16 replies to this topic

[Buzzkill Pauley]

Thanks much to DanooME in the Manager's Decision thread -- he put up some very interesting data about how pitchers have worked through opposing lineups there. But my observations aren't about BobbyV's hook, so they didn't seem to fit there.

________________


Sox starters are getting through opposing lineups three times at a 73.1% rate. That isn't good, but the AL average is only 74.7%, so there's only a net (1.6%) difference in how often the Sox starters make it through three full cycles of a lineup. Still slightly below average, but not too significantly.

What's significant is getting to that third plate appearance:
1st time through: .859 to .719 = net (.140) OPS difference
2nd time through: .899 to .751 = net (.148) OPS difference
3rd time through: .666 to .741 = net .073+ OPS difference

First the good news. As might be expected of a rookie and a converted reliever, Doubront and Bard get worse results as they get deeper into games. For both guys, though, the steady increase of HR and BABIP rates suggests both newbies have normal problems that could be managed to success by BobbyV using a quick hook in the 6th-7th innings and appropriate calls to the bullpen.

#4 SP: Felix Doubront

Doubront 2012	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip
1st PA in G, as SP	63	10	0	10	15	1.50	0.192	0.317	0.288	0.606	0.263
2nd PA in G, as SP	63	13	1	6	14	2.33	0.232	0.317	0.357	0.675	0.293
3rd PA in G, as SP	40	14	2	0	8	--	0.350	0.350	0.575	0.925	0.400

#5 Starter: Daniel Bard

Bard 2012	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip
1st PA in G, as SP	55	10	0	7	7	1.00	0.213	0.327	0.277	0.604	0.250
2nd PA in G, as SP	53	13	0	5	7	1.40	0.271	0.340	0.313	0.652	0.317
3rd PA in G, as SP	48	14	1	3	9	3.00	0.326	0.383	0.488	0.871	0.394

What's much, much more worrisome is how Lester, Beckett, and Buchholz have been unspeakably terrible the first and second times through the order, and consistently excellent the third time through. Meaning, it's up to that day's offensive output whether BobbyV can stomach leaving them in to get better.

#1 SP: Jon Lester

Lester 2012	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip
1st PA in G, as SP	63	18	2	4	12	3.00	0.305	0.349	0.492	0.841	0.356
2nd PA in G, as SP	62	15	1	10	9	0.90	0.300	0.410	0.480	0.890	0.341
3rd PA in G, as SP	49	5	0	2	7	1.75	0.114	0.204	0.159	0.363	0.135

#2 SP: Josh Beckett

Beckett 2012	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip
1st PA in G, as SP	55	13	3	5	9	1.80	0.271	0.333	0.542	0.875	0.270
2nd PA in G, as SP	52	17	4	4	7	1.75	0.354	0.404	0.688	1.091	0.351
3rd PA in G, as SP	39	7	2	0	9	--	0.179	0.179	0.385	0.564	0.179

#3 SP: Clay Buchholz

Buchholz 2012	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip
1st PA in G, as SP	64	21	4	9	8	0.89	0.396	0.500	0.736	1.236	0.415
2nd PA in G, as SP	63	17	3	8	5	0.63	0.327	0.419	0.538	0.958	0.311
3rd PA in G, as SP	50	12	2	1	4	5.00	0.245	0.260	0.367	0.627	0.238

If it were a problem with pitch-calling, I would expect to see similar issues with Bard and Doubront. Nothing there. And all three of these guys should​ have good enough stuff to get MLB hitters out -- just check those 3rd-time-around splits, nevermind their career numbers. Not one of them can claim the excuse of injury, either.

Which seems to me to leave only sloppy pre-game preparation and poor in-game execution -- the implication of both being that these three guys just aren't consistently getting ready to compete when it's their turn to pitch.

If that's really the case, then sadly they can have all the talent in the world, and it still won't be worth a damn.

Edited by Buzzkill Pauley, 14 May 2012 - 12:38 PM.

[absintheofmalaise]

Good idea for a thread. Have you checked on Brooks Baseball to see what types of pitches they are throwing in the innings? I don't have time right now, but will tonight. That might tell more of the story here.

