Wednesday, July 30, 2008

Optimal Roster Construction, part 1,401

The recent roster moves have made it hard to tell what the Orioles are doing with the pitching staff right now, but in the end, there are 13 pitchers on the roster, with only four of them - Dennis Sarfate, Jeremy Guthrie, Garret Olson, and Daniel Cabrera - in the starting rotation.

For the time being, at least until the Orioles make a move to get a fifth starter on the team, there are nine relievers down in the pen. Nine!

Is there ANY need to ever have that many guys in your bullpen? Right now, with Juan Castro slowly taking over the starting SS job, the bench consists of Brandon Fahey, Jay Payton and Guillermo Quiroz. It's been discussed many times before, but that's just an awful bench. Payton is the closest thing to a hitter in the bunch, and he's currently sporting a .653 OPS. The biggest piece of strategy that Trembley can employ in the late innings of a close game is to pinch-hit Payton for Castro, move Payton to LF as a defensive sub for Luke Scott, and insert Fahey as the new SS. That's not really all that viable, but hey, they've made their choice to go for broke on having pitching and defense (supposedly) over any type of offensive flexibility.

But hey, Juan Castro smacked a homerun today, so maybe we've found our SS for the next 25 years? Castro did manage to beat Fahey in the race to get to 1 HR first, even with the head start he spotted Fahey. I'm actually shocked because looking at Castro I thought that even if he made good contact he'd just slap the ball to the outfield at best.

Speaking of Sarfate (we were earlier at least), his first start wasn't a total disaster. Of course I'm grading that from the perspective of "Sarfate won't last three innings" as a baseline. He actually lasted four, striking out five, and allowing just five base runners. Unfortunately all five of those guys (three of whom reached on walks) came around to score so that made for an ugly line. The guy who didn't help himself at all was the guy who just lost his starting spot, Brian Burres. He came into the game in the fifth with the game still in reach (5-2 Yankees) and after pitching a fine fifth inning, got only one out in the sixth before being lifted, eventually being charged with four runs. His roster spot is getting to be pretty tenuous as that ERA climbs close to 6.00 and he struggles in his new role as a long reliever.

Next up could be Hayden Penn or Brad Bergesen. Possibly both, as the Orioles will be needing both a fifth starter and a replacement for Sarfate in the coming weeks. Bergesen just threw a shutout for Bowie in his last start to get his 13th win of the season, and Penn had a nice start down at Norfolk. Neither of these guys are among the best arms down on the farm, but they may be the most ready, so they'll be pressed into action soon enough. Here's hoping Bowie's Jason Berken gets a look too. He's a guy I think has been a bit underrated this season.

Wieters Watch

Matthew "Babe" Wieters and his big blue Ox took it to New Britain last night to the tune of 4-5 with a double and another HR, his 21st of the year. That brings his current line up to a modest .350/.449/.600 with Bowie, .347/.448/.584. With that monster night he actually nudged his AA OPS back ahead of the OPS he had in A ball. Wieters currently ranks fifth in the minors in OPS among those with at least 300 PA, 13th among those with 150 PA.

Monday, July 28, 2008

Prospect Watch (7/29/08)

98 games (Frederick/Bowie): .341/.446/.572
29 games (Bowie): .330/.440/.560

Wieters drew four walks last night, including two of the intentional variety. That brings his BB/K ratio for the season to 62/63, which shows a pretty damn terrific sense of the strike zone.

Speaking of prospects that have shown great plate discipline, 2008 third round draftee LJ Hoes has done a nice job for the Gulf Coast League Orioles, drawing 17 walks and striking out just 9 times in 99 PA on his way to .333/.449/.444 line. Hoes is still in the very early days of his development, not even having made his way out of the complex leagues yet, but he's had an encouraging debut season so far. Hopefully the Orioles will challenge at least at Bluefield in the last month of the season.

Saturday, July 26, 2008

The Adam Loewen Career Retrospective Bonanza

It was announced last week that after suffering yet another injury, Adam Loewen's pitching career is finished. He'll try to make himself into a 1B/DH/OF to salvage his career, but he won't have nearly the upside he did as a pitcher.

