Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Baseball Newsstand > Baseball Primer Newsblog > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Tuesday, January 27, 2009

Daneeka’s Ghost: Bullpen Usage

A look at the winning percentage for the home team with a one-run lead entering the eighth and ninth innings for every season from 1977 through 2006. 

1977 - 1989 (over 40% SV = 1+ IP)  -  9 of 14 seasons above average at preserving 1-run leads in the ninth. (2561 out of 2925 - 87.56% of leads preserved)

1990-2006 (over 40% SV = 1 IP)  -  4 of 16 seasons above average at preserving 1-run leads in the ninth. (3133 out of 3651 - 85.81% of leads preserved)

So, why has the most current bullpen setup been subpar in holding those leads?

WillYoung Posted: January 27, 2009 at 07:43 PM | 42 comment(s) Login to Bookmark
  Tags: sabermetrics

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. Barry`s_Lazy_Boy Posted: January 27, 2009 at 08:00 PM (#3061796)
Higher run scoring environment.
   2. DCA Posted: January 27, 2009 at 08:24 PM (#3061825)
Man, there were a lot of blown leads in 1994. What happened in 1994? Expansion, shortened season, first year of "new" higher run environment.

If it's the run environment, and that makes sense, it's interesting that there's no real spike in 1987. If it's expansion, it's interesting there's no similar effect in 1998. 1994 could just be a fluke which just happened to be timed with three interesting events.
   3. Sheer Tim Foli Posted: January 27, 2009 at 08:53 PM (#3061852)
Which book had an article showing that closers have not done anything to statistically save games? The Book? Thinking Fan? I am not sure what value I get from reading all these only to not remember the facts or where I got them from.

Anyway - there has been some good math questioning the value of the closer role.
   4. Walt Davis Posted: January 27, 2009 at 10:03 PM (#3061911)
The formula for the standard error of the difference in proportions between two independent samples is:

sqrt [ p1(1-p1)/n1 + p2(1-p2/n2 ]

These two samples probably aren't technically independent (due to some relievers carrying over between periods and possibly other factors) but any covariance is likely small. Assuming independence, I get a standard error estimate of .008. The estimated difference is .0175 so this is significant at the .05 level and we would reject the null hypothesis that the two rates are the same. But the difference isn't so much greater than what we might get by random chance that we'd be betting the house -- especially given the change in run-scoring environments.

In short, I'm not convinced there is any difference to explain after you control for run-scoring environment.
   5. GuyM Posted: January 27, 2009 at 10:37 PM (#3061933)
I think it's mainly or entirely the run environment. Break up the data in smaller segments, and you get this:
1977-1985: 87.3%
1986-1993: 86.9%
1994-2000: 85.5%
2001-2006: 86.1%

The one big (kinda big) drop is pre/post-1994, when offense surged. The change from 77-85 to 86-93 is tiny, despite a huge increase in 1-run saves. And preservation of leads actually increased after 2000, even though 1-run saves continued to increase.
   6. Steve Treder Posted: January 27, 2009 at 11:14 PM (#3061956)
In short, I'm not convinced there is any difference to explain after you control for run-scoring environment.

Which means that the full-scale redesign of bullpen/pitching staff usage modes over the past 30 years, whatever else it may have achieved, has had virtually no significant impact on late-inning 1-run lead preservation. Making the enormous focus on the save stat over the same period, well, rather amusing.
   7. snapper (history's 42nd greatest monster) Posted: January 28, 2009 at 12:14 AM (#3061986)
Which means that the full-scale redesign of bullpen/pitching staff usage modes over the past 30 years, whatever else it may have achieved, has had virtually no significant impact on late-inning 1-run lead preservation. Making the enormous focus on the save stat over the same period, well, rather amusing.

To add to this, they've increased the number of relievers on the roster from 5 to 7, greatly restricting bench flexibiltiy.
Also, I would guess the fraction of payroll spent on the bullpen has gone up as well. All to no avail.
   8. Steve Treder Posted: January 28, 2009 at 12:29 AM (#3061990)
To add to this, they've increased the number of relievers on the roster from 5 to 7, greatly restricting bench flexibiltiy.
Also, I would guess the fraction of payroll spent on the bullpen has gone up as well. All to no avail.