[czar]

I agree with abs, it's interesting stuff. That said, while I don't have the time to look into it myself, I'd have to believe there probably is some serious confirmation bias occurring with the Lester, Buchholz, Beckett triumvirate.

When they suck, they've been pulled early (Beckett's last start, Lester's TEX start, and a few of Buchholz's starts come to mind) so their first-time through numbers are going to be skewed with those "bad starts." Games where they work deep into games are going to add weight to 2nd-3rd time through the lineups, so you are going to have a tough time teasing out "fatigue/batters figuring out pitchers" from this noise. Especially when we are talking about a sample of 6-7 starts.

There probably is some way to rudimentarily normalize the data that I can't think off the top of my head right now, but I'd think you'd need to do that with the above stats.

Another interesting thing that we commonly discuss in game threads, but outside of some Lackey (2011) and Bard (2012) threads hasn't made it to the main board is the trend in pitch velocity throughout the game. Lackey (possibly injured) and Dice-K (right before TJ) had very precipitous drop offs in velocity after the first inning or two (corresponding to a fast start-slow finish trend) Buchholz on the other hand has consistently ramped up velocity from the 1st through the 3rd-4th innings in 2012 (this would agree with a pitcher who gets better as the game goes along). While velocity is not going to drive things like OPS directly, studies show it plays a pretty important role in K% and batted ball profiles, so you'd expect it to play some role in determining how a pitcher's rates change throughout the game.

[24JoshuaPoint]

Which seems to me to leave only sloppy pre-game preparation and poor in-game execution -- the implication of both being that these three guys just aren't consistently getting ready to compete when it's their turn to pitch.

Does any of that fall on the catcher too? I.e. Variteks intangibles.

[absintheofmalaise]

One reason Bard and Doubront do better earlier could be not that many hitters are familiar with them yet. This should be something to watch further into the season.

[Buzzkill Pauley]

I agree with abs, it's interesting stuff. That said, while I don't have the time to look into it myself, I'd have to believe there probably is some serious confirmation bias occurring with the Lester, Buchholz, Beckett triumvirate.

When they suck, they've been pulled early (Beckett's last start, Lester's TEX start, and a few of Buchholz's starts come to mind) so their first-time through numbers are going to be skewed with those "bad starts." Games where they work deep into games are going to add weight to 2nd-3rd time through the lineups, so you are going to have a tough time teasing out "fatigue/batters figuring out pitchers" from this noise. Especially when we are talking about a sample of 6-7 starts.

There probably is some way to rudimentarily normalize the data that I can't think off the top of my head right now, but I'd think you'd need to do that with the above stats.

I would agree about normalizing the data -- if one were planning on using the data as predictive. However, there's no way I can think of to rationally explain why the top three Sox starters should continue pitching so much worse than an average AL pitcher, so I didn't see any reason to try tweaking the stats.

And there certainly is confirmation bias among the top-three starters -- when the Sox offense can explode for 7 runs over the first three innings, BobbyV can afford to let Buchholz face Fielder a third time in the 4th inning. But even so, here are the following percentages for getting thrice through a lineup:

AL-avg = 74.7%:
Lester: 77.8%
Beckett: 70.9%
Buchholz: 78.1%
Doubront: 63.5%
Bard: 87.3%

Bard has been by far the best at plowing multiple times through a lineup, even if that doesn't necessarily mean he's getting equivalently deep into games, when measured by innings pitched.

__________


Also, I decided not to breakout each pitcher by catcher or pitch type, mostly in order to keep the initial post somewhat manageable in length -- any additional information and analysis would, of course, be greatly appreciated.

Edited by Buzzkill Pauley, 14 May 2012 - 12:42 PM.

[finnVT]

One question/comment about this data. It could easily be explained by having some "good" starts (in which they make it through the lineup 3x) and some bad ones (where they're removed sooner). So the 3rd time through looks better because it just happens to show data from their good starts only, whereas the other data shows all starts (which have been pretty bad on the whole).

To be more convinced that there's actually a change occurring on the 3rd time through the rotation (and not just from start-to-start), it'd be interesting to see this data only from the starts in which they made it through 3x, and see if the numbers still hold up.

edit: czar beat me to it

Edited by finnVT, 14 May 2012 - 12:17 PM.

