The Hardball Times -- Steve Sommer

The Hardball Times -- Steve Sommer http://www.hardballtimes.com/main Baseball. Insight. Daily. en studes@hardballtimes.com Copyright 2013 2013-06-18T08:13:15+00:00 General draft trends http://www.hardballtimes.com/main/article/general-draft-trends/ http://www.hardballtimes.com/main/article/general-draft-trends/#When:08:04:15
With that in mind, I thought it would be interesting to see what draft trends, both league-wide and by team, have looked like over the last 10 seasons. The data pulled for the investigation were from the first 10 rounds of the drafts from 2001-2010.

High school versus college

The traditional draft debate is whether it is better to take players out of high school, which are generally characterized as high-risk/high-reward, or players out of college, characterized as safer picks. Keith Law had a nice write-up on the subject here. The following graph shows how many players have been chosen—though not necessarily signed—in the draft from the various levels of school over the span of the data discussed above.

First a quick explanation of the terminology. High school draftees are represented in the chart with HS, four-year collegians with 4Y, junior college players with JC and others—independent league players as an example—with OT. The numbers represent the total number of players of that type picked. Clearly, the majority of players selected were from four-year universities; however, there was still a significant number of high school players taken.

This chart, as it is from an aggregate perspective, sets the stage for the rest of the analysis. Further analysis will slice the above data by team and round. The following table breaks the data out by team, sorted by percentage of players drafted from high school.

Team	HS	JC	OT	4Y
Rays	52%	10%	0%	38%
Dodgers	51%	8%	1%	40%
Angels	50%	14%	0%	36%
Marlins	47%	6%	0%	47%
Braves	44%	20%	1%	35%
Brewers	44%	7%	0%	50%
Royals	42%	6%	2%	50%
Twins	41%	5%	0%	54%
Phillies	41%	5%	0%	54%
Red	39%	5%	0%	55%
Rangers	37%	6%	1%	56%
Pirates	37%	6%	0%	57%
Yankees	37%	3%	0%	60%
Mariners	36%	4%	2%	58%
Expos/Nats	36%	13%	0%	52%
Astros	33%	7%	0%	60%
Orioles	30%	10%	0%	60%
Reds	29%	7%	1%	63%
White	28%	9%	0%	62%
BlueJays	28%	4%	0%	69%
Mets	28%	11%	1%	60%
Indians	26%	4%	1%	70%
Tigers	23%	3%	0%	74%
Cubs	23%	9%	0%	69%
Cardinals	22%	5%	0%	73%
Diamondbacks	21%	6%	0%	73%
Padres	19%	5%	0%	76%
Giants	18%	8%	1%	74%
Athletics	17%	4%	0%	79%
Rockies	15%	7%	0%	78%

My first observation when going down that list was that the teams that were picking the most high school players were located in the states that generally produce a large chunk of the amateur talent. Could it be that those teams were simply able to more efficiently scout high school talent because it was in their back yard?

The observation that quickly counters that theory is that the bottom five is made up of three California teams and one Arizona team.

The final observation is that the A's draft philosophies touched on in Moneyball were held to across the decade investigated, as Oakland drafted the most four-year college players in the league over the ten years studied.

As stated, the previous data was based on the first ten rounds; however, it is also interesting to limit the search to just how teams spend their first-round picks (For the next chart, I will define first-round as the first round plus first-round supplemental picks ). The league aggregate graph looks like this:

On an aggregate level, a higher percentage of high school players are taken in the first round than in subsequent rounds. The same data broken out by team looks like this:

Team	HS	JC	OT	4Y
Phillies	90%	0%	0%	10%
Braves	76%	0%	0%	24%
Angels	72%	0%	0%	28%
Marlins	71%	0%	0%	29%
Dodgers	67%	7%	0%	27%
Mariners	58%	0%	0%	42%
Rays	58%	0%	0%	42%
Twins	56%	0%	0%	44%
Yankees	53%	7%	0%	40%
Rangers	53%	0%	5%	42%
Brewers	50%	0%	0%	50%
Royals	50%	0%	14%	36%
Tigers	46%	0%	0%	54%
Astros	45%	9%	0%	45%
Giants	44%	0%	0%	56%
Indians	44%	6%	0%	50%
Orioles	43%	7%	0%	50%
Rockies	43%	0%	0%	57%
Expos/Nats	36%	7%	0%	57%
Reds	36%	0%	0%	64%
Cardinals	35%	0%	0%	65%
Mets	33%	0%	0%	67%
Padres	32%	0%	0%	68%
Red	32%	5%	0%	63%
Diamondbacks	30%	0%	0%	70%
Pirates	27%	0%	0%	73%
BlueJays	26%	0%	0%	74%
Cubs	21%	7%	0%	71%
White	14%	0%	0%	86%
Athletics	4%	0%	0%	96%