Roger that.
   9. GuyM Posted: January 28, 2009 at 01:03 AM (#3062000)
Steve/Snapper: You can't tell how valuable the bullpen usage has been without knowing how much the increased run environment would reduce lead preservation. You'd have to calculate a win expectancy table for the pre-1993 environment and compare that to today's game.

It's clear that modern usage wins games. If team scoring were random, a team with a 1-run lead after 7 innings would win 73.4% of the time (according to The Book). In fact, they've won 76.6% in the 2000-2006 period. I don't know how that +3.2% compares to earlier periods, but that's a key question. And we do know for sure that today's short-inning relievers are MUCH more effective at preventing runs than the manly relievers of the 1960s and 70s.
   10. Steve Treder Posted: January 28, 2009 at 01:10 AM (#3062003)
It's clear that modern usage wins games. If team scoring were random, a team with a 1-run lead after 7 innings would win 73.4% of the time (according to The Book). In fact, they've won 76.6% in the 2000-2006 period. I don't know how that +3.2% compares to earlier periods, but that's a key question.

Well, if that's a key question, than how is it so clear that modern usage wins games?

And we do know for sure that today's short-inning relievers are MUCH more effective at preventing runs than the manly relievers of the 1960s and 70s.

We do know that pitchers are much more effective in run prevention working shorter stints than longer stints. What isn't so clear is to what degree the addition of a 15-20% additional marginal pitchers into the mix, pitchers who wouldn't have had major league jobs under the earlier model, have had as a net effect; obviously they're better working short than they would be working long, but what of the issue of their intrinsic relative quality? And what, of course, of the non-insignificant issue of the sacrifice of position player roster spots to support the expanded bullpen?

The point is that it's complicated; it isn't clear.
   11. philly Posted: January 28, 2009 at 01:44 AM (#3062015)
Which means that the full-scale redesign of bullpen/pitching staff usage modes over the past 30 years, whatever else it may have achieved, has had virtually no significant impact on late-inning 1-run lead preservation. Making the enormous focus on the save stat over the same period, well, rather amusing.


Doesn't it also make the ewqually enormous complaining about the save stat and modern bullpen usage also amusing?
   12. Steve Treder Posted: January 28, 2009 at 01:51 AM (#3062017)
Doesn't it also make the ewqually enormous complaining about the save stat and modern bullpen usage also amusing?

I don't know. What's more amusing: fascination and celebration of achievement in a statistical category which has no apparent relationship with the increased achievement of wins, or enormous complaining about the fascination and celebration of achievement in a statistical category which has no apparent relationship with the increased achievement of wins?
   13. GuyM Posted: January 28, 2009 at 02:31 AM (#3062027)
Well, if that's a key question, than how is it so clear that modern usage wins games?


As I said, it's clear that it wins games compares to random usage of relievers, regardless of whether you have a lead or trail. What's not clear from these conversion rates alone is how much better it is at preserving leads than prior usage patterns.

Keep in mind that there's a limit to how much difference a league-wide strategy can possibly have in terms of wins. The team with a lead uses short relievers, but often the opponent does as well. So it's hard to see the gain. But it may still be true that pushing your relievers to get 7-8 outs, or pushing your starters to go 7 IP, would result in many fewer wins. (We can't see that easily, because no team in today's game has yet made the mistake of following your advice).

"The point is that it's complicated; it isn't clear."

It's not that complicated. The third time through a lineup, a starter gives up about 13% more runs (+0.60). In contrast, a short reliever gains about a one run advantage. The combination is incredibly powerful. In the 7th inning, an average pitcher coming in for one inning becomes Brandon Webb. There is simply no question that, in general, the modern usage pattern is effective. Would Tony LaRussa be better off with 1 less LOOGY and one more position player? Maybe so. But the general pattern of shorter outings for starters and shorter outings for relievers is unquestionably the right way to go.
   14. Wes Parkers Mood (Mike Green) Posted: January 28, 2009 at 02:58 AM (#3062034)
You'd think that with the decline in run-scoring over the last few years that bullpen usage would at least move towards the patterns of 1993 (i.e with the six man bullpen being common). Instead, the average length of relief outings continues to shorten. The run-scoring environment was comparable in 2008 to that of 1993, but the average relief outing over that span decreased from 1.24 innings per appearance to 1.07.