[Buzzkill Pauley]

Fuck -- for some reason various stats got doubled when I ran them through the table-izer.

I am editing the OP right now.

edit --

OP and follow-up post now edited with correct numbers. I knew Beckett had given up a lot of HRs, but should also have known even he couldn't be up to 18 by mid-May.

It wasn't the table-izer, but something odd happened when I brought the BP data over to excel or when I was deleting cells.

Regardless -- the problem seemed to primarily effect the raw data by doubling numbers, so although there are a few random rate-stat changes, the analysis follows the same trend lines.

I am embarrassed, but no further damage. I guess.

Edited by Buzzkill Pauley, 14 May 2012 - 12:47 PM.

[czar]

And there certainly is confirmation bias among the top-three starters -- when the Sox offense can explode for 7 runs over the first three innings, BobbyV can afford to let Buchholz face Fielder a third time in the 4th inning. But even so, here are the following percentages for getting thrice through a lineup:

AL-avg = 74.7%:
Lester: 77.8%
Beckett: 70.9%
Buchholz: 78.1%
Doubront: 63.5%
Bard: 87.3%

Bard has been by far the best at plowing multiple times through a lineup, even if that doesn't necessarily mean he's getting equivalently deep into games, when measured by innings pitched.

Those percentages don't add up-- are the numbers supposed to be the percentage of starts getting through 27 PA? If so they should all be divisible by the (inverse) # of starts the player has made. Lester/Buchholz have each made 7 starts so it's impossible for them to be off by ~0.5%.

But this really doesn't tell you anything. If the 22% of starts Buchholz doesn't get through 27 PA (-27PA) are 7x worse than his 78% +27PA starts it will disproportionately skew his stats more than if Doubront's 37% -27PA are only 2x worse than his 63% +27PA (sorry if that's confusing). This bias will probably be especially evident in SSS.

I like what finnVT suggested. Unfortunately, the sample is way too small to worry about that in 2012, but you might be able to go back to 2011 and look at all start where Lester or Beckett makes it 3 times through the lineup and see if these trends (better/worse with batters faced) hold up.

Edited by czar, 14 May 2012 - 01:28 PM.

[Buzzkill Pauley]

Those percentages don't add up-- are the numbers supposed to be the percentage of starts getting through 27 PA? If so they should all be divisible by the (inverse) # of starts the player has made. Lester/Buchholz have each made 7 starts so it's impossible for them to be off by ~0.5%.

The numbers you quoted are the number of 3rd PA's as a percentage of 1st PA's, per starter. Basically, it rates how deep a pitcher gets through an opposing lineup after it turns over twice. This is regardless of the number of innings (outs), and usually has some small amount of noise based on pinch-hitters, defensive subs, and/or injury replacements faced by a starter. This noise is why those percentages don't add up.

If one prefers, it would be possible to divide by the closest multiple of 9 less than the actual number of 1st PA's. But that's not how BP tracks this split, so I didn't think about doing that.

Take Doubront for example -- since he's got no noise in his results:

63 1st PA = (7 GS * 9 BF) Doubront made it once through each opposing lineup in each start, and had no later pinch-hitters, defensive subs, or injury replacements.

63 2nd PA = in each of his starts, Doubront made it twice through the opposing lineup, and was never pulled without having pitched through the second full cycle.

40 3rd PA = Doubront averaged about 6 batters deep (40/63, or 63.5%) through the third cycle of the opposing lineup, in his seven starts.

2 4th PA = Doubront almost never pitches a fourth time through the lineup.


I'm not claiming this breaks any new ground or otherwise is critically-important information. I simply thought the OPS/BABIP trends were interesting enough between the 1-2-3 starters and the 4-5 starters, to merit a thread.

[czar]

The numbers you quoted are the number of 3rd PA's as a percentage of 1st PA's, per starter. Basically, it rates how deep a pitcher gets through an opposing lineup after it turns over twice. This is regardless of the number of innings (outs), and usually has some small amount of noise based on pinch-hitters, defensive subs, and/or injury replacements faced by a starter. This noise is why those percentages don't add up.

If one prefers, it would be possible to divide by the closest multiple of 9 less than the actual number of 1st PA's. But that's not how BP tracks this split, so I didn't think about doing that.

