Leads Lost by Top Starters, 1954-2008
Definition of a lead lost:
A pitcher is considered responsible for losing a lead if:
1. he is still in the game when the tying runs scores, OR
2. he leaves the game with the tying run in a position where it can score without a hit - e.g on second with 0 out, on third with 0 out or 1 out.
I’ve identified 38 starting pitchers who are either in the Hall of Fame or who are considered to be HOF-worthy, who pitched between 1954 and 2008. I don’t have complete records for those pitchers who were active in the first part of that era (i.e. Robin Roberts, Warren Spahn, Whitey Ford) because Retrosheet doesn’t go back to the beginning of their careers and doesn’t have full event data for the 50s and 60s.
Player Starts HadLead PctHad AvgLeadHeld LostLead PctLost AvgLeadLost
Roberts R 395 251 63.5% 2.29 133 53.0% 1.38
Niekro P 710 491 69.2% 2.41 240 48.9% 1.40
Perry G 690 468 67.8% 2.37 224 47.9% 1.40
Morris J 527 385 73.1% 2.54 182 47.3% 1.41
Smoltz J 466 341 73.2% 2.36 157 46.0% 1.37
Hunter J 476 340 71.4% 2.47 155 45.6% 1.25
Blyleven B 685 473 69.1% 2.51 214 45.2% 1.37
Bunning J 516 346 67.1% 2.47 156 45.1% 1.34
Kaat J 625 409 65.4% 2.53 184 45.0% 1.39
Spahn W 357 272 76.2% 2.47 122 44.9% 1.39
Carlton S 709 493 69.5% 2.58 219 44.4% 1.42
Glavine T 682 467 68.5% 2.54 207 44.3% 1.34
Stieb D 412 288 69.9% 2.40 127 44.1% 1.32
Drysdale D 463 311 67.2% 2.58 136 43.7% 1.32
Gooden D 410 292 71.2% 2.50 127 43.5% 1.31
Sutton D 756 499 66.0% 2.42 215 43.1% 1.38
Schilling C 436 315 72.2% 2.55 134 42.5% 1.34
Koufax S 314 238 75.8% 2.50 101 42.4% 1.34
Johnson R 586 426 72.7% 2.55 180 42.3% 1.37
Ford W 396 289 73.0% 2.55 122 42.2% 1.27
Tiant L 484 356 73.6% 2.49 150 42.1% 1.29
Cone D 419 299 71.4% 2.58 125 41.8% 1.33
Jenkins F 594 414 69.7% 2.52 173 41.8% 1.33
Mussina M 536 390 72.8% 2.64 162 41.5% 1.37
Marichal J 457 338 74.0% 2.62 139 41.1% 1.31
Brown K 476 331 69.5% 2.40 136 41.1% 1.38
John T 700 479 68.4% 2.33 196 40.9% 1.31
Ryan N 773 485 62.7% 2.32 196 40.4% 1.36
Guidry R 323 260 80.5% 2.57 105 40.4% 1.32
Maddux G 740 521 70.4% 2.40 208 39.9% 1.35
Appier K 402 256 63.7% 2.56 101 39.5% 1.26
Seaver T 647 439 67.9% 2.42 171 39.0% 1.29
Saberhagen B 371 248 66.8% 2.47 96 38.7% 1.34
Perry J 447 293 65.5% 2.60 113 38.6% 1.49
Gibson B 482 335 69.5% 2.58 129 38.5% 1.30
Palmer J 521 376 72.2% 2.68 142 37.8% 1.28
Clemens R 707 520 73.6% 2.53 180 34.6% 1.32
Martinez P 400 295 73.8% 2.60 92 31.2% 1.26
Median 69.8% 2.52 42.2% 1.34
Columns are:
Number of starts included
Number of starts in which the pitcher held a lead
Percentage of starts in which the pitcher held a lead
Average size of lead held in those starts
Number of starts in which the pitcher gave up a lead
Percentage of starts in which the pitcher held a lead and gave up a lead
Average size of leads given up
I listed the medians for the group at the bottom.
There is a correlation between the average lead size and the percentage of leads lost in the expected direction (r=-0.33).
Morris, Koufax, and Pedro jump out at me here, alonbg with Smoltz and Glavine.. I didn’t check Sandy’s numbers for 1962-1966 alone, and I probably should. I don’t know how much of Pedro’s remarkable ability to hang on to leads was a function of his managers getting him out of games at the first sign of trouble, but basically, give him a two-run lead and the opposition might as well go home.
