The Progression of Pitchers under Jim Hickey
by FreeZorilla on Jun 19, 2009 1:04 PM EDT
Email
PrintDRB contributor Matthan has done some very comprehensive work on the performance of pitchers in year 1 and year 2 under the direction of pitching coach Jim Hickey. He came to the conclusion that there is a strong statistical evidence of a decline in pitchers'' K/9 under Hickey. The next step is establishing why. Is it intentional? If so, how come?
For my numbers, I divided the pitchers into 2 groups based on the number of innings pitched. Over 100 IP in each year crowned you a starter. One pitcher, Brandon Backe, was thrown out for being over 100 and under 100 (injury). In total, there were 10 relievers and 8 starters. This field of pitchers comes from the 2005 and 2006 Astros and the 2007 and 2008 Rays. The starters are divided evenly between the teams while 6 of the 10 relief pitchers came from the Rays.
The metrics I looked at were strikeouts per nine innings, unintentional walks per nine innings, strikeout percentage of total batters faced, strikeout to walk ratio, pitches per inning, and strike %. I looked at each pitcher's change from year 1 to year 2 under Hickey, the average change per pitcher, and the change in the total performance of the group (For example K/9 would be combined strikeouts / combined innings * 9).
Below is the data of the changes. The hard #'s are available on the google spreadsheet at the end of the article:
Starters
|
Player |
K/9 |
uBB |
K% |
K/uBB |
P/Inn |
Strike% |
|
Andy Pettitte |
0.55 |
0.9 |
-0.38% |
-1.25 |
1.91 |
-2.31% |
|
Roger Clemens |
0.22 |
-0.2 |
0.54% |
0.4 |
0.93 |
-0.65% |
|
Roy Oswalt |
-0.08 |
-0.29 |
0.16% |
0.79 |
-0.15 |
0.26% |
|
Wandy Rodriguez |
0.91 |
0.15 |
1.75% |
0.18 |
0.63 |
0.38% |
|
Andy Sonnanstine |
-0.91 |
-0.02 |
-2.37% |
-0.5 |
-0.94 |
1.01% |
|
Edwin Jackson |
-1.85 |
-1.02 |
-3.32% |
-0.08 |
-1.75 |
1.21% |
|
James Shields |
-1 |
0.17 |
-2.81% |
-1.11 |
-0.21 |
-1.27% |
|
Scott Kazmir |
-0.6 |
0.19 |
-1.05% |
-0.27 |
0.59 |
-0.28% |
|
|
|
|
|
|
|
|
|
Average Change |
-0.35 |
-0.02 |
-0.93% |
-0.23 |
0.13 |
-0.21% |
|
Group Change |
-0.57 |
-0.02 |
-1.67% |
-0.2 |
0.13 |
-0.19% |
Notice there is a strong difference between the ballclubs. All four of the Rays starters had significant declines in K% and K/9 while also throwing fewer pitches per inning outside of Scott Kazmir. Its probably worth looking into pitches/plat appearance also. Both the Astros and Rays had improved defenses in their second year. In the case of the Rays, it was a conscious effort to upgrade defensively. The Astros UZR/150 jumped from 2.9 to 6.1. However Batting Average on Balls in Play went up from .289 to .294. The Rays UZR jumped from a MLB worst -6.5 to an MLB best 11. As you would expect their BABIP fell from .338 to .285. This large a decline in BABIP would have an unavoidable drop in K/9 all things being equal, but K% should remain unaffected. However, if pitchers were pitching to contact, their walks would decrease as well.
There is also the chance there is much deeper strategy. Perhaps, the Rays decided to go aggressively after players with low slugging due to their defense, while at the same time pitching around sluggers more than the average team due to their faith in the defense helping to strand the runners. There is no way to draw any conclusions at this point without tying other metrics into the equation. The next stats to look at in my opinion would be first strike%, evaluating the characteristics of the batters being walked, and pitches/plate appearance. Comparisons of teams with similar defensive improvements also might be worth a look.