The Phillies stand out the most when compared to the previous list with an astounding 90 percent of first-round picks being high school players after being only moderately high in the overall list. Another team that is interesting to compare is the Giants. In the first ten rounds, they hardly draft any high school players, ranking near the bottom of that list. However, in the first round they have taken a decent amount of high school talent. Finally, the A's, as expected given the information presented in Moneyball, reside at the very bottom of the list.

So far I have shown data aggregated across the first ten rounds and zoomed in on the first round. What about everything in between? Are there any interesting trends buried in that data? The following table looks at league-wide data for each individual round:

Round	HS	JC	OT	4Y
1	44%	2%	1%	53%
1s	42%	2%	1%	55%
2	47%	3%	1%	49%
2s	50%	0%	0%	50%
3	36%	6%	0%	58%
3s	40%	20%	0%	40%
4	39%	6%	1%	54%
5	30%	4%	0%	66%
6	29%	9%	0%	62%
7	29%	8%	0%	62%
8	22%	14%	0%	64%
9	24%	12%	0%	64%
10	24%	10%	0%	65%

There is a clear trend that, as the draft gets deeper into the first ten rounds, teams shift from high school players to four-year college players. The observed gradual decline in high school players is not a surprising result.

As the draft gets deeper and expected bonuses get lower, there is a higher incentive for high school players to attend college and improve their draft stock (along with getting an education). With the incentive there for high school players to attend college, teams have an incentive to draft players they will have the ability to sign and, thus, drift towards players from four-year colleges.

Clearly there is context—that is not explored in this article—behind all of these picks, as the best player available at the time a team picks could clearly be a high school player some years and a college player others. That said, general trends do exist for specific teams as to whether they prefer drafting collegians or high school players, as do general trends from round to round.

Geography

To this point, the focus has been on the type of school the draftee attended, with a minor reference to the states the players came from. This section will look at that geography in additional detail. First, in general, where state are players drafted from? The following table shows the unsurprising top five for players drafted in the top ten rounds.

State	Players Drafted
CA	614
FL	349
TX	325
GA	150
NC	131

The state represented in the table above is the state the player went to college in, not necessarily the one in which he was born. An interesting note from the data is that these five states make up nearly 50 percent of the players drafted in the first ten rounds. If we cut the data to just the first round and the first-round supplemental, the list only changes slightly, with Tennessee replacing North Carolina and the order shifting around slightly.

One final interesting tidbit from the geographical data, the Braves' fondness for Georgia-based players has been discussed before. The data back up the statements, as they have selected 22 players from Georgia over the last 10 seasons with the next-highest total from any other team being nine (the Rangers and Red Sox). They also lead in drafting players from Georgia in the first round with five over the last ten years.

Summary

This article was not meant to be a study on how successful drafting teams went about their business at draft time. It set out to investigate trends in types of players that were being drafted both in relation to types of schools and geography.

To that end, I found the teams on the extremes, especially with their first-round picks, to be very interesting. The A's have spent one lone pick on a high school first rounder in the years investigated—Jeremy Bonderman—while the Phillies used nine of their 10 ten first rounders on high school players. In the geography section, I found the data on the Braves and players from Georgia to be fascinating.

Click here to learn about THT's download subscriptions.]]> Steve Sommer 2011-04-11T08:04:15+00:00 Observations on leverage http://www.hardballtimes.com/main/article/observations-on-leverage/ http://www.hardballtimes.com/main/article/observations-on-leverage/#When:06:52:15 the primary function, of a manger is to place the correct players in the correct situations for the good of the team. Bullpen usage, while maybe not the most important of the playing time decisions, is one that tends to stand out in fans' minds. In an attempt to get our minds wrapped around which players are being put in the correct situations, who is being used in the toughest spots, and other similar questions we will turn to an investigation of leverage index.