Ryan and Appier are the guys who should really sue for non-support, based on the number of starts in which they *never* had a lead. Guidry, on the other hand, perhaps should have won a lot more often than he did, although he did OK with the leads that he did have.
As requested below, here is the same list, including only starts in which the pitcher held or lost a lead at the end of a complete inning:
Player Starts HadLead PctHad AvgLeadHeld LostLead PctLost AvgLeadLost
Roberts R 393 246 62.6% 2.29 117 47.6% 1.33
Morris J 524 378 72.1% 2.54 162 42.9% 1.37
Niekro P 707 481 68.0% 2.40 204 42.4% 1.35
Smoltz J 466 338 72.5% 2.34 141 41.7% 1.34
Perry G 685 459 67.0% 2.37 191 41.6% 1.35
Hunter J 474 333 70.3% 2.44 138 41.4% 1.26
Gooden D 406 285 70.2% 2.48 118 41.4% 1.32
Bunning J 513 339 66.1% 2.45 137 40.4% 1.28
Glavine T 679 460 67.7% 2.53 185 40.2% 1.35
Blyleven B 683 457 66.9% 2.50 183 40.0% 1.30
Carlton S 707 484 68.5% 2.57 193 39.9% 1.38
Mussina M 534 384 71.9% 2.60 153 39.8% 1.34
Schilling C 436 314 72.0% 2.51 125 39.8% 1.34
Stieb D 410 278 67.8% 2.39 108 38.8% 1.25
Ford W 391 280 71.6% 2.53 108 38.6% 1.20
Kaat J 618 394 63.8% 2.51 151 38.3% 1.28
Johnson R 586 421 71.8% 2.53 161 38.2% 1.32
Jenkins F 592 406 68.6% 2.52 155 38.2% 1.29
Marichal J 455 336 73.8% 2.60 128 38.1% 1.30
Sutton D 750 486 64.8% 2.40 185 38.1% 1.31
Maddux G 738 512 69.4% 2.39 194 37.9% 1.32
Cone D 419 288 68.7% 2.54 109 37.8% 1.31
Brown K 473 323 68.3% 2.40 121 37.5% 1.36
Seaver T 646 435 67.3% 2.41 162 37.2% 1.28
Drysdale D 460 305 66.3% 2.55 113 37.0% 1.26
Spahn W 354 270 76.3% 2.45 100 37.0% 1.32
Ryan N 761 476 62.5% 2.30 174 36.6% 1.28
Gibson B 480 331 69.0% 2.56 120 36.3% 1.28
Guidry R 323 248 76.8% 2.57 89 35.9% 1.29
John T 693 468 67.5% 2.27 167 35.7% 1.25
Appier K 398 251 63.1% 2.52 89 35.5% 1.21
Saberhagen B 369 243 65.9% 2.45 85 35.0% 1.34
Tiant L 480 344 71.7% 2.47 120 34.9% 1.20
Clemens R 705 505 71.6% 2.51 171 33.9% 1.31
Palmer J 516 367 71.1% 2.66 122 33.2% 1.18
Koufax S 305 226 74.1% 2.50 74 32.7% 1.21
Perry J 439 279 63.6% 2.60 84 30.1% 1.43
Martinez P 399 292 73.2% 2.59 86 29.5% 1.24
Median 68.7% 2.51 38.1% 1.30
Mike Emeigh
Posted: February 17, 2009 at 01:20 PM |
28 comment(s)
Login to Bookmark
Related News:
Reader Comments and Retorts
Go to end of page
Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.
1. Mike Emeigh-- MWE
So this is saying that both Morris and Blyleven were among the worst at holding leads?
It also shows that 2 HOF'ers held leads at a worse % than Blyleven, and he is within 2% pts of 5 other HOF'ers.
I'd say this bolsters his case.
Question: can a pitcher blow a lead and hold a lead in the same game? Pitcher given a 1-0 lead in the first inning, gives up 2 runs in the third, blowing the lead, then his team scores a few more runs while he's still in the game and ends up winning 6-2. Is this counted as a lead blown or lead held or both? Size of lead I assume is max lead attained.
I'd also like to see the numbers using only starts where the pitcher finished his last inning (whether replaced by a RP the next inning or finishing the game). It's not obvious to me that the runners on situations are a fair comparison across run environments and bullpen usage patterns.
Home and road breakdowns seem like they would be interesting too, perhaps not for individual pitchers but certainly for the group. Obviously more likely to have a lead on the road since your team bats first, but distribution of lead size might also differ and affect the results.