The more collaborative this effort is, the better the results will be. I look forward to your feedback and improvements.
Below is the relief matrix. I am struggling to do much with this due to the smaller inning samples and the greater variance.
|
Player |
K/9 |
uBB |
K% |
K/uBB |
P/Inn |
Strike% |
|
Brad Lidge |
-0.63 |
1.04 |
-4.81% |
-1.43 |
1.92 |
-4.39% |
|
Chad Qualls |
-1.09 |
-0.14 |
-2.51% |
-0.31 |
-0.66 |
-1.21% |
|
Dan Wheeler |
0.11 |
0.06 |
-0.91% |
-0.06 |
1.61 |
-3.52% |
|
Russ Springer |
-1.3 |
-0.48 |
-2.78% |
0.07 |
-1.06 |
1.29% |
|
Al Reyes |
-2.85 |
0.21 |
-7.77% |
-1.13 |
-0.17 |
-0.31% |
|
Dan Wheeler |
-2.17 |
-0.44 |
-2.34% |
-0.31 |
-3.14 |
0.23% |
|
Gary Glover |
-0.11 |
0.65 |
-1.52% |
-0.43 |
2.2 |
-0.67% |
|
Grant Balfour |
1.61 |
-2.99 |
10.39% |
1.88 |
-3.61 |
2.15% |
|
J.P. Howell |
0.63 |
0.12 |
4.78% |
0.09 |
-2 |
1.31% |
|
Jason Hammel |
-1.72 |
-0.57 |
-3.95% |
-0.22 |
-1.92 |
0.05% |
|
|
||||||
|
|
||||||
|
Average Change |
-0.75 |
-0.25 |
-1.14% |
-0.19 |
-0.68 |
-0.51% |
|
Group Change |
-0.35 |
-0.04 |
-0.36% |
-0.08 |
-0.53 |
-0.72% |
54 comments
|
4 recs |
Do you like this story?
Comments
Part 2 of my study is more relevant to FreeZorillas work
Anyways what really stands out to me is Balfour and Howell from a managerial standpoint. They were the only relievers to really increase their K/9 from year 1 to year 2 with Hickey however their usage climbed dramatically in year 2. That leads me to believe that Maddon (and Hickey) were able to act upon these changes and allocate innings effectively (especially into high leverage situations)
Can we just give an award or something to matthan and FreeZorilla?
Fantastic work all around and the kind of shit that makes DRays Bay great.
+1
Kaz/Shields/Garza/Sonny/Price/Davis/Hellickson-necessitate a drool cup or a 7 man rotation
by CubFanRaysaddict on Jun 19, 2009 3:01 PM EDT up reply actions
Also I think the most important change is the change in pitch types. As FreeZorilla kindly pointed out that K/9 is at least linked to defense, walks, IBBs, etc. So a decline in that rate is explainable in many different ways (and not all ways would be bad).
However I really can’t get around the change in pitch types. I really can’t see why there would be any other reason for Hickeys pitchers to essentially move from changeups to curveballs other than Hickeys bias.
Also playing Devils Advocate to Freezorilla….perhaps the increase in curveball usage in year 2 (and decrease of changeup usage) is also a factor in the decline in K/9?
I haven't really looked into your pitch type work yet
Did you look at the change in linear pitch weight?
Follow Me on Twitter @FreeZorilla
I only looked at usages. The change in the usages were stronger than K/9 and obviously is tougher to explain away. The change in CH% is the strongest. Over this data we can pretty much say CH% decreased with a confidence of over 99 or so %. It most definitely happened across the board. There isn’t even a discrepancy among teams, years, or anything. Pretty much every single year every single pitcher threw their changeup less.
If I recall 20 of 28 in the sample threw the CH less. 3 of those 8 had increases less than 1% whereas 16 of the 20 decliners had decreases of over 1%.
By percent change I mean of overall pitch usage. As in if someone threw 30% changeups a 1% change would make it either 29% or 31% changeups….not a 1% percentage change off of 30%
Definitely appreciate the effort.