Take Doubront for example -- since he's got no noise in his results:

63 1st PA = (7 GS * 9 BF) Doubront made it once through each opposing lineup in each start, and had no later pinch-hitters, defensive subs, or injury replacements.

63 2nd PA = in each of his starts, Doubront made it twice through the opposing lineup, and was never pulled without having pitched through the second full cycle.

40 3rd PA = Doubront averaged about 6 batters deep (40/63, or 63.5%) through the third cycle of the opposing lineup, in his seven starts.

2 4th PA = Doubront almost never pitches a fourth time through the lineup.

OK, thanks for clearing that up!

EDIT: I'm not trying to argue the topic isn't interesting by any stretch-- I just think there are so many variables at play it's quite unfair to argue something like "sloppy preparation" based on a few start sample size. I think there likely are very real signals in how pitchers "evolve" over a game but I'd argue you'd need really big sample sizes (and careful data parsing) to see them clearly.

But it's a good point for discussion.

Edited by czar, 14 May 2012 - 03:36 PM.

[DanoooME]

Almost two more times through the rotation and let's see if things are normalizing

5/14 data:

Times Facing Opponent in Game - Sox Total 5/14

PA	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st	32	291	73	9	32	48	1.5	0.289	0.372	0.486	0.859	0.323	116
2nd	32	282	77	10	31	39	1.26	0.314	0.393	0.506	0.899	0.338	126
3rd	29	214	48	7	8	37	4.63	0.236	0.272	0.394	0.666	0.258	66
4+th	14	32	8	1	11	3	0.27	0.381	0.594	0.571	1.165	0.412	198

5/24 data (eliminated some of the columns):

Times Facing Opponent in Game - AL Total

Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	620	5611	1244	146	474	1105	2.33	0.247	0.315	0.395	0.710	0.287	98
2nd PA in G, as SP	616	5441	1288	151	447	931	2.08	0.264	0.329	0.419	0.748	0.296	108
3rd PA in G, as SP	584	4194	980	138	328	692	2.11	0.259	0.323	0.432	0.755	0.282	110
4th+ PA in G, as SP	209	474	114	11	46	81	1.76	0.272	0.348	0.418	0.766	0.313	114

Times Facing Opponent in Game - Sox Total

Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	44	401	90	11	45	70	1.56	0.260	0.350	0.431	0.781	0.294	106
2nd PA in G, as SP	44	388	100	12	40	60	1.50	0.297	0.378	0.463	0.841	0.330	123
3rd PA in G, as SP	41	296	69	8	15	48	3.20	0.249	0.295	0.379	0.674	0.276	78
4th+ PA in G, as SP	18	46	13	1	13	7	0.54	0.406	0.565	0.594	1.159	0.480	209

Some regression to the norm, significant improvement in the first time through the lineup (79 pts of OPS), almost as much improvement the second time through (55 pts of OPS). A small increase in the 3rd time through and not enough data the 4th time through to have any significant change. Is this just regression to the mean? Or is it some sort of focused improvement?

Edited by DanoooME, 24 May 2012 - 11:02 PM.

• Back to top

[Plympton91]

Tonight, with Beckett at 91 pitches, he pulled him just as the lineup was about to turn over a 4th time.

Is it just me, or have BV's hooks come much more quickly since this thread appeared? And, if so, is that a coincidence, or a function of the bullpen suddenly becoming historically awesome?

[Soxfan in Fla]

The fact the pen has been terrific over the last month allows Booby to get the starters out at the right time.

[seantoo]

Tonight, with Beckett at 91 pitches, he pulled him just as the lineup was about to turn over a 4th time.

Is it just me, or have BV's hooks come much more quickly since this thread appeared? And, if so, is that a coincidence, or a function of the bullpen suddenly becoming historically awesome?

I was at last nights game and Becket breezed through the game until the last inning. He had only thrown sixty something through six IP and reached 91 through 7 IP. He allowed only 1 hit through 6 and 3 in the seventh inning alone as both runs against him crossed the plate in the seventh too. Clearly he labored at the end, therefore, quality of bullpen aside it was the right move to make.
Your larger point still stands despite last nights example not fitting the mold. Initially the bullpen was horrible but has been terrific for an even longer stretch of time and in doing so apparently earned BV's trust. I think that he's starting to find the happy medium now balancing the starting pitching and bullpen. It's amazing how simple winning can also shape our perception too.