I suspect that both tended to stray in games far longer than is common for starters nowadays. Especially early in their careers. There's a lot of anectdotal testimony by Twinkies fans that Blyleven was basically left in games come hell or high water- and often blew leads late for that reason.
Hence this may say as much about the quality of the bullpen and the manager's proclivities as it does about the pitcher himself.
Ideally, wouldn't you want to break this down by game situation (Inning & lead to start inning, I guess) and then compare each pitcher's performance in that "cell" to the avg. SP. Sort of how the PBP defensive metrics do it?
Not the way that I count it. The Had Lead column contains games in which he HAD a lead, at any point, while he was in the game; the Lost Lead column contains games in which he gave up a lead, at any point; you only get credit for doing either one once in a game. I have more detailed breakdowns by inning, which I'll post later (for now, I'll simply say that the anecdotal evidence on Blyleven is wrong). The average lead held is the average lead held at any point in the game - if you pitched five innings in a game with a lead of 1,1,1,2, and 3 runs at the start of each inning, you get an average lead held for that game of 1.6 runs. The average lead lost is the same - if you lost a lead of 1 run and 2 runs in the same game, you get an average lead blown of 1.5 for that game.
I didn't see any particular pattern of era bias which would have led me to do this, but I'll run the numbers based only on complete innings and report on it.
-- MWE
-- MWE
I distinctly rememberthis one from listening to it on the radio.
It would have been even more amazing if it had happened a year later :)
-- MWE
Two thoughts about that:
(1) That was an awfully slow hook. Letting the Giants bat all the way around and more before pulling the plug? But it was a slow hook because he was Bob Gibson and he was already legendary. You just didn't do that to him.
(2) He wasn't having a very good year in 1967 before Clemente broke his leg. In his 175.1 IP through July, his ERA was just 3.52, against a park-adjusted league average of 3.27 for the year. That's an ERA+ of 93. Sure, it would look better - up to about 111 - if you were to remove the June 29 game from the record. But June 29 did happen, and even so, that's just 111.
I'll claim that Gibson's 1967 injury recovery was the most amazing injury recovery of any baseball player ever. Does anyone have anything to top it? He was out from July 15 to Sept. 7, which is just 7 weeks. That seems pretty quick for a broken bone. But as I've said above, he left as a famous pitcher having a sub-par, even mediocre season. When he returned to the lineup, he returned as a fire-breathing monster. The way to put it is that his "1968," which is what Mike is referring to in his quote, started in Sept. 1967.
In 37.1 IP the rest of the 1967 regular season, his ERA was 0.96.
In 27 IP in the WS (three complete games), his ERA was 1.00. Without him, the Cardinals don't win that Series.
In 300+ IP in the 1968 regular season, his ERA was ... well, you know about that one.
And his 1969 was the second best year of his career.
When I did the by-whole-innings numbers, they were coming out wrong, so I thought. Then I went back and looked at what I did for the originals, and realized that instead of taking the "average" lead I was taking the "maximum" lead in each game. It's now been fixed; the average lead held and the average lead lost are now as I stated they should be, above.
-- MWE
Blyleven had 16 starts in which the only time his team took a lead occurred just before he came out of a game mid-inning. That was the highest number of such starts. By contrast, Schilling had just one such start, Spahn and Drysdale had 2.
Ryan had 12 starts in which he did not complete at least one inning; no one else had more than 9. I missed David Cone and Kevin Brown earlier in my list of pitchers who never came out before the end of the first inning, so it's not quite as rare as I thought it was. Seaver had one, Gibson 2. The modern pitchers do have an advantage here, because teams are far less likely to remove a pitcher in the first than they used to be (this coincides with the disappearance of the long reliever).
Koufax looks a LOT better when you take partially completed innings out of the mix. I need to take a closer look at his bullpen support.
Gibson's other early exit occurred in this game, in which he was handed a six-run lead early and gave it all back, although three of the runs scored after Broglio came into the game, so it's Broglio who gets the "blame" for losing the lead, as I define it. It was Gibson's first start of the season.
-- MWE
Thanks for doing the whole inning thing. That'll be interesting to see. I think, though, that what I was getting at might be better expressed as taking out those games where the lead blown is definition (2): he leaves the game with the tying run in a position where it can score without a hit - e.g on second with 0 out, on third with 0 out or 1 out.