I am not complaining about the save stat and modern bullpen usage. I just wish that one team would have the guts to act as if the "modern" way is a horribly inefficient use of personnel and roster spots, and to simply not play along.
   15. Walt Davis Posted: January 28, 2009 at 03:24 AM (#3062041)
The third time through a lineup, a starter gives up about 13% more runs (+0.60). In contrast, a short reliever gains about a one run advantage.

Generally speaking, you need to go through the lineup three times just to get through 6 innings. Short relievers ain't got nothing to do with that really.

Maybe you meant 4th time through the lineup.
   16. Steve Treder Posted: January 28, 2009 at 05:21 AM (#3062083)
it's clear that it wins games compares to random usage of relievers

Well, great, but random usage of relievers has never been employed in any league since the invention of baseball.

the general pattern of shorter outings for starters and shorter outings for relievers is unquestionably the right way to go.

Absolutely. Except, of course, for the questions.
   17. Steve Treder Posted: January 28, 2009 at 05:27 AM (#3062086)
I think either one requires an awfully low Amusement Threshold.

You need to find more enthrallment in the inherent delightfulness of being. Duh.
   18. GuyM Posted: January 28, 2009 at 10:13 AM (#3062124)
Maybe you meant 4th time through the lineup

Actually, 3rd or more time through the lineup. The point is, by the 6th or 7th inning, your starter is at a huge disadvantage compared to a reliever. And of course teams use short relievers that early in today's game (as well as in the 8th and 9th).
   19. GuyM Posted: January 28, 2009 at 03:16 PM (#3062257)
Looking some more at this data, there appear to be two things going on: 1) a one-time shock in 1993/1994, where the surge in offense reduced preservation rates dramatically, and 2) a general tendency toward improved preservation, presumably resulting from change in bullpen usage. Here’s the preservation rates for 1 run lead, top of 8th:
1977-84 76.5%
1985-92 79.7
1993-94 72.1
1995-00 75.9
2001-06 76.8
Note the solid improvement in the 1985-92 period, as one-inning saves surged. Increased offense reduced the preservation rate by 7-8% in 1993-94, but then preservation rates resumed their climb and are now back to the 1977-84 level despite higher scoring rates. What’s interesting is how fast preservation rates moved back up. I wonder what teams did in 1995-96 to respond? Candidates include:
Increased use of high-quality set-up men in the 8th;
Shorter appearances by set-up men and closers;
More pitcher changes to gain platoon advantage;
More defensive replacements in late innings;
Some/all of the above.

Steve: if you’re out there, are you aware of any significant tactical changes along these lines in the 1995-96 period?
   20. Mike Green Posted: January 28, 2009 at 03:41 PM (#3062283)
The logical consequence of the deterioration in performance by starters after about 75 pitches is to move to increased use of tandem starters. Roy Halladay or Brandon Webb can deteriorate and still be valuable after the 5th inning. Jon Garland or Jarrod Washburn cannot. Recognizing the importance of a fresh arm would logically lead to longer average relief outings than shorter ones. Ideally, these should be scheduled.

The modern bullpen use is driven not by the understanding that GuyM is describing, but rather by the vagaries of the save role and a belief that gaining the platoon advantage is exceptionally important in a game which can be saved.
   21. TomH Posted: January 28, 2009 at 03:47 PM (#3062287)
Is the "3rd or 4th time thru the lineup" decreaase in effectiveness a modern thign, or was it true in the 60s when pitchers were men? :) I mean, if Gibson/Koufax/Marichal et al really were worse in the 8th inning back then and no one noticed, that's a major point. But it could be that guys today aren't effective in the 8th because they don't do it much (of course it could also be more related to pitch count, and obvioously today's 7th inning is more like the 8th or 9th of many years ago). I hadn't seen the study that was ref'd by GuyM in #13, so I don't know the answer. But it seems important enough that somebody oughta.
   22. Steve Treder Posted: January 28, 2009 at 04:42 PM (#3062347)
Steve: if you’re out there, are you aware of any significant tactical changes along these lines in the 1995-96 period?

No. Frankly I think you're making way too much of year-to-year fluctuations that are more likely just random perturbation than anything meaningful.
   23. GuyM Posted: January 28, 2009 at 04:42 PM (#3062348)
Recognizing the importance of a fresh arm would logically lead to longer average relief outings than shorter ones.

