggest the kid had some stuff. But that was too much work for a kid in his first season as a starter. He never again had a winning season or an ERA+ over 100, and was done at 28.

      Reply ↓
5. kds May 11, 2012 at 5:49 am

   I guess Urban Faber had red hair, so he got the nickname. We can’t say this is no Shocker. Not an urban legend, in any case.

   Phil Collins must have quit his day job. Whether that was a good thing or not depends upon your taste in music and pitching.

   AFAICT, the Nationals have had no disaster starts yet this year. What are team records for fewest in a year, and latest with first?

   Reply ↓
   1. Doug Post authorMay 11, 2012 at 1:43 pm

      And, Ted sure didn’t blank ’em.

      Reply ↓
   2. John Autin May 11, 2012 at 1:48 pm

      kds — Washington has had one Disaster Start, by Gio Gonzalez on 4/7 (4 R in 3.2 IP):
      http://www.baseball-reference.com/boxes/CHN/CHN201204070.shtml

      But there is one team that hasn’t had one yet — the Blue Jays. (Neil, take a bow.)

      Reply ↓
   3. John Autin May 11, 2012 at 2:04 pm

      The fewest DS this century is 13 by last year’s Phillies.

      For the ’90s (excluding ’94) the low mark was 13 by the ’98 Astros and the ’92 Reds. (The latter is a bit surprising, as only 2 in their rotation had ERA+ over 93. But they had a strong bullpen, so I guess the starters would get yanked quickly if they were giving up runs.)

      For the ’80s (excluding ’81), 10 by the ’85 Mets. Note that’s ’85, not their title year.

      For the ’70s, 12 by the ’71 Angels, another surprise. They had a ridiculously deep bullpen, but a losing record.

      For the ’60s, 13 by the ’65 Yankees. The big surprise is that nobody in ’68 was special in this regard (Mets were low with 16 that year).

      For the ’50s, 16 by the ’52 Phils.

      For the ’40s, 14 by the ’46 Tigers.

      For the ’30s, 17 by the ’34 Giants and ’37 Yanks.

      For the ’20s, 14 by the 1920 Brooklyn Robins.

      And for 1918-19 (with 128- and 140-game schedules), 8 by the 1918 Red Sox.

      Reply ↓
   4. John Autin May 11, 2012 at 2:17 pm

      Too much work to figure out the all-time latest for a team to have its first DS, but for 2002-11 it was the 2010 Rays; their first was in game 45. They wound up with 19, tied for 4th fewest.

      Reply ↓
   5. Doug Post authorMay 11, 2012 at 2:48 pm

      In the live-ball era, fewest disaster starts in a non-shortened season are 10 by the 1985 Mets, followed by the 1986 Mets, 1985 Blue Jays and 1971 Angels with 12.

      The most disaster starts in a season – 63 by the 1939 Browns followed by 61 by the 1930 Phillies and 59 by the 1996 Tigers. Led by Pat Caraway the all-time single season leader, the 1931 White Sox were next with 58.

      It appears the latest a team has gone before recording its first disaster start is game 51 by the 1967 Reds. Billy McCool allowed 4 runs in 2.2 innings on June 3, 1967 but the Reds still won.

      The 2010 Rays had their first disaster start in game 45, but then had 4 more in their next 16 games.

      Reply ↓
      1. Richard Chester May 11, 2012 at 5:28 pm

         I also searched for the latest a team had its first DS and came up with the same result– the 1967 Reds with 51 games.

         Reply ↓
6. RJ May 11, 2012 at 9:16 am

   Jonathan Sanchez frustrated with his inconsistency in San Francisco. In 2008 he had 13 quality starts but 6 disaster starts (19 J&H of 29 total) ending up with a 5.01 ERA and an ERA+ of 88. In fact, whenever Sanchez went six innings that year, it was a quality start.

   Reply ↓
7. Evil Squirrel May 11, 2012 at 10:28 am

   Seeing Bobby Jones’ 2001 season as the most recent entry on the Jekyll and Hyde list makes me thinks of the back to back matchups by Jones and the Cardinals’ Bud Smith that season, where they both made the the Jekyll and Hyde (or I guess it should be Hyde and Jekyll) switch in dramatic fashion….

   August 29, 2001 in St. Louis (I was at this slugfest!):

   http://www.baseball-reference.com/boxes/SLN/SLN200108290.shtml

   Five days later, same two pitchers, this time in San Diego:

   http://www.baseball-reference.com/boxes/SDN/SDN200109030.shtml

   Reply ↓
8. Wine Curmudgeon May 11, 2012 at 11:04 am

   This is the kind of post that makes me what wonder what Baseball-Reference was smoking when they kicked you guys out.

   Reply ↓
9. Voomo Zanzibar May 11, 2012 at 1:34 pm

   Geoff Zahn managed to be 4 games under .500 in a season where he had 5 Shutouts.