If a pitcher gives up the tying run, and is then pulled mid-inning, it doesn't matter since he would certainly still have blown the lead if allowed to finish the inning. However, in the blown lead definition (2) it's not clear whether he would actually have blown the lead or not if left in (or whether the reliever blew the lead or not). A team with a good fireman might be tempted to put in the fireman even though the starter is still likely to escape with the lead (say 60% chance of the fireman keeping the lead, 50% chance the starter could have kept the lead, yet the starter is given a blown lead if the fireman comes in). Whereas a team with a crappy bullpen (say 40% change of keeping the lead) might leave the starter in, and he gets a blown lead only 50% of the time, resulting in fewer blown leads even though he's no better a pitcher, and his team is actually *more* likely to blow the lead.
Possibly, non-blown leads where the pitcher exited mid-inning would also have to be excluded from the denominator -- since had the pitcher been left in he could have given up more runs (e.g. the discussion about Blyleven) -- but I'm not sure whether it's all such situations that should be removed (even a 10 run lead seems extreme), a slightly looser criteria that you use for definition 2 (e.g. tying run could score with a single and outs, or tying run on base), or if this is too hairy and the cleaner definition of complete innings is the best way to do it. I'd need to put some more thought into what's appropriate.
I remember that Ryan had persistent blister problems on his pitching fingers, and this was mostly in his days with the Mets. Do the short outings cluster in that time, and are some of them games in which the box score/game recap doesn't make it obvious why he would have been pulled?
They do not cluster. Two were with the Mets, five were with the Angels, two were with Houston, and three were with Texas. Ryan was ineffective in nine of the 12. In the other three, this one was the result of a mistake by Dick Williams in filling out the lineup card, this one was likely due to an injury (Ryan made just one more start that season, lasting just 1 1/3 in that one), and this one was due to a strained hammy.
-- MWE
While I don't know how I'd do it, this seems odd. A 1-run lead held for 5 innings is the same "average lead" as a 1-run lead held for 1 inning. I know, a lead lost is a lead lost in some sense, but one of those pitchers has done a really good job while the other hasn't. I'm not sure it will make a difference in the results really (I suspect it's pretty random), I just find it a strange way to conceptualize it.
I think I would probably just rather see it broken down by size of lead lost (i.e. % of 1, 2, 3 run leads lost) though that too ignores how long they held the lead.
Since you've got the data, maybe it would be more interesting (to me) to look at "average number of innings they held a 1-run lead; average for two-runs" etc?
it's more variance than I would have expected.
I gather that comment was made for Mike's old numbers which I never saw. These numbers don't show much variance really. 42.2% is the median but I'm too lazy to figure the mean and so will treat it as the mean. Assuming a binomial distribution and 350 leads, which should be about the median of that, by random chance you'd expect a relative standard error of about 2.6% giving us a 95% CI of about 37% to 47.4%. By random chance you'd expect 2 of those 40 pitchers to be outside that CI and instead we get 5 -- so there's probably something going on but it's not far off random.
The binomial might not be right -- but then if we had the correct one we might find it looks even more random. Using actual # of starts by pitcher to see how many fall outside the 95% CI for their # of leads would be better and maybe that would show more variation than my quickie above.
It's also interesting (odd? inexplicable?) that, near as I can tell, these numbers have not been adjusted for run-scoring environment yet there doesn't seem to be any relationship between the average lead, the average lead lost or the lost %age and whether they've pitched during the 90s-00s. But again, I haven't actually calculated it.
But I'm kinda unimpressed with the results (not Mike's work, the pitchers'). The best pitchers give up 42% of their leads. And given the average lead given up is much lower than their average lead, one assumes they must be giving up at least 50% of their 1-run leads. My guess is that the size of the lead in any inning has rather small impact on the pitcher's likelihood of giving up a run or the expected number of runs (except maybe when they have a big lead).
I would guess (and am probably wrong) that HR pitchers (Roberts, Blyleven and Jenkins at least from the above list) would be more likely to give up a big lead. But maybe not as they tended to compensate for HR with Ks and few BBs.
And it cracks me up that Morris & Smoltz are right next to each other yet are probably both best-known for a 1-0, extra-inning WS game. Clearly the key is that neither had a lead to protect except for 1 inning. :-)
The pitcher's team is visiting, he pitches the whole game and the scores are as follows.
V 1 0 1 0 0 0 0 0 0 = 2
H 0 1 0 0 0 0 0 0 0 = 1
My reading of your method is that this would count as a game where the pitcher had a lead and lost it. Seems to me it would be better to count this as a case where the pitcher held a lead. The held lead is more important to the outcome.
I agree with Walt that it would be nice to see how long the leads were held.
That's actually not the way to read it. The best pitchers give up a lead *at least once* in 42% of the starts in which they had a lead. I think what was a more surprising finding for me is that even the best pitchers never had a lead in 30% of their starts.