No, it would lead logically to more innings thrown by relievers (and fewer by starters), which is exactly what has happened. In addition, there is evidence that relievers are more effective over 1 IP than over 2+ IP, creating an incentive for shorter outings by relievers (which has also happened). Short version: the fewer batters a pitcher has to face, the better he is (mainly because of increased K rate).

TomH: "The Book" by Tango/Lichtman/Dolphin has analysis of the times-thru-lineup effect, and the B-Ref split page shows it as well. And Tango has written a lot at his blog on the pitch count issue. I believe his data indicates that -- as you suggest -- today's pitchers throw about as many pitchers as the he-men of yesteryear. In terms of comparison to the past, I took a quick look at the mid-70s (1975-77) and it looks like pitchers' OPS+ for batters' 3rd+ PA was about 107, compared to around 112 today. Of course, the 1970s pitchers faced a lot more of those batters, which impacts their total OPS, so that serves to narrow the spread a bit. Still, it looks like the performance dropoff may have increased somewhat. I think that's a function of the greater need for strikeouts to succeed in today's game (the biggest advantage of pitching fewer innings/batters is a higher K rate), rather than what pitchers are "trained" to do.
   24. GuyM Posted: January 28, 2009 at 04:49 PM (#3062354)
No. Frankly I think you're making way too much of year-to-year fluctuations that are more likely just random perturbation than anything meaningful.


C'mon Steve, I'm not looking at y-t-y changes -- I'm comparing 6-year samples. The exception is 1993-94, which I've isolated because of the huge change in scoring. Do you really think the plunge in lead preservation rates then is a coincidence? It's entirely consistent with what we know about win probability: according to The Book, increasing RS by .9 runs decreases the chance of throwing a scoreless inning by about 6%.
   25. Steve Treder Posted: January 28, 2009 at 04:59 PM (#3062366)
C'mon Steve, I'm not looking at y-t-y changes -- I'm comparing 6-year samples. The exception is 1993-94

Well, it isn't a 6-year sample, is it.

which I've isolated because of the huge change in scoring. Do you really think the plunge in lead preservation rates then is a coincidence?

Maybe it is or maybe it isn't, but being as it is a 2-year sample, I'm not confident in concluding much of anything about it, especially since in the years following the rate reverts back to virtually exactly where it was in 1977-1984. And overall, outside of the 1993-94 exception, the rate changes over the 30-year period -- a very long time -- are so miniscule as to suggest that the larger point is that the major modification of bullpen/pitching staff construction has had no meaningful impact on late-inning lead preservation. Whatever else it's done, it hasn't impacted that factor, and this, I think, is the big deal here, not the 1993-94 blip.
   26. Mike Green Posted: January 28, 2009 at 05:20 PM (#3062399)
GuyM,

I am not aware of any studies comparing the performance of a pitcher who throws 3-4 innings every 5 days, with that pitcher throwing an inning or less 3 times every five days. It is not a given that the second will perform better, and I can't think of good evidence either way. The only mildly suggestive thing is that AL starters do in fact pitch well during the 26-50 pitch range (NL starters obviously do because of they almost invariably face the pitcher during that range, and often do not during the 1-25).
   27. Steve Treder Posted: January 28, 2009 at 06:11 PM (#3062480)
I am not aware of any studies comparing the performance of a pitcher who throws 3-4 innings every 5 days, with that pitcher throwing an inning or less 3 times every five days. It is not a given that the second will perform better, and I can't think of good evidence either way.

The study I did here pretty strongly showed that a given pitcher performs more effectively in relief than when starting, and one can logically deduce from it that pitchers perform better in shorter stints than longer stints. But it doesn't directly isolate the sort of 3-4 innings every 5 days pattern you're envisioning; the issue with relative ineffectiveness in starting assignments is that it the guys will often have their roughest innings in the fifth or sixth.
   28. GuyM Posted: January 28, 2009 at 07:28 PM (#3062604)
Maybe it is or maybe it isn't, but being as it is a 2-year sample, I'm not confident in concluding much of anything about it, especially since in the years following the rate reverts back to virtually exactly where it was in 1977-1984. And overall, outside of the 1993-94 exception, the rate changes over the 30-year period -- a very long time -- are so miniscule ..

