21 February 2016

The Last Item In The Pitching Display Case: Yovani Gallardo

Yesterday, we learned that the Orioles were done hemming and hawing over retaining their 14th overall selection in the 2016 draft.  They would willingly discard it in order to acquire Yovani Gallardo, a pitcher whose performance does not match what front offices seem to think of him.  In other words, a pitcher who is considered a middle of the rotation arm is one that clubs are typically willing to sacrifice a draft pick in addition to offering a sizeable bundle of cash (see Ian Kennedy, Jeff Samardzija).  Instead, the pitcher who often struggles is one who has accomplished quite a bit, but whose repertoire has become highly questioned (see Ubaldo Jimenez, Ervin Santana).

Back in the fall, I used the BORAS model to project future contracts.  BORAS is a relatively simple model.  It looks at several years of data as well as age.  It does not consider Qualifying Offer status.  BORAS pegged Gallardo for a four year deal at 58.4 MM.  Last I heard, Gallardo was trying to secure himself a three year deal for 45 MM, which is in the ballpark.  That would be a 15 MM annual salary as opposed to a 14.6 MM annual salary.

At this point, it might be good to compare different pitchers projected by the BORAS model:

Annual Salary
BORAS (MM) Actual
Jeff Samardzija 15.8 18
Scott Kazmir 15.6 16
Yovanni Gallardo 14.6 12
JA Happ 14.4 12
Hisashi Iwakuma 13.5 12
Mike Leake 13.4 16
Ian Kennedy 9.6 14

Of the pitchers who are considered similar by BORAS, Gallardo is rather reasonably priced.  Not shown here, is that besides earning a rather reasonable salary, Gallardo is also only on the hook for three years while pitchers like Leake, Samardzija, or Kennedy are locked in for five years.  However, JA Happ certainly makes more sense as he would have provided the club with a left handed starter and not resulted in a lost draft pick.  But just to reiterate, Gallardo's salary is quite reasonable (though the lost draft pick certainly adds a wrinkle and I will get to that later).

A while back I also put together a comparison model for Gallardo, which differs from the BORAS model in that it looks forward beyond the off season market.  I never published it because I figured it would be silly for the Orioles to give up a mid-range first round pick on a relatively forgettable arm.  I read the market wrong on that it seems.  Anyway, the comps at his age for a variety of key characteristics (e.g., walk rate, strikeout rate, peripheral runs allowed indices, handedness) were: Brad Radke, Carl Pavano, Todd Ritchie, James Baldwin, Pedro Astacio, Jon Garland, Jeff Suppan, Ramon Ortiz, Matt Morris, Jason Johnson, Jake Westbrook, and Kyle Lohse.  Pedro Astacio was his closest match.

That group led to this five year mean projection, one year beyond the contract's option year:
IP
K
BB
FIP
WAR
30
176.2
113
53
4.45
1.7
31
157.2
95
49
4.47
1.5
32
166.2
93
50
4.60
1.3
33
168.0
92
53
4.81
0.9
34
161.0
76
50
4.97
0.5

Next, here are the low, mean, and high expectations by WAR:
WAR
Low
Mean
High
28
1.3
1.7
2.5
29
0.7
1.5
2.3
30
0.0
1.3
2.7
31
-0.4
0.9
2.3
32
-0.9
0.5
1.9

Part of the decrease is due to the injury rate of pitchers.  As is pretty common, injury rate is quite low in the first year of a contract.  For this comp population, 8% of the pitchers were unable to provide 100+ IP.  Years two through four increases the attrition rate to 40% and year five has a 60% loss.  If healthy, the contract could actually be a modest boon,  The chance that Gallardo stays healthy and productive for the next three seasons is 33% and 13% through the option year.  Purely on a coincidence, 33% is the historical average during the 2000s for a drafted player in the teens to turn into a regular which is what the Orioles gave up to sign Gallardo.

That draft pick, of course, is the cost associated with Gallardo that is the most difficult to pin down.  Based on data science, a mid teens draft pick is worth about 20 MM.  That number comes from the odds of a player panning out, cost controlled surplus value, and decreased value based on that value coming years down the line.  In practice, teams appear to value these picks about half of what the data science suggests they are worth, so about 10 MM in market terms.

With the Gallardo deal, the Orioles saved ~10 MM on what BORAS thought Gallardo's salary should be.  The Orioles saved additionally by only having to offer a three year deal, while BORAS suggested four.  That probably saved the club another 7-9 MM.  In other words, the Orioles appear to have gotten true data science value in signing Gallardo.  However, if we look at it from a comp model perspective, he should have earned 3/33 and the only value saved is on not having to ensure a fourth year.  So, that would be a savings of 7-9 MM which is closer to what market value of picks suggests.  This paragraph is somewhat confusing.  The take home is basically, this is relatively fair.