Essentially, this is true. Most pitchers, from what I can tell, have their worst results when they have a big lead than they do at any other time in the game - and the later in the game, the worse they tend to do. Blyleven is the one case that ran counter to the general trend that I have found so far; he tended to pitch better when he had a big lead.
Makes sense.
Most leads are lost early in games, as you might expect - primarily because early leads are typically smaller. What I want to see is the distribution for 1-run and two-run leads across innings.
FWIW: Blyleven's rough patches tended to occur in the middle innings. I think he got the rep of losing leads late because he lost a lot of small ones, and people remember those; he had a higher percentage of leads lost than the norm in the late innings but a smaller average lead lost. (He also had a lot of BIG leads late.) This is why I need to look at the distribution of leads across innings.
-- MWE
Fair enough. Still given the %age and the fairly large gap between "average lead" and "average lead lost" that they gave up at least once a quite high percentage of their 1-run leads.
I guess what I would want to see would be # of innings started with a 1-run lead, # of times that lead was "lost" (by your definition or another) in that inning. Repeat for 2-run leads, etc. Do a run sub-setting to inning 6 or later. I'm assuming your data could support that analysis though that might be a pain to program.
And I think that's closer to what you want to get at ... or maybe I misunderstood your motivation as it's been expressed in the occasional post. I thought you were arguing from the perspective of "if a pitcher has a lead, his job is to hold it." And while I appreciate you're thinking (I think) along the lines of "by golly, if he has to hold it for 5 innings and he doesn't, then he failed" I still think it's just cleaner to view it on an inning-to-inning basis. Strictly speaking, the pitcher only knows how big a lead he needs to protect through this upcoming inning -- he doesn't know if his team is going to score again when they come to the plate.
I still think it will show things are pretty random and the size of the lead doesn't have a lot of impact on how well a pitcher pitches.
I actually have all of that. One of the joys of Retrosheet event data - once you get it into a database engine - is that it becomes relatively easy to slice and dice the data. I created a table in my database to expedite this:
CREATE TABLE `retrosheet`.`defensiveinnings` (
`Game_Date` date NOT NULL,
`DH_Game` int(1) unsigned NOT NULL,
`Visitor` varchar(3) NOT NULL,
`Home` varchar(3) NOT NULL,
`DefTeam` varchar(3) NOT NULL,
`Inning` int(2) unsigned NOT NULL,
`MinEvent` int(3) unsigned DEFAULT NULL,
`MaxEvent` int(3) unsigned DEFAULT NULL,
`CloseGame` char(1) DEFAULT NULL,
`PitcherAtStart` varchar(8) DEFAULT NULL,
`StarterInGame` char(1) DEFAULT NULL,
`LeverageAtStart` decimal(4,2) unsigned DEFAULT NULL,
`LeadAtStart` varchar(1) DEFAULT NULL,
`TrailAtStart` varchar(1) DEFAULT NULL,
`TieAtStart` varchar(1) DEFAULT NULL,
`DiffAtStart` int(3) DEFAULT '0',
`PitcherAtEnd` varchar(8) DEFAULT NULL,
`StarterStillPitching` char(1) DEFAULT NULL,
`LeadAtEnd` varchar(1) DEFAULT NULL,
`TrailAtEnd` varchar(1) DEFAULT NULL,
`TieAtEnd` varchar(1) DEFAULT NULL,
`DiffAtEnd` int(3) DEFAULT '0',
`BlownLead` varchar(1) DEFAULT NULL,
`BlownLeadEvent` int(3) unsigned DEFAULT '0',
`RespPitcher` varchar(8) DEFAULT NULL,
PRIMARY KEY (`Game_Date`,`DH_Game`,`Home`,`DefTeam`,`Inning`)
)
Most of the field names are self-explanatory. MinEvent and MaxEvent are pointers into the play-by-play event table that set the bounds for the inning.
I still have that perspective.
Agreed.
Well, there are two aspects to this. You need to look at both the distribution and whether individual pitchers vary from the distribution. The distribution itself doesn't appear to be random, but individual pitchers (so far, anyway) don't seem to have a lot of variance from the distribution.
-- MWE
Just curious: how do other, "lesser" starting pitchers compare from this era? 42% leads blown certainly seems high for a HOF-type group, but maybe that figure blows Jim Deshaies' percentage out of the water? (Although one voter apparently thought Deshaies was HOF-worthy...)
If he is, then of course we can rule out the bullpen. If he is, and he's "blown" a lead ... well, isn't that characteristic of a bad team?
You must be Registered and Logged In to post comments.
<< Back to main