But we don't just have a 2-year sample. We know mathematically that a given lead is less secure in a higher run-scoring environment (just as it's less secure earlier in a game than later). The 1993-94 scoring surge meant it became about 4% less likely a team could be shut out for one inning, and 6% less likely to be shut out for two consecutive innings. I don't know the precise implications for win expectancy, but it certainly became significantly harder to hold a lead after 1994. For you to talk about this as a single 30-yr-period and ignore the changed run environment is, well, it's like saying AL pitchers are worse than NL pitchers at striking out batters while ignoring the existence of the DH in one league.

There are other ways to see the benefit of current bullpen usage as well. For example, in late/close situations (B-Ref) over the last three years, pitchers have posted an OPS against that's about 28 points below average. But from 1974-76 (3 of the glory years I picked at random, honest), pitchers actually had a worse than average OPS (+3 points) in those situations. Teams are getting better pitching performances when it matters now. (Which isn't to say they couldn't do even better -- I certainly support using "closers" for high-leverage PAs in the 8th or even 7th inning).

**

Mike: No one can do that study, because no pitchers get used in that manner. But starters definitely perform worse on 3 days rest. Over past 4 years, starters have posted an OPS+ against of 118 on 3 days rest, vs. 102 on 4-5 days rest. That's a huge difference. And you have to think that the starters asked to go on short rest were usually those the manager felt were most able to handle it!
   29. Karl from NY Posted: January 28, 2009 at 07:58 PM (#3062642)
Over past 4 years, starters have posted an OPS+ against of 118 on 3 days rest, vs. 102 on 4-5 days rest. That's a huge difference.

Not sure if you're being facetious, but there's a serious selection bias there. Only good pitchers get selected by managers to go on 3 days rest. The right comparison is performance on 3 days rest compared to the performance by those same pitchers on 4-5 days rest, adjusted for park and opponent.
   30. GuyM Posted: January 28, 2009 at 08:12 PM (#3062657)
The right comparison is performance on 3 days rest compared to the performance by those same pitchers on 4-5 days rest, adjusted for park and opponent.

Not facetious, but imprecise. I was using B-Ref's "tOPS+" which compares the pitcher's performance to his own mean peformance: http://www.baseball-reference.com/pi/psplit.cgi?team=TOT&lg=ML&year=2007. So on three days rest, a starter is substantially worse than he is on 4+ days of rest. (But the 3-day rest figure is actually 116, not the 118 I reported).

Tango et.al. also did a much more rigorous study on this, and found a significant performance decline on 3 days rest.
   31. Steve Treder Posted: January 28, 2009 at 08:56 PM (#3062706)
The 1993-94 scoring surge meant it became about 4% less likely a team could be shut out for one inning, and 6% less likely to be shut out for two consecutive innings. I don't know the precise implications for win expectancy, but it certainly became significantly harder to hold a lead after 1994. For you to talk about this as a single 30-yr-period and ignore the changed run environment is, well, it's like saying AL pitchers are worse than NL pitchers at striking out batters while ignoring the existence of the DH in one league.

Yeah, but both the team ahead and the team behind are playing within the same changed run environment; just as the team behind is more likely to score, the team ahead is more likely to tack onto its lead. The run environments between 1977-1984 and 2001-2006 were different, sure, but they weren't THAT different, and so the 1-run 8th-inning lead preservation percentage going from 76.5% to 76.8% honestly doesn't impress me as all that phenomenal an achievement. Add the costs involved in swapping one, two, or even three position player rosters spots for bullpen spots, and the ROI seems even less compelling.
   32. Proo Posted: January 29, 2009 at 01:36 AM (#3062928)
Add the costs involved in swapping one, two, or even three position player rosters spots for bullpen spots


Which in and of itself has an effect. Managers do not have as many late-inning pinch-hitting options, which means overall rosters will be less effective at scoring runs in those situations. This needs to be factored into the equation as well.
   33. Mike Emeigh Posted: January 29, 2009 at 02:15 AM (#3062954)
I've been studying this for 18 months now. The biggest issue we have is that reliever usage patterns changed rapidly between 1975 and 1990, and the only reasonably stable period we have had is the one we are in right now - which is the one that we want to evaluate for effectiveness. We really don't have any prior baseline for comparison because we don't have any sort of stable long-term pattern to compare *to*.