However, I do not think Gallardo is considerably better than what the Orioles already have shuttling between Baltimore and Norfolk.  Additionally, the minor league system is nearly bereft of talent and an additional pick would be helpful to make the future look a little brighter.  Some have suggested that due to the Orioles history of poor development that the club should value their draft picks less, which is a weird perspective to have.  Personnel turnover has occurred regularly over the past couple decades and poor performance is not hereditary.  The logo on the polo shirt does not provide automatic deficiencies.

From 2001-2010, the Orioles were below average in developing first round talent, but not by much.  This range had a wide range of competency, from the wreck of the Syd Thrift era to better organizational efforts of Jim Beattie, Mike Flanagan, and Andy MacPhail.  Over the course of these years and considering the probability of developing regular starters by draft pick, the Orioles would have been expected to produce 4.3 starters.  Instead, they developed three: Nick Markakis, Matt Wieters, and Manny Machado.  There is nothing exceptionally bad or good about that.  I think that there is a "feeling" of poor development of first round picks that simply is not shown well in that actual production of their draft selections.  Keep in mind, the decision to focus greatly on pitchers also plays a bit of a role on the below average success rate as pitchers are less likely to develop into starters than position players are.

Anyway, even if the Orioles have difficulty developing their own drafted talent, it helps picking higher up in the order where evaluation and development is often easier to accomplish.  I think that argument holds no water.  However, again, the play here is for the next couple years before the core gets old and Manny leaves for a 400-500 MM deal somewhere else.  The draft pick would have no bearing on this window beyond a chance it could be used as trade bait.  I know if my contract was coming up after 2018, I would not value these picks so greatly and would be looking to see how I could swing some success into another position.

8 comments:

Roger said...

I'm glad the statistical analysis made the deal come out "fair", because that could make the Fowler deal look even better (the last analysis showed acquiring both to be fair even if Gallardo got 3/40). One note in the "organizations seem to think ..... instead of what the data shows" department. Apparently, the O's as well as others don't fully believe in xFIP as a reliable statistic to make decisions on. They lean toward actual performance and like pitchers who routinely exceed their expectations. Is it possible that the statistic is not good enough?

I agree that Gallardo is likely not too much better than Wilson/Wright/Worley/Jones but he comes with more consistency. I still think the O's could trade Miguel and Walker and/or one the four above for better prospects or draft picks and get back in the game. Even when Miguel is good, he is inconsistent and currently holds the least WAR value of the starting five. That would be a much better place to start with finding a spot for Wilson/Wright/Bundy than having Miguel higher in the rotation and an empty fifth spot. Who knows what injuries there will be but Miguel/Walker/Wright is a good place to start for a deadline deal too.

Roger said...

One last question. Can you prove that teams that hew more strictly to the statistical data actually perform better? That seems like quite an interesting analysis to make - lots of regression on individual situations and than a regression at the team level.

Don't get me wrong. Advanced Statistics is incredibly valuable but the tenor of discussion here is that "I can't believe the O's would sacrifice #14 for Gallardo" and then you ultimately show the deal is "fair".

Roger said...

Oh, sorry, one more thing. With the "successes" you named (Markakis/Wieters/Machado), I realize you were only talking about drafting talent but teams (see, Braves, Atlanta) often restock their farm system via trade and the Orioles farm system actually produced Jones, Tillman, Schoop, Britton, Givens, Joseph, who are all major leaguers with Jones, Tillman, Schoop, and Britton as starters and Joseph would be if Wieters had not stayed.

Jon Shepherd said...

Basically everyone has an analytics department and Duquette has been one of the leading people in advanced metrics.

Jon Shepherd said...

Teams typically do both.

Jon Shepherd said...

No one thinks xFIP is the end all be all. Otherwise we would not be looking at multiple quantitative and qualitative metrics.

Anonymous said...

Aren't you overlooking depth here when you're comparing Gallardo's projected WAR and the Norfolk shuttle's? Perhaps the Norfolk shuttle is good for a win as the fifth spot in the rotation but won't there be diminishing returns when they're asked to cover two spots in the rotation, which is an inevitability? They're going to need more than five starters.

Jon Shepherd said...

Projection models think Miguel Gonzalez pitches similarly to the Norfolk shuttle. Maybe he goes back to defying expectations or maybe those shoulder and elbow injuries continue to derail him.