   Check out this game:
   http://www.baseball-reference.com/boxes/MIN/MIN198004270.shtml

   Staked to a 10-0 lead in the First inning.
   Couldn’t go 5.
   14 hits in 4.1
   Got relieved with a 17-8 lead

   Reply ↓
   1. Doug Post authorMay 11, 2012 at 2:03 pm

      That’s why Geoff was 4 games under .500 – his manager was taking away his wins by not letting him pitch 5, even with a 9-run lead :).

      Reply ↓
   2. Lawrence Azrin May 11, 2012 at 3:19 pm

      In 1952 Virgil Trucks had two no-hitters, but finished 5-19. BTW, he is uncle to Butch Trucks, founding member of The Allman Bros.

      Reply ↓
      1. Mike L May 11, 2012 at 3:36 pm

         The Tigres were 50-104. Trucks wasn’t that awful-he had an ERA+ of 95 and WAR of 1.7. Yeesh

         Reply ↓
         1. Lawrence Azrin May 11, 2012 at 3:58 pm

            I was not implying that Trucks had a terrible year, merely that a pitcher can have some spectacular starts the two no-hitters) and still be rather ordinary overall for the year.

            Per the example in #16 above – Geoff Zahn in 1980 had five shutouts, which is very impressive, but finished 14-18 in 232 IP, with a 99 ERA+. These are decent numbers, but not as good as you might expect from someone who threw five shtouts, second in the league.

            That is all that I meant.

            Reply ↓
            1. Mike L May 11, 2012 at 7:34 pm

               LA, no offense intended. An observation on how a 5-19 pitcher can be not all that bad. Anyone remember Anthony Young’s run of horrible in 92 and 93. A combined 3-30. In 1993, he had an ERA of 3.77 in 100 inning, but gave up an additional 20 unearned runs. Young ended up a wreck. He was 15-48 for his career, with and ERA+ of 100.
      2. Richard Chester May 11, 2012 at 3:47 pm

         He also had a one-hit shutout, a CG win giving up one run and 6 hits and a win where he pitched 7.2 innings and gave up 2 hits.

         Reply ↓
         1. Richard Chester May 11, 2012 at 3:59 pm

            Trucks and Nolan Ryan are the only two pitchers with two no-hitters and a complete game one-hitter in the same season.

            Reply ↓
         2. Voomo Zanzibar May 11, 2012 at 4:10 pm

            Run Support for Trucks that year: 2.65

            How well did he have to pitch to get a win?
            Here are the batters’ slash lines in those 5 wins:

            .064 .166 .107 .273

            For Virgil that would be:
            0.21 era
            43.2 IP
            9 H

            Reply ↓
   3. Richard Chester May 11, 2012 at 4:49 pm

      There have been 66 instances of a pitcher throwing 5 or more shutouts and finishing with a sub-.500 W-L%. Camilo Pascual, Bullet Joe Bush, Rube Waddell and Christy Mathewson each had a season of 8 shutouts.

      Reply ↓
10. no statistician but May 11, 2012 at 3:51 pm

    I’d be interested in seeing career % of disaster starts for, let’s say, hall of famers, given that it isn’t possible to chart the earlier ones.

    Looking at the second list above, with 40% of the worst performances in the 1930s, I’m also wondering a little about changes from era to era. Given the tendency of managers in the last twenty years or so to pull the starter on almost any pretext, one would expect fewer of these disasters to show up in the records, possibly skewing ERAs lower, WARs higher.

    Reply ↓
    1. Richard Chester May 11, 2012 at 4:12 pm

       I did a quick check. Early Wynn’s DS/GS ratio is 111/612 which is 18.1%. Wynn’s 111 DS, along with Tommy John’s, are the most for the game searchable era. Nolan Ryan clocks in at 103/773 = 13.3%. Red Ruffing is also high, 88/538 = 16.4%.

       Reply ↓
11. Paul E May 11, 2012 at 4:31 pm

    JA:
    I like this Jekyll & Hyde concept-very interesting stuff. You could probably do it on a larger scale with whole seasons for batters and pitchers. You know, 120 OPS+ or ERA+ followed up with a sub 100 season…where’s Lon Chaney when you need him

    Reply ↓
12. Hartvig May 14, 2012 at 10:53 am

    In a normal run scoring environment a starter with an ERA around 4, and an ERA+ in the mid to high 90’s is usually best suited to maybe 4th starter status on a decent team. If a starter gives up 4 or more runs for a team with an average offense in a normal run scoring environment he’s going to normally lose the vast majority of those games. Does anyone know: has there ever been a game-to-game Jekyll & Hyde type pitcher at least for the majority of his career? I’m thinking of someone who gives up 2 runs or fewer in close to two-thirds of this starts but usually 6 runs or more in the remainder? I assume that in most seasons that would give him 4th-starter type peripherals but with a better than expected winning record?

    I’m also surprised that in an article about Jekyll and Hyde pitching performances and all the comments that there was no mention of probably the pitcher best known for up and down seasons in my direct baseball memory (going back to around 1960): Bret Saberhagen (I know some of his issues were injury related but still…).

    Reply ↓