The one underlying trend throughout all of this period is that managers have been reducing the average stint for ALL of their pitchers - starters and relievers. The long reliever has virtually disappeared.

-- MWE
   34. Mike Emeigh Posted: January 29, 2009 at 02:34 AM (#3062964)
Here's an interesting little table that I pulled from my Retrosheet files:

Season     Games     RelieverUsed     RPUsedPerGame     LongStintPct     ShortStintPct
1954       2260      1488             1.26              26.2
%            24.0%
1955       2186      1536             1.41              27.5%            29.8%
1956       2274      1570             1.40              25.3%            31.0%
1957       2380      1696             1.43              27.5%            30.5%
1958       2430      1698             1.44              26.4%            31.7%
1959       2456      1719             1.40              25.5%            29.0%
1960       2464      1801             1.45              27.6%            29.7%
1961       2860      2115             1.44              24.7%            31.6%
1962       3216      2377             1.55              24.1%            36.3%
1963       3232      2367             1.48              24.0%            35.0%
1964       3238      2446             1.58              24.1%            35.1%
1965       3212      2479             1.65              24.3%            37.4%
1966       3198      2468             1.63              25.2%            38.7%
1967       3210      2437             1.60              23.2%            37.3%
1968       3188      2305             1.47              19.8%            36.0%
1969       3892      2910             1.60              20.7%            39.0%
1970       3888      3034             1.66              21.3%            43.1%
1971       3846      2770             1.49              19.3%            37.8%
1972       3710      2702             1.45              20.6%            33.9%
1973       3870      2813             1.37              25.9%            32.1%
1974       3890      2801             1.40              24.5%            33.6%
1975       3868      2816             1.40              25.2%            33.7%
1976       3878      2839             1.41              24.6%            33.0%
1977       4206      3299             1.53              27.0%            34.8%
1978       4204      3170             1.40              24.4%            31.5%
1979       4196      3283             1.52              25.5%            37.3%
1980       4210      3354             1.56              26.3%            36.9%
1981       2788      2278             1.67              25.8%            40.1%
1982       4214      3480             1.62              27.6%            37.3%
1983       4218      3473             1.60              25.2%            37.1%
1984       4210      3578             1.65              23.9%            34.5%
1985       4206      3579             1.74              22.5%            39.1%
1986       4206      3627             1.80              22.5%            43.7%
1987       4210      3649             1.89              21.5%            45.9%
1988       4200      3578             1.75              18.7%            42.3%
1989       4212      3729             1.88              19.8%            42.8%
1990       4210      3781             2.02              17.7%            49.6%
1991       4208      3842             2.13              17.0%            55.3%
1992       4212      3793             2.15              14.8%            59.5%
1993       4538      4167             2.27              12.6%            64.9%
1994       3200      2945             2.33              13.8%            71.5%
1995       4034      3759             2.45              14.2%            72.4%
1996       4534      4244             2.44              13.0%            70.0%
1997       4532      4266             2.50              11.9%            75.7%
1998       4864      4562             2.46              10.6%            71.1%
1999       4856      4619             2.56              11.0%            72.3%
2000       4858      4624             2.54              10.5%            71.0%
2001       4858      4659             2.63              10.2%            73.1%
2002       4852      4638             2.63              9.2%             73.1%
2003       4860      4651             2.67              9.7%             74.4%
2004       4856      4706             2.76              8.2%             77.7%
2005       4862      4673             2.71              6.5%             78.4%
2006       4858      4714             2.85              8.0%             80.9%
2007       4862      4750             2.97              7.3%             85.5%
2008       4856      4720             2.92              6.4%             81.0


The columns are, in order:
Games played
Games in which at least one reliever was used
Number of relievers used per game
Percentage of games in which at least one reliever pitched three or more innings
Percentage of games in which at least one reliever pitched less than one inning.

-- MWE
   35. Mike Emeigh Posted: January 29, 2009 at 02:47 AM (#3062972)
Add the costs involved in swapping one, two, or even three position player rosters spots for bullpen spots, and the ROI seems even less compelling.