Loewen was drafted fourth overall in the 2002 draft out of high school, three spots above Prince Fielder and eighteen spots above Jeremy Guthrie. It took forever for the Orioles to actually get Loewen signed, long enough for him to play one season of JUCO baseball. The present rules say that all draft picks must be signed by August 15th, but back then (back then? it was only six years ago) teams had a full year to sign draft picks. The O's finally got it done by giving Loewen a Major League contract worth over $4M in late May of 2003.

What that "Major League" adjective meant is that Loewen was immediately put onto the 40 man roster which caused him to be optioned to the minors every season, in effect accelerating his development schedule. In the end, the injuries are what did Loewen in as a pitcher, but he was, in my opinion, rushed through the minors too hastily based on how he actually performed.

Loewen was the prototypical great-stuff-no-control guy in the minors, walking an absurd 67 batters in 93 innings as a 20 year old in Delmarva and Frederick in his first full season. Spending the next season exclusively with the Keys, he walked 86 and hit 14 batters in 142 innings.

Loewen showed improvement at Bowie in 2006 and was soon pitching in Camden Yards. It was a fairly memorable debut in the rotation for Loewen as in his first four starts he was matched up with Randy Johnson, Roy Halladay in back to back starts, and then Tom Glavine in his fourth start. Loewen struggled, and then after a brief stint at Ottawa (his AAA debut), he came back and finished out the season in the Orioles' rotation, finishing with an extremely promising 13 start stretch. In that baker's dozen of starts, Loewen won six times, posting a 4.44 ERA with 66 K, 33 BB, and just 3 HR in 73 innings of work.

It was a fine finish for a then 22 year old Loewen, but the rest is history as injuries marred his 2007 and 2008 seasons. Loewen would make just 10 more starts before calling it quits as a pitcher. Sadly, Loewen won't go down as the biggest recent first round draft bust for the Orioles, or even particularly close - see e.g. Chris Smith, picked seventh overall in 2001.

Now he begins the long road to becoming a major league hitter. While the recent conversion of Rick Ankiel to a bona fide Major League hitter probably gives some the impression that Loewen could do the same, the odds are long. Even if Loewen were a third-round caliber pick, well, only five of the fifteen hitters drafted in the third round in 2002 have made the bigs (one of them being the immortal Val Majewski of the O's). Only Curtis Granderson, and to a lesser extent, Elijah Dukes, have done anything of note, and that's before accounting for the six years of development that Loewen has missed which others with similar skill did not.

So it's been real, Adam. See ya in 2012?

Friday, July 25, 2008

Does Cabrera Have A Future?

It's hard not to be pessimistic about Daniel Cabrera. Even before being tagged for 14 tuns in his last two starts, Cabrera had yet to show much tangible improvement in his fifth season aside from a nice string of games in May and July which included two complete game victories over the Royals. If anything, Cabrera had gone the opposite way, missing fewer bats and in general looking less overpowering than he had in the past.

Let's look at some of the key Pitch Data Summary from baseball-reference relating to Cabrera's trends which show how "overpowering" he has been over the past three seasons.

%Strikes Thrown: 57%-->58%-->59%
%1st Pitch Strikes: 52%-->55%-->57%

Those are both nice trends for a pitcher that has notably struggled with his control as a young guy. Unfortunately, the average pitcher throws 62% of his pitches for strikes and 58% of his first pitches for strikes so even with steady improvement since 2006, Cabrera is still below average at throwing strikes and at getting ahead in the count.

%Strikes Swinging:18%-->14%-->10%
%Pitches StrS: 10%-->8%-->6%

The first number is the percentage of Cabrera's strikes that have been of the swinging variety while the second number is the percentage of all Cabrera's pitches that have been swinging strikes. That's quite a marked decline. Cabrera is missing fewer and fewer bats, a fact you would expect with his declining strikeout rates. But maybe this, combined with his slightly improved command, means that Cabrera is getting more called strikes, perhaps on the corners.