My question is, does this really tell us much? If I were to pick out 20-30 random pitchers, would we see similar results?
No it doesn't
This would/will take a lot more effort from a lot more people to draw any substantial conclusions. I don’t even have much of a conjecture at this point.
Follow Me on Twitter @FreeZorilla
We shouldn’t draw any actionable conclusions. We can’t say Hickey is making our pitchers worse (as the FIP increase wasn’t significant). But we can most definitely say pitchers have had declining K/9 and increase CB% and decreased CH%. You’ve been trying to explain the K/9 which is critical. Without having a significant FIP change we have to understand why something is happening and if that something is a good or bad reason. Which makes your work here very critical. However the k/9 decrease is significantly significant. At this point its in the ‘draw your own conclusions’, but the decline happened.
I would disagree with 'most definitely' and say it's 'possible' because of the sample sizes you guys are working with.
"Where we all wait in earnest with pudding in hand for the Upton comet to sail through the roofed skies, so that we may meet Him."
Certainly fair enough
Although my sample of 06/07 astros, 07/08 rays, 08/09 rays was near normal. For example if i recall correctly we have 3 pitcher 2 k/9 above the mean change and 3 pitchers 2k/9 below the mean change…
Of course a bigger sample would make it ‘more normal’ if there was such a word.
'normaler'
"Where we all wait in earnest with pudding in hand for the Upton comet to sail through the roofed skies, so that we may meet Him."
I was more referring to my work, than your conclusions
Mine was based on your conclusion. Its there but why?
Follow Me on Twitter @FreeZorilla
That is exactly my point
My work was more about that it did in fact happen. Now your work and even more work needs to be done about why it happened and the consequences of it. There is no way anyone can draw conclusions without knowing the why/consequences part of the equation.
You would rarely find the same results among other samples.
My assumption is that these metrics shouldn’t change on a per pitcher basis year over year given the same conditions (same park, same team, same division, same PC, etc)
So the whole point is testing the significance of it is to determine just that. What is the likelihood that this is just random chance? Given that the P Values for K/9, CH%, CB% were very small then the chances of these changes happening due to chance is very small. Of course these changes may be a good thing (or bad) as that is up to interpretation, but they most certainly happened.
One more thing...
This data is year 1 to year 2 under Hickey. However year 2 to year 3 shows the same trends.
As I mentioned in another thread if you guys are looking to do this for other PC's (ALE to start?)
I would be willing to compile the data, but it would go faster if I have a template to just plug names and numbers into which would also help in following the same methodology. I’ve never seen this information publicly (nor privately, though I’m sure someone has it) and would like to take it to the next step.
Rays Win!
My google doc (bottom link) has a pretty simple template for the various metrics
Unintentional walks I did manually. Its probably a decent starting point. I’d be extra cautious with the averages due to cell references but the formulas are there.
Follow Me on Twitter @FreeZorilla
OK I'll take a look at Toronto and maybe Boston tomorrow since I think they've had some continuity there
For uBB did you just use BB-iBB? HBP?
Rays Win!
by Sandy Kazmir on Jun 19, 2009 2:27 PM EDT up reply actions
I'll try to get an easy to use template for my analysis
Basically if we are going to look at year over year changes we have to narrow the pitching coaches down to guys who have coached 06-09. So three consecutive years. I do think its important to at least be looking at the same time period and same season sample size.
Just a FYI looking at how a pitcher changes under one pitching coach was far easier and quicker to compile than see how pitchers changed between pitching coaches. So we should definitely stick with that (for now) even though the biggest change is probably from the year with a different coach to the year with the coach we are lookign at.
Who are some of the "best" pitching coaches in baseball?
I would imagine that many staff’s have players that perform differently year to year. Some guys get injured and try to pitch through it(Kazmir). Some guys get off PED’s(Petitte). Some guys arent in a contract year.
Brad Lidge is a nice example of a guy that had a HUGE yr last year. Does his pitching coach get some credit for that? What about now that he sucks again? Does that same pitching coach get blamed?