Throughout most of the remainder of this article I will be referring to different leverage index bands. I'm defining those bands the same way that Baseball Reference does: high leverage is a leverage index greater than 1.5, medium leverage is a leverage index between 0.7 and 1.5, and low leverage is below 0.7.

The good

The first, and simplest look at pitcher leverage index is to see which pitchers have entered the game with the highest percentage of high leverage situations. The following table was generated from data from 2006-2010. Pitchers needed to have made at least 150 appearances over the five-year span to be included in the list.

Pitcher	High Lev %
Francisco Cordero	61%
Brian Wilson	61%
Joakim Soria	61%
Mariano Rivera	61%
Brian Fuentes	60%
Joe Nathan	60%
Jonathan Papelbon	58%
Trevor Hoffman	57%
Huston Street	57%
Joe Borowski	56%

Despite the way managers handle most closers, using them only in the ninth inning when there is a save situation, this list remains populated with pitchers who are closers. This list in not meant to imply that these closers are being used optimally or even near optimally. There are plenty of times where these pitchers were used in low leverage situations, or there were higher leverage situations in the eighth inning that these guys would have been better suited to pitch in.

The list does say, though, that these pitchers have appeared in a higher percentage of high leverage situations than have their bullpen brethren. As a point of reference, a glut of top setup men over that time period— Scot Shields, Luke Gregerson, Matt Thornton)—are bunched at approximately 50 percent high leverage situations.

This list was built over an extended time period, so it includes many factors including true talent changes, role changes and team changes. With that in mind it might be insightful to look at some of the top single seasons over the same time frame. The following table summarizes the results.

Pitcher	Season	High Lev %
David Aardsma	2010	79%
Francisco Rodriguez	2008	75%
Rafael Soriano	2010	70%
Joakim Soria	2009	69%
Brian Fuentes	2008	69%
Jose Valverde	2007	69%
B.J. Ryan	2008	68%
Mariano Rivera	2008	68%
David Aardsma	2009	68%
Jonathan Papelbon	2010	68%

The table is again a list populated with all pitchers who were closers at the time. It appears that it is possible, depending on the season, to have pitchers appear in high leverage situations 70 to 80 percent of the time, with top setup men appearing on the list in the mid 60 percent range. We will talk about Mr. Aardsma and his appearance on this list twice in some depth a little later.

The bad

The previous two lists have looked at the top of the crop, but what does the bottom look like? Who are the pitchers who have been sent to the mound in the least stressful situations over the past five seasons? First from the perspective of lowest percentage of high leverage situations:

Pitcher	High Lev %
Edward Mujica	19%
Lance Cormier	20%
Blaine Boyer	21%
Rudy Seanez	21%
Clay Condrey	22%
Brandon Medders	22%
Brian Tallet	23%
Jason Grilli	23%
Juan Cruz	25%
Julian Tavarez	25%

and then from the perspective of highest percentage of low leverage situations:

Pitcher	Low Lev %
Rudy Seanez	62%
Edward Mujica	62%
Clay Condrey	61%
Blaine Boyer	60%
Jesus Colome	58%
Jesus Colome	55%
Doug Slaten	55%
Lance Cormier	55%
Guillermo Mota	54%
Brian Stokes	54%

As expected those lists are populated with journeymen relievers. These are the guys who are sucking up innings in blowouts, the infamous long relievers who are the 11th man on a pitching staff (12th or 13th if your manager happens to be Tony LaRussa). These pitchers, in any particular season, are pitching approximately 60 percent of the time in situations that are not vital to that particular game. That is not to say that these pitchers do not have roles, as some pitcher clearly has to pitch in those situations. It would be folly to pay these pitchers anything over the league minimum for the low impact service they are providing.

Young or old?

I thought would be interesting to investigate the relationship between age and leverage. As a Cardinals fan I have watched Tony LaRussa make young pitchers prove themselves before he will put them in a high leverage role. My going-in assumption was that this was probably a fairly common managerial tactic across the league. What do the data say?