%Strikes Looking: 30%-->27%-->28%
%Strikes In Play: 26%-->31%-->34%

Nope. Cabrera is actually getting fewer of his strikes as called strikes when compared to 2006. He's simply become more hittable with an 8 percentage point increase in the number of his credited strikes being on balls put into play. The average pitcher, meanwhile, gets 14% of his strikes swinging and 31% on balls in play compared to 10% and 34% in 2008 for Cabrera. One more number.

%Contact: 75%-->81%-->87%

When hitters do swing, they're making contact some 87% of the time, far above the league average of 80%. Once again, Cabrera is simply not missing bats which, when combined with subpar (4.14 BB/9, 14 HBP), is a bad way to go.

Now to look at some fangraphs data. Simply put, Daniel Cabrera is fairly predictable. He's going to throw a heavy dose of fastballs with the occasional slider mixed in. Of all pitchers with at least 80 innings this season - 123 in total - none have used their fastball as often as Cabrera who has done so on 83.6% of his total pitches. Fangraphs says that most (14.1%) of the rest of Cabrera's tosses have been sliders, with the occasional change up being used, although Josh Kalk's Pitch F/X (more from there later). says that all of the surplus non-fastballs are classified as

All told, he's throwing tons more fastballs than he did in his best season, 2005, going from about 65% fastballs to nearly 84% fastballs. He has also, according to fangraphs, lost about three mph on his fastball, dropping from an average of a bit over 96 mph to just over 93 mph.

His slider is used most often as his out pitch. The three most likely situations for you to see Daniel Cabrera throw a slider is on 1-2, 2-2, and 3-2 counts. Here is a nifty chart which plots the movement on Cabrera's pitches:

What you're seeing here is how Cabrera's pitches move relative to a pitch thrown with no spin or movement. That green cluster of triangles are Cabrera's sliders. And as you can see, they're clustered somewhere very close to the origin of the graph. Looking at the numbers, Cabrera's slider has little movement, clocking in at (.79, .89) inches in the x and y directions.

Now that's not really all that good. Now I'll be the first to admit that I'm really just starting to learn about how to interpret these pitch F/X graphs and data, but it seems to me that a slider that doesn't really move is going to be troublesome, especially when

a) you use it about once every six pitches and
b) it's really the only thing you throw besides a fastball

Here are a few good pitchers who rely heavily on their slider, and the graphs of their pitch movement, C.C. Sabathia and Jake Peavy.

Sabathia's average slider (-5.04, -0.41) and clocks in at 81.2 mph while Peavy's slider breaks at (5.83, 0.92) while averaging 83.1 mph. Cabrera, meanwhile is at 82.1 mph on his velocity. It shouldn't come as any surprise that Daniel Cabrera's slider isn't as good as the sliders of two of the best pitchers in baseball, but looking at the different in movement is striking. Cabrera's fastball, at least in movement and velocity, does stack up well against that of Peavy and Sabathia, but those pitchers rely on the fastball far less, both having multiple other plus pitches.

For one last piece of information, here is Cabrera's pitch type-movement chart from last season.

Cabrera, last season, got about four inches of movement on his slider. Of course, that didn't lead to results that were notably better than this season's, but qualitatively, it seems logical that a slider with more movement is a key, especially for a two-pitch pitcher. [One note - the 2007 data was based on only about 650 pitches, a small sample of Cabrera's total work. The 2008 data is far more complete.]

So does Cabrera have a future? Unfortunately it would seem unlikely that he'll ever be even a consistently average starter. He's still under team control until after the 2010 season so if he can avoid completely falling off a cliff, he'll have a chance to stick around, at least as a fifth starter, given his cheap price tag. But 138 starts into his Major League career, it's definitely time to stop thinking of Cabrera as a young guy with great stuff who only needs to harness his immense potential. It's just not the case anymore.

Sunday, July 20, 2008

First Half Value: Pitchers

I'm back with the second part of my first half team value breakdown with an analysis of the pitching staff. This one should be easy, right? We have ERA which already gives us our number of runs allowed, so all we need to do is compare that to the average to get a runs above average estimate.