The problem here is that there are fewer opportunities to use position players 13 through 15 than there are to use pitchers 10 through 12. You can always find 40-50 low leverage innings a year for your lesser pitchers; it's awfully hard to find 40-50 high-leverage pinch-hitting opportunities for those bench guys, PLUS enough starts during the course of a season to keep them sharp. It's been a lot easier to keep guys in AAA where they can actually face live pitching on a regular basis, and rotate them in and out of the 3 bench slots that most teams have; you will note that AAA is chock full of the guys who used to live on major league benches getting 5 PAs a week. End-of-the-roster pitchers are, by and large, cheaper in terms of salary as well.

It is my opinion that the decision to go to the 13/12 roster splits that you see today was largely driven by cost and opportunity. There are about twice as many low-leverage situations as there are high-leverage situations for a typical major league team, and they get far more value per dollar spent by using those last three roster spots on pitchers who can soak up low-leverage innings (letting them keep fresher arms for higher-leverage outings) than they do by getting guys who might get, at most, 20-30 meaningful late-inning PAs where they might deliver one time in four.

-- MWE

EDIT: Just for the record, I dislike it about as much as anyone else, but I don't see it changing because there is no strong on-field reason for it to change.
   36. Chris Dial Posted: January 29, 2009 at 02:50 AM (#3062974)
Thanks, Mike. That is awesome data. The 10 year change from 1984 to 1994 is just crazy. And I don't believe for a second it is necessary to win games.
   37. Wes Parkers Mood (Mike Green) Posted: January 29, 2009 at 03:26 AM (#3062996)
Yes, Mike E. That is helpful data.

"There are about twice as many low-leverage situations as there are high-leverage situations for a typical major league team, and they get far more value per dollar spent by using those last three roster spots on pitchers who can soak up low-leverage innings (letting them keep fresher arms for higher-leverage outings)"

I am of the view that they could get by easily with two pitchers throwing 80-100 low leverage innings, instead of three pitchers throwing 50-65 low leverage innings. Incidentally, the cost of the lost roster spot can be any number of things. Many AL teams try to ensure that the DH has a secondary defensive role because of the short bench. The cost of the lost roster spot can in those cases be measured in the difference in offensive performance between a pure bat and a player with a dual role.
   38. GuyM Posted: January 29, 2009 at 04:26 AM (#3063022)
And I don't believe for a second it is necessary to win games.

Why would you doubt it? Do you really think teams steadily kept moving in the same direction (shorter outings, more pitchers) because it was putting them at a competitive disadvantage? Strikeout rates have increased 35-40% since 1980. Good relievers are much more effective today than 30 years ago. A few ERA+:
Fingers 119
Gossage 126
Lyle 127
*
Hoffman 144
Wagner 180
Rivera 199
Papelbon 254 (just for fun -- I know he's still young)

Mike is right -- it's not going to change because it works. Shorter outings make all your relievers more effective. It gives you more opportunities to gain the platoon advantage. And pitchers can be deployed much more effectively than extra hitters, in terms of matching skill to leverage, because they aren't constrained by a lineup (and the constraint that the weak hitters you want to sub for often play key defensive positions).
   39. Mike Green Posted: January 29, 2009 at 02:46 PM (#3063180)
GuyM,

You're looking at it from a narrow perspective. There are two phenomena at work which result in the decreasing length of relief stints- starters being held into games to be given the chance to "win" and the pen optimized partially for the chance of saves and partially to increase the possibility of actually holding leads.

When Jon Garland or any other comparable average starter has given up 3 runs, say single runs in the 1st, 2nd and 4th, and is in a 3-3 tie after 4 innings having thrown 65 pitches and having gone through the lineup twice, how many times is he pulled? Answer: almost never. Is it because the evidence supports the view that Jon Garland is likely to pitch better in the 5th, 6th, and possibly 7th innings than whoever would replace him? No. He stays in the game because managers want to give the starter an opportunity to win the game, and also because the current bullpen usage dictates that relievers rarely come in during the 5th inning or earlier (i.e. only when the starter obviously doesn't have it that day).

There are not two binary choices- the pitcher usage of today and the pitcher usage of the 70s.
   40. GuyM Posted: January 29, 2009 at 03:52 PM (#3063276)
Mike G:
My comments are all about why the changes that have occurred since the 1970s make sense, and why MLB will never go back to the 1970s system. I haven't said, and don't believe, that future improvement is impossible.