Maybe these guys are performing worse, but without Hickey they would have been even worse thenthat. Kinda like how the economy sucks now, but Obama saved jobs….so it could have been far worse.
Just playing devil’s advocate.
I agree with your mindset for the most part, but the more I'm reading into this the more it seems like
certain components and metrics may show sustained differences. After seeing other teams it will make for an interesting comparison.
Rays Win!
by Sandy Kazmir on Jun 19, 2009 2:29 PM EDT up reply actions
And I'm not dismissing all the data...
I just think there are a lot of things that go into having a successful pitcher…
Also another thing to note…just because over a period of time we see a change in K/9, CH, and CB doesn’t mean that will happen with every single pitcher. When we say we see a change of like .6 K/9 per pitcher (or broken down by starter or reliever) we don’t mean that is what will happen for every single pitcher. There will certainly be some pitchers that see an increase.
Obama saved jobs?
Anyway, you’re looking at it the wrong way: it’s not Hickey or nothing. It’s a matter of Hickey or someone else. I think the question being debated now is would that make a difference.
I think you emphasized the point that there are two very crucial ways to look at it. I attempted both, although my part 1 needs a lot of work statistically.
You have to see how a pitcher changes when he switches from 1 pitching coach to another pitching coach
We also have to see how a pitcher progresses over time while under the same pitching coach
Also I don’t think there is any way we can statistically come up with a way to “blame” a pitching coach on the failures or success of one pitcher. That is why you need a large sample size. For every guy who randomly became great you’ll have a guy who suddenly sucked. It would be more looking at the group as a whole.
You definitely put forth a ton of work...
The data is interesting. I wonder if the Rays use this sort of data when evaluating new hires…
The problem is most new hires and coaches in general don't have a diverse and large enough sample size to draw any real conclusions with
Heck first time hires have absolutely no data to look at
Although I’m sure they have really smart stats guys that are able to figure out a way to do things like this in some sort of reliable fashion.
I wouldn’t be surprised if somehow they did some leaguewide analysis to see what traits in certain coaches cause certain metrics to rise and fall and then look for coaches that are good or emphasize those traits.
There are just way too many factors in play here to pin much, if anything, on the coach.
Especially pitch selection. How many times do you see a pitcher shrug the catcher’s sign off? How do we know if he’s shrugging to or away from the pitching coach’s strategy? When you’re on the mound, I don’t think you’re doing much overstrategizing, just going with what’s working, what’s comfortable. That’s an assumption, though, maybe td32 can shed more light on that.
I admire the thought that went into this, but, in the end, nothing persuaded me one way or the other.
So long, Sweet Lime!
A perfect example of POW's point was yesterday....
When Barmes hit the HR off Garza. He had absolutely blown him away with FB’s in the first AB. Now, this doesnt mean that you can’t mix it up, but Barmes WAS overpowered and a FB would have been a better choice then a breaking ball. Whose fault was that? If Navi chose that pitch, it is both their faults. If Garza shook to that pitch, it is mostly Garza’s fault. Hickey can only prepare a pitcher for battle, but he doesnt fight for him.
The other problem with that pitch is that it sucked out loud. He hung it. If he throws a good breaking pitch, I doubt it goes for an XBH. Maybe a blooper, but you cant control that. It was poor execution. Garza clearly knows how tothrow a breaking pitch, but he didnt execute.
The problem is you would then be assuming pitchers are shrugging off one pitch in favor of another pitch more often in the 2nd year rather than the 1st.
Either way the data shows the change. The question is why. It could be because of Hickey and it may not be. Thats why more work is needed.
This was never said to be any sort of make or break evaluation of Hickey
Its a look at pitcher performance under the guidance of Hickey, a look at any potential trends, and attempts to rationalize why the trend exists
Follow Me on Twitter @FreeZorilla
Maybe the pitchers arent executing pitches...
Shaking off pitches has little to do with this argument. The pitches arent being called from the dugout.