ERA = pitching + defense. We've already attempted to account for defense in an earlier post, so now we want to attempt to isolate the pitching component. So how do we do that? We use a defensive-independent pitching metric, in this case FIP (Fielding Independent Pitching), a metric that relies solely on those things a pitcher has great control over (K, BB, HR). Lucky for us, FIP is both

a) readily available and
b) scaled to "look like" ERA

So that's nice. Now what we'll do is park adjusment to account for OPACY playing as a slight hitters' park. Luckily, Baseball Reference does that already, telling us that the park adjusted league average ERA for pitchers who play half of their games in OPACY is 4.22 rather than 4.14. What we'll also do is adjust this to scale to Runs Allowed rather than Earned Runs Allowed (as ER vs R is often an arbitrary and nonsensical decision) by adding 0.46 (lgRA - lgERA) to our FIP values. We'll also break this down into starters and relievers; starters have an average park adjusted RA of 4.75 this season while relievers have an average park adjusted RA of 4.24.

Finally, we'll use this equation to convert our numbers into an estimate of runs above or below average for each pitcher:

RAA = ((lgFIPR - FIPR)/9)*IP

where lgFIPR is FIP, adjusted to scale to RA, adjusted for park, and customized for each player based on his mix of starting and relief innings. FIPR is each player's FIP scaled to RA.

Now onto results. Warning: Not for the faint of heart.

Bradford 5.00
Johnson 3.98
CabreraF 1.17
Guthrie 0.90
Castillo -0.01
Cormier -0.16
Bierd -0.64
Sherrill -1.20
McCrory -1.35
Albers -3.23
Bukvich -3.34
Sarfate -3.45
Walker -4.65
Burres -6.69
Olson -6.70
Loewen -7.41
Aquino -9.05
Liz -11.11
CabreraD -13.55
Trachsel -15.25

Oh that's not good. No that isn't good at all.

From a defense-independent perspective, Chad Bradford and Jim Johnson are leading the way, while Jeremy Guthrie is barely above average in 135 IP so far this season. The rest of the team have somewhat obscene values on the negative side of the ledger.

First let's talk about Guthrie. Simply, his peripherals have lagged behind his ERA for two seasons now. His strikeout rate is around average, he gives up a below-average number of walks, and he allows homers at a clip that rates as a bit above average. The key for Guthrie is that he has been much better (.652 OPSA) with runners on base than he has been with the bases empty (.722 OPSA). His BABIP is also low, sitting at .272 for a second straight year, which is interesting now that the sample is about 1300 PA. So, the book here is that Guthrie is stilly likely to regress, but then given that we're working with a decent sample of data here, it's also possible that his peripherals are underrating his true talent.

Either way, there isn't really a ton of dispute about Guthrie being an above average starter, which does have a lot of value, especially when accompanied by a price tag of less than $1MM. On the other hand, the rest of the starters on the team have been absolutely abysmal. Daniel Cabrera has put together some good results so far, but his collapsing strikeout rate and not-so-improved command means he's having a very poor season. Olson and Liz, the two young guys in the rotation, have both been awful, which has to be very discouraging.

Still worse, bullpen guys like Albers and Sarfate that have been fairly successful this season don't fare out too well under this analysis. Albers doesn't get enough strikeouts and Sarfate walks way too many hitters. The result? Both have contributed below average performances.

All told, the pitching staff comes in at an amazing 70+ runs below average. You could just look at the team-by-team FIP numbers and tell that the Orioles weren't going to rate favorably as they are the worst team in the AL in FIP, by a fair margin. So we'll see how this shakes out in a few months. I suspect the analysis will be similar, in which case we'll go back to our annual topic of overhauling the entire staff.

Wednesday, July 16, 2008

First Half Value: Position Players

Nearing the 100 game mark, let's take a look at the first half of the season and try to figure out how much each Oriole has been worth this year in terms of runs saved and runs allowed. Today I'm going to evaluate the position players, starting with their offense, which I'll evaluate using BattingRuns, a linear weights style system available at Baseball Reference. Later in this post I'll look at their fielding, and hopefully I'll look at the pitchers in a subsequent post.