One trend I'm noticing is that BV likes to sacrifice alot more than TF. I'm not a big fan of small ball even though I acknowledge certain situations do call for it and last night with 2 pitchers clearly on their game qualified as one of those situations.

[DanoooME]

It's been a couple of months, let's check again to see if there's normalization:

Times Facing Opponent - AL thru 7/29

Times Facing Opponent in Game - AL 2012
Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	1422	12866	2978	370	987	2519	2.55	0.256	0.319	0.411	0.730	0.296	101
2nd PA in G, as SP	1408	12395	2972	351	962	2185	2.27	0.266	0.327	0.424	0.751	0.300	106
3rd PA in G, as SP	1334	9532	2302	334	728	1587	2.18	0.267	0.328	0.445	0.773	0.291	112
4th+ PA in G, as SP	475	1112	263	24	99	192	1.94	0.264	0.335	0.410	0.745	0.306	106

Times Facing Opponent - Sox thru 7/29

Times Facing Opponent in Game - Red Sox 2012
Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	102	929	224	27	76	170	2.24	0.270	0.335	0.445	0.780	0.307	112
2nd PA in G, as SP	101	894	207	24	82	159	1.94	0.264	0.341	0.427	0.767	0.301	110
3rd PA in G, as SP	97	677	161	27	38	122	3.21	0.257	0.305	0.440	0.745	0.279	101
4th+ PA in G, as SP	33	84	22	1	14	12	0.86	0.324	0.440	0.485	0.926	0.375	154

Interesting, almost no improvement the first time through the order in two months (1 pt of OPS). However, 78 points of improvement the second time through the order is nothing to sneeze at. Now they are within shouting distance of the AL average. They are still better than average the 3rd time through, even though they gave back almost everything they've gained the second time through (+71 points of OPS). They are still getting hammered the 4th time through the lineup, but the sample is still small.

So the Sox continue to have the reverse pattern of the rest of the AL, where they give up more early in the game as opposed to later in the game. Is this unusual? It seems so.

Times Facing Opponent in Game - Red Sox 2011

Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	162	1477	346	35	122	279	2.29	0.263	0.334	0.391	0.725	0.308	103
2nd PA in G, as SP	162	1422	304	40	136	254	1.87	0.243	0.322	0.394	0.716	0.272	100
3rd PA in G, as SP	149	1055	247	37	87	184	2.11	0.266	0.338	0.452	0.791	0.292	120
4th+ PA in G, as SP	54	127	30	1	13	17	1.31	0.278	0.368	0.435	0.803	0.319	125

Times Facing Opponent in Game - Red Sox 2010

Split	G	PA	H	HR	BB	SO	SO/BB	BA	OBP	SLG	OPS	BAbip	tOPS+
1st PA in G, as SP	162	1477	322	26	139	307	2.21	0.247	0.326	0.369	0.695	0.302	92
2nd PA in G, as SP	161	1422	318	26	110	289	2.63	0.248	0.311	0.382	0.694	0.299	91
3rd PA in G, as SP	157	1232	294	33	105	212	2.02	0.268	0.336	0.432	0.769	0.303	111
4th+ PA in G, as SP	80	210	46	4	29	25	0.86	0.263	0.370	0.400	0.770	0.284	113

Some of the data has normalized, and I'd expect it to continue to normalize, but this is an unusual pattern so far. When I get some more time this evening, I'll break it down by starter like Buzzkill did starting this post.

• Back to top

---

#17 BCsMightyJoeYoung

• 
• 
• 1,309 posts

Posted 30 July 2012 - 10:51 AM

The "4th time through the order" already seems to have normalized. Seeing as, as of the 24/may, the tOPS+ was 209 and is now down to 154 .. which means the last couple of months it's been right around league average

• Back to top

---

[BCsMightyJoeYoung]

The "4th time through the order" already seems to have normalized. Seeing as, as of the 24/may, the tOPS+ was 209 and is now down to 154 .. which means the last couple of months it's been right around league average