I don’t follow. If the pitchers are executive their change up less the 2nd year than the 1st year to the point they stop throwing it as much wouldn’t that be a definite indication of Hickey? That would point to him being a bad teacher rather than him just being in favor of the curve over the change.
Well the data isn’t showing that Hickey somehow persuaded these guys to throw different pitches. The data is showing that the pitchers are throwing more curves and less change ups. This could very well be because Navi is the one with the bias. Or perhaps whoever or however they labeled pitches changed over the course of one season.
Don’t lose sight of what the data is saying. I know we are centering it around Hickey but I think it is crucial that the data just shows what has happened and not why it has happened.
Okay, but this organization is very uniform in its approach
So why would they have a pitching coach who seems to not like changeups when they sent their top pitching prospect down to the minors to learn one? And they have Jimbo Shields teaching everyone in sight how to throw one? It doesn’t add up.
So long, Sweet Lime!
Shields is actually a big culprit...
Every single year his CH% has decreased….
And you are right. That is what they preach. But the data shows that on average each individual pitcher is throwing about 2% less change ups.
If pitchers are learning a "new" changeup, they may be less confident at first
than with their old one. This could result in short-term decline of usage. As they gain confidence and improve with the pitch it is quite possible for usage not only to catch bak up, but surpass the usage of the old change up.
Follow Me on Twitter @FreeZorilla
Of course that doesn't explain Shields
And also Kazmir in 2008 had a very large increase in the change up. So that kind of goes against that theory.
That's always been my first instinct
Which further ensconces me in the opinion that there’s not enough data here (and probably never will be) to isolate and define Hickey’s influence.
So long, Sweet Lime!
Lets take a look at changeup usage with our two "best" CH guys
Shields
2006 NonHickey- 34.1%
2007 Hickey-30.2%
2008 Hickey-26.3%
2009 Hickey-24.3%
Nelson
2008 NonHickey w/ Marlins-32.7%
2009 Hickey-28.5%
FYI Nelson wasn’t used in my analysis as he didn’t qualify with IP (and when he does it will make the evidence even stronger).
I see where you're going, but there are so many possibilities when you try to crawl inside of a person's mind to evaluate the decision making process
So long, Sweet Lime!
Hence again why you need a large enough sample
For a singular pitcher there could be tons of reasons why they throw one pitch compared to another.
When looking at a large and diverse group of pitchers it is a different story.
But it still doesn't add up, for reasons I cited with Price and Shields' CH tutelage
And, if FreeZo is right, and I suspect he is, then the picture isn’t complete.
So long, Sweet Lime!
The data is the data. Although FreeZorilla hasn’t looked at the pitch type usage. I think what he was trying to show was that a change in one metric (BABIP) essentially caused the change in another metric (K/9).
But either way I didn’t just merely show that the average % of changeups declined…I show that the decline was statistically significant from zero. There is a huge difference. There is like a 99% chance that it wasn’t due to random chance.
Maybe it isn’t Hickey…I don’t pretend to know why they are throwing less changeups. I just know that they are.
I think I'm going to do a big piece on the changeup and curveball
Hopefully I’ll make the front page.
To clarify the decline is K/9 is exaggerated by the decline in BABIP
K% still did decline as well
Follow Me on Twitter @FreeZorilla
Still excellent work
It was an angle I didn’t even think of. At first I just wanted to see what was happening (why I did my analysis) but of course the why is important to. And it didn’t even occur to me that BABIP would hurt K/9.
I just ran some simple regressions for K/9 and BABIP
One with all year over year data (Astros and Rays) and one with just the Rays data
The R-Squared for all the data is just 4.6%
The R-Squared for just the Rays data is 13.9%
So clearly the change in BABIP has become a bigger factor in the change of K/9 during his Rays tenure…but it doesn’t totally explain the change in K/9
Comments For This Post Are Closed
Secondary Sidebar
User Tools
Please read our Community Guidelines.
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
DRB Suggestion Box
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