Here are every Oriole position player ranked by Batting Runs.
Markakis        21.3
Huff 17.1
Roberts 15.7
Scott 8.3
Salazar 0.7
Moore -0.3
Torres -1.1
Cintron -2.1
Millar -2.4
Jones -3
Fahey -3.9
Quiroz -5.2
Mora -5.6
HernandezL -5.6
Payton -7.4
HernandezR -9.9
Bynum -13.2
TOTAL -1.3
First of all, the values here are the number of runs above or below average a player has been offensively this season. A value of '0' indicates that a player has performed at exactly the league average. This number has been adjusted based on league and park, but it does not account for positional differences (we'll look at that later), nor does it account for the difference in competition between leagues. It is a counting stat, which means that the absolute value of the statistic will tend to increase as playing time does.

The results here should not be particularly surprising. The entire team rates as being very slightly below average offensively with about 98% of the positive contribution coming from Luke Scott, Brian Roberts, Aubrey Huff, and Nick Markakis. In fact, the ONLY other player of the remaining 13 that has generated positive value is Oscar Salazar who had just 18 PA before being send back to Norfolk.

The SS production has been terrible, as expected, with a combined contribution of about -25 runs from the four primary SS options. Freddie Bynum and his .444 OPS over 121 PA has been the worst offender at -13.

So that's interesting. Now let's look at fielding value. This is based off of The Hardball Times' Revised Zone Rating (RZR) and converted to a runs saved estimate using the method used by Justin over on his excellent Reds' site (I also used his method for evaluating catcher defense) to convert to a "+/-" number of plays above or below average before converting to runs using the fact that 0.8 runs are saved for each play made.

What IS included in these rankings: runs saved on balls in and out of zone for non-catchers, relative to the average at their position; runs saved on stolen bases and errors for catchers.

What IS NOT included in these rankings: runs saved based on throwing arm for outfielders or double plays for infielders; runs saved on errors for non-catchers; runs saved on passed balls and wild pitches for catchers.

I hope to expand these numbers at the end of the season by including some of those other aspects of defense, but for now, I'm more interested in a quick-n-dirty look at defensive value.

Markakis 7.49
Roberts 5.63
Jones 2.80
Bynum 2.46
Quiroz 1.20
Millar 0.62
Huff 0.44
Moore 0.40
Torres 0.37
Cintron 0.13
Salazar -0.66
Fahey -0.87
Payton -1.33
HernandezL -4.03
HernandezR -4.22
Scott -6.07
Mora -15.24

First, a note. Ideally, we want to have a larger sample size of data. And even more ideally, we would want to use UZR (Ultimate Zone Rating) rather than RZR as UZR takes a much more granular approach to analyzing the data, or at least another more advanced defensive system. However, RZR does bring us a reasonable approximation of defensive value and indeed correlates very strongly with UZR.

So, how about Melvin Mora? He turns in an absolutely terrible performance at over 15 runs below average. This is mainly as a result of making 15 fewer plays out of zone than would be expected from the number of chances he has had in his zone this season. When someone says that Mora has no range anymore at age 36, it's hard to look at these numbers and argue. He'll likely regress towards the mean somewhat, but there's a good chance that he is the worst fielding regular at 3B in the league.

Nick Markakis rates as the teams best fielder, even without the benefit of accounting for his excellent arm. Adam Jones also rates very well, as we would expect, though perhaps a touch lower than you would guess.

And going back to SS once again, look at those defensive values! The Orioles aren't trading offense for defense; they're punting offense just to get average defense. The numbers rate Bynum as a plus fielder, but his contributions are more than zeroed out by Luis Hernandez poor performance in the field. Is there any question he was the worst Opening Day starter in the league?

[To be fair, Hernandez' error and double play rates look solid at a glance, so perhaps factoring those into the analysis would push him back toward average. Still, average fielding from Luis Hernandez means he isn't particularly close to being a major league starter.]

So now let's add these two values up to get a complete rating system for all the Orioles, right? Nope. We still have to make a positional adjustment to account for the fact that Adam Jones playing CF is more valuable than Aubrey Huff playing DH. That's what our numbers so far don't show. There are a number of run values that are floating around the internet for positional adjustments, but I'll just a variation on the one floating around on Tom Tango's blog.