The chart depicts on the aggregate how often players of each age are placed in high leverage situations. The generic trend in the data does seem to back up my assumption. Generally speaking, older pitchers are placed in a higher percentage of high leverage situations than younger pitchers. Clearly the trend is not overwhelming, and there are some sample size and selective sample issues on the periphery. Additionally, the talent level for each age needs to be considered and corrected for before any age bias in managers could be concluded with any level of certainty. All caveats aside, it does appear that LaRussa is not the only manager who slightly prefers veterans in high leverage situations.

A few anecdotes

David Aardsma has one of the more interesting reliever journeys, at least when it comes to leverage index, that I found while compiling the research for this article. In 2006 and 2008 he posted some of the lower single season totals for percentage of appearances in high leverage situations, 11 percent and 21 percent respectively. Then, after being traded to Seattle and installed as the closer and showing better peripherals, he promptly posted two of the highest single season totals as seen in the table near the beginning of the article.

One of the more dramatic changes in the other direction is Scot Shields. After posting seasons above 50 percent high leverage appearances from 2006-2008, Shields had injury trouble in 2009 which appears to have led to a higher walk rate and lower strikeout rates. Those lower peripherals, and the true talent level that produced them, led the Angels to use Shields in high leverage in only four of his 43 appearances.

Those are just a couple of examples of how usage can change over time because of changes in scenery, role or talent level. Clearly we could investigate the data and find other such examples. I thought those two were interesting because of how high they two ranked on the various lists.

Summary

This article merely scratched the surface on reliever usage. There are many more things that could be done with the data, correlating salary with leverage for example. That said, I found the fact that the top 10 high leverage percentage pitchers were all closers. I expected to find a few more setup men higher on the list. Also, while the age graph was not surprising, I did think it was interesting and worthy of additional investigation.

Click here to learn about THT's download subscriptions.]]> Steve Sommer 2011-03-28T06:52:15+00:00 Adjustments by the Freak http://www.hardballtimes.com/main/article/adjustments-by-the-freak/ http://www.hardballtimes.com/main/article/adjustments-by-the-freak/#When:09:01:15 Tim Lincecum is one of the top pitchers in Major League Baseball. He has ranked first, third and fifteenth in the league in Field Independent Pitching (FIP) over the past three seasons, respectively.

Despite the consistent success, Lincecum has had some underlying changes to his repertoire. He burst onto the scene in 2007 and 2008 sporting a 94-mph fastball. Since then, however, his fastball velocity has been in steady decline.

The following table summarizes the loss of velocity by year using BIS data for fastballs, PITCHf/x data for fastballs and—to take out some of the classification issues—Tango's trick of looking at the top 25% of pitches in velocity according to PITCHf/x.

	BIS	Pfx	top 25%
2008	94.1	94.1	95.8
2009	92.4	92.4	93.7
2010	91.3	91.3	92.6

For some additional pieces on Lincecum and velocity, Dave Cameron has previously discussed it here and here.

Investigating an effect

One of the first logical questions when presented with the declining velocity data is, "How big of an effect, if any, does the velocity loss have on performance?" Should it be something we pay attention to this coming season?

Mike Fast found that in general a starter losing one mph will see his runs allowed rise by approximately 0.25 runs, but how can that generalized result be applied to a specific pitcher?

To offer insight into that question, I will introduce two specific data points:

1. Run value per pitch by velocity
2. An at-bat-level metric

What happens if we take a granular view and look at success on a pitch-by-pitch basis? The following graph plots run values per 100 fastballs (rv100) versus velocity over the 2008-2010 seasons. Lower numbers (more negative) equate to a more effective pitch.

There are more variables at play here than simply velocity—sequencing and location to name a few. However, this chart does seem to be evidence in favor of velocity being important to the success of Lincecum's fastball, at least up to a point.

Velocities between 93 mph and 97 mph all have better run value results than do those at 91 mph or 92 mph. The 91-92 mph range is one Lincecum spends a lot more time in now than he did in 2008. One key distinction to make is that the analysis to this point only focuses on fastballs. The rv100 chart merely shows how the fastball is affected; it doesn't show how other pitches, or entire at-bats, are affected.

In an attempt to garner insight into that very question, I binned at-bats by counting the number of 94-plus mph fastballs in an individual at-bat and then calculated the weighted on-base avearge (wOBA) against for each bin. The following table summarizes the results:

94+ fastballs	PA	wOBA
0	1977	0.279
1	359	0.288
2	197	0.229
3	95	0.187
4	55	0.265
5	26	0.232
6	9	0.377
7	6	0.657

Again, this table generally points to velocity being helpful to Lincecum's cause. There are some sample size issues given the low number of plate appearances in some bins, but in general it appears that some 94-plus mph fastballs are better than none. As a whole, the wOBA against when there is at least one 94-plus mph fastball is 0.260 compared to 0.279 when there is none.