Similar to what you're saying, I think it's likely that we'll see a team experiment at some point with a 4-starter-and-1-bullpen-game rotation. If you've got 13 pitchers, it's hard to believe you couldn't cobble together a bullpen game that's better than what many 5th starters offer. And you wouldn't even have to use the bullpen strategy in 20% of your games, only when you have games on 5 consecutive days. I think there's a good chance this would work (driving down salaries for 5th starters).

Another possible variation is putting four starters on a 3-day rest rotation, while aggressively pulling them early in the game when either team has a big lead. However, unless MLB changes the 5 IP requirement for starter wins, that will be a hard sell. Less radical is having teams move from a 5-man rotation (those that do) to a 4-day-rest rotation, skipping their 5th starter when possible, while pulling starters earlier when they have a big lead or big deficit. That would get you more starts out of your good starters, but same number of IP.
   41. Steve Treder Posted: January 29, 2009 at 05:19 PM (#3063360)
Mike, that data above is fabulous.

The problem here is that there are fewer opportunities to use position players 13 through 15 than there are to use pitchers 10 through 12. You can always find 40-50 low leverage innings a year for your lesser pitchers; it's awfully hard to find 40-50 high-leverage pinch-hitting opportunities for those bench guys, PLUS enough starts during the course of a season to keep them sharp.

I just don't think so. Look at the way a Casey Stengel used his rosters; by implementing multi-position platoon arrangements (and platooning not only on the L/R basis, but also on many other modes of complementarity) it's isn't difficult to find ways to get 150-250 at-bats a season almost to the 15th man. The best way to leverage the bench isn't to only seek late-inning PH ABs, but instead to liberally platoon throughout the starting lineup.

You must be Registered and Logged In to post comments.

 

 

<< Back to main

News

All News | Prime News

Old-School Newsstand


BBTF Partner

Dynasty League Baseball

Support BBTF

donate

Thanks to
BFFB
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Hot Topics

NewsblogSavages in the box go bust with slew of strikeouts
(14 - 2:02pm, Oct 20)
Last: Omineca Greg

NewsblogClinching heroics net Altuve ALCS MVP honors
(11 - 1:59pm, Oct 20)
Last: Howie Menckel

NewsblogWaiting game: Nationals get 6 days off before World Series
(4 - 1:54pm, Oct 20)
Last: Howie Menckel

NewsblogCC Sabathia leaves possible last appearance with injury
(30 - 1:37pm, Oct 20)
Last: The Yankee Clapper

NewsblogReady for the World Series? 10 legends that should throw out the first pitch
(54 - 1:25pm, Oct 20)
Last: Tin Angel

NewsblogReport: Yankees fans taunted Zack Greinke about his mother, battles with anxiety and depression
(45 - 1:13pm, Oct 20)
Last: Astroenteritis

NewsblogChampionship Series OMNICHATTER!
(1125 - 12:54pm, Oct 20)
Last: Bote Man sez Deivi is MoY

NewsblogOT- Soccer Thread- October 2019
(184 - 12:44pm, Oct 20)
Last: Fancy Crazy Town Banana Pants Handle

NewsblogOT - NBA thread (pre-season)
(566 - 11:44am, Oct 20)
Last: DCA

NewsblogEx-Red Sox knuckleballer Steven Wright to undergo Tommy John surgery
(1 - 9:13am, Oct 20)
Last: michaelplank has knowledgeable eyes

Hall of MeritRoy Halladay, Mariano Rivera, and "Cannonball" Dick Redding elected - Hall of Merit Class of 2019
(56 - 7:16am, Oct 20)
Last: bjhanke

Hall of MeritMost Meritorious Player: 2019 Discussion
(6 - 6:53am, Oct 20)
Last: bjhanke

NewsblogCC Sabathia has a case for Cooperstown
(20 - 1:50am, Oct 20)
Last: DanG

NewsblogBA: MLB floats proposal that would eliminate 42 minor league teams.
(33 - 1:28am, Oct 20)
Last: Greg Franklin

NewsblogOT - October 2019 College Football thread
(282 - 12:44am, Oct 20)
Last: Red Voodooin

Page rendered in 0.5080 seconds
46 querie(s) executed