Markakis      26.16
Roberts 21.88
Huff 10.82
Jones 2.03
Moore 0.12
Salazar -0.04
Torres -0.85
Scott -0.91
Cintron -1.32
Quiroz -2.54
Fahey -4.05
Millar -6.99
HernandezL -8.70
Bynum -9.57
HernandezR -9.58
Payton -10.07
Mora -20.35
So there it is, our midseason estimate of player value, in terms of runs above or below average.

Melvin Mora has been, by far, the worst player on the team, in terms of cumulative value. While folks talk about the need to upgrade SS (total of about -24), the situation is almost as bad at 3B with Mora accounting for a -20 at the hot corner. That will happen when you play a power position and don't provide offense or defense to the team.

Another interesting observation is that after adding defense and our positional adjustment to the equation Luke Scott moves to the negative side while Adam Jones moves to the positive side. I think this is one of the areas where crunching the numbers is really instructive. It's sometimes hard for folks to look past a difference of over .100 points of OPS that exists between Scott and Jones, even if those people are fully aware that Jones plays CF very well and Scott plays LF mostly adequately. It should also be noted that Scott has been platooned somewhat with Jay Payton and also plays DH occasionally, leaving him far behind Jones in PA as well as innings in the field.

Did we learn anything here? I'd like to think we did, but if not, there are plenty of pretty numbers to look at. Stay tuned for more updates on this. I'm planning to at least do a final results post when the season is over.

Markakis, Year 3

Even as the Orioles slip further away from the elusive .500 mark, the season still has its small pleasures. Chief among them is watching the growth of Nick Markakis into a legitimate star caliber player. Now in his third season, Markakis has shown improvement in many areas of his game.

First let's look at the basic Markakis lines over the past two seasons. In 2007, Markakis went for a .300/.362/.485 line (121 OPS+) before bumping that up to .299/.401/.492 (139 OPS+). What has been impressive is that not only have Markakis' raw rate stats shown a jump, they have done so in as the run environment has become more receptive to pitchers in the AL this season. AL scoring has dropped by about 0.27 runs - a not insignificant drop - relative to pre-All Star Break levels from 2007.

By Batting Runs (a linear weights method), Markakis rates as +21.5 runs above average, as compared to +19.5 RAA last season, even though he has 300 fewer PA this season. That is important because lwts is a counting stat that, all things equal, accumulates as playing time accumulates.

The most striking difference in Markakis rate stat line is the jump in OBP, especially considering that his AVG has been nearly identical. His BB% (measured per PA) has jumped from 8.7% to 14.3%, an enormous jump, and good for 18th in all of baseball, 9th in the AL. Markakis has also increased his K% (20.3%), which is a concern, but his strikeout rate is not especially concerning. Among the top 20 in BB%, he ranks 10th in BB/K ratio at 0.80, a well above average ratio.

The rise in strikeouts might ordinarily forecast a drop in AVG as less balls in play mean less opportunities for hits. But what Markakis has done in conjunction with the increase in BB and K is to hit more line drives. His LD% has skyrocketed to 22.9%, which fits in well with his BABIP of .346 (a quick "rule of thumb" is that BABIP = LD% + .120). Markakis is becoming a bit more selective at the plate (swinging at just ~41% of pitches this season as compared to ~45% last season) and has simultaneously hit the ball harder when he does swing.

Markakis' defensive reputation has been that of a Gold Glove caliber RF for some time, at least among Orioles fans, but the numbers did not agree. At least not the RZR numbers of The Hardball Times, which rated him a very poor 16th out of 20 qualifying RF last season (although he did have an excellent 45 OOZ plays). This year those numbers peg him as being 5th of 19th in RZR as well as tied for 4th in OOZ plays, a strong case for being named the best defensive RF in the AL. Of course, we should take note of the small sample size involved in both seasons (as fielding statistics, even the best, have more year to year noise than do hitting statistics), and it may be correct to conclude that Markakis deserves to be ranked as just slightly above average in the field.

It has been interesting to see the strides that Markakis has taken this season, although it is probably best to be somewhat cautious while seeing how he finishes the season. The next step will be to add more power, although even if he just stays to his first half pace, he will end up with about 25 HR and 45 2B, very strong totals when combined with a .400 OBP. Here's to hoping