The combination of the two presented pieces of data points us towards at least a small correlation between velocity and success for Lincecum, both on the individual pitch level and on the at-bat level.

Adjusting in light of the evidence

Now that I have shown some evidence towards at least a slight correlation between fastball velocity and fastball success for Lincecum, it follows that the next question to ask is, "What adjustments has Lincecum made to maintain success?" The clearest adjustment would be a change in pitch usage as a lesser fastball might not be used as frequently. The following chart summarizes how frequently Lincecum has used his pitches over the three-year span being investigated.

Clearly, the chart shows that he has been less reliant on his fastball as his velocity has decreased, with a good portion of the slack being picked up by his curveball. These data point to Lincecum being aware that his fastball is less effectiveness with less velocity, and he is adjusting his approach as required to get hitters out.

As an addition to the overall usage, I thought it might be telling to see how Lincecum has used his fastball by count over the three-year span. The following table summarizes that data, with the percentages representing the percent of fastballs thrown in that count.

Count	2008	2009	2010
0-0	72%	67%	65%
0-1	61%	48%	43%
0-2	49%	29%	44%
1-0	75%	66%	64%
1-1	61%	46%	41%
1-2	44%	39%	33%
2-0	88%	85%	84%
2-1	82%	64%	54%
2-2	47%	35%	31%
3-0	98%	100%	96%
3-1	88%	86%	80%
3-2	69%	42%	42%
All	65%	56%	53%

The most telling lines of the table are the 3-2 line and the 2-1 line. Counts where Lincecum had previously been throwing predominately fastballs are now mixed. Sure, some of that shift is probably due to game theory between hitter and pitcher and hitters making adjustments, but some of the change is likely a realization that, if a hitter knows the fastball is coming, it is less likely to be successful than it has been in the past.

Another way to combat declining velocity would be with location. That said, the only point worth mentioning that I could find was that in 2010 he threw approximately three percent fewer fastballs in the generic strike zone than he had in 2008.

Summary

Lincecum's fastball effectiveness and general effectiveness when looked at on an at-bat by at-bat level is affected by velocity. That said, Lincecum has made some adjustments in the way he approaches hitters that seem to have counteracted some of the diminishing velocity. What does this mean for this year? Clearly, he has shown an ability to succeed with the altered approach, but it will still be something to keep an eye on as the season plays out.

Click here to learn about THT's download subscriptions.]]> Steve Sommer 2011-03-07T09:01:15+00:00

Round	HS	JC	OT	4Y
1	44%	2%	1%	53%
1s	42%	2%	1%	55%
2	47%	3%	1%	49%
2s	50%	0%	0%	50%
3	36%	6%	0%	58%
3s	40%	20%	0%	40%
4	39%	6%	1%	54%
5	30%	4%	0%	66%
6	29%	9%	0%	62%
7	29%	8%	0%	62%
8	22%	14%	0%	64%
9	24%	12%	0%	64%
10	24%	10%	0%	65%

Round	HS	JC	OT	4Y
1	44%	2%	1%	53%
1s	42%	2%	1%	55%
2	47%	3%	1%	49%
2s	50%	0%	0%	50%
3	36%	6%	0%	58%
3s	40%	20%	0%	40%
4	39%	6%	1%	54%
5	30%	4%	0%	66%
6	29%	9%	0%	62%
7	29%	8%	0%	62%
8	22%	14%	0%	64%
9	24%	12%	0%	64%
10	24%	10%	0%	65%

Round	HS	JC	OT	4Y
1	44%	2%	1%	53%
1s	42%	2%	1%	55%
2	47%	3%	1%	49%
2s	50%	0%	0%	50%
3	36%	6%	0%	58%
3s	40%	20%	0%	40%
4	39%	6%	1%	54%
5	30%	4%	0%	66%
6	29%	9%	0%	62%
7	29%	8%	0%	62%
8	22%	14%	0%	64%
9	24%	12%	0%	64%
10	24%	10%	0%	65%