Diamond Dollars: Finding Baseball’s MVPitcher The University of Chicago Booth School of Business Sean McCluskey, Andrew Ungerer, Mike Velcich, and Joe Puccio
Nov 30, 2014
Diamond Dollars:Finding Baseball’s MVPitcher
The University of Chicago Booth School of Business
Sean McCluskey, Andrew Ungerer, Mike Velcich, and Joe Puccio
2
Executive Summary
4) Identification – 3 Most Valuable Pitchers
5) Case Study – In-Depth on Each Pitcher
3) Analysis – Statistical Modeling
2) Preliminary Process – Pool, Comps, Contracts
1) Situation – Understanding “Value”
6) Risks –Potential Pitfalls
Value in the Eyes of the Beholder
3
Yankees Cubs
Pirates Marlins
Financial Resource
s
Contending Rebuilding
Low Discount Rate, Low
Payroll
High Discount
Rate, High
Payroll
Veteran vs. Prospect
4
Veteran: Current Production Surplus
Prospect: Delayed Gratification
Budget and proximity to contention is key in identifying the most valuable pitching assets for each specific team.
Our analysis is focused on identifying the best pitching values, agnostic but mindful of team situations.
Performance projections, age, years of control, injury risks, makeup and pitching profile are the key components to a proper
analysis.
The Preliminary Process
5
Phase 1 Phase 2 Phase 3
Objective: Identify Baseball’s Most Valuable Pitchers
Identify target population
• Build database of 75 target pitchers based on performance
• Used last 2 years of performance data, also included top minor leaguers
Determine comparable pitchers
• Identify 10 comparable pitchers for each candidate
• Comps based on Bill James’ Similarity Scores
Analyze contract data and project
arbitration
• Aggregate salaries from Cot’s Baseball contracts
• Create dynamic arbitration salary projections
Population Comparables Contracts
N = 75 N = 16 N = 3
Population Comparables Contracts
6
• Started with PECOTA’s top 40 pitchers for projected 2014 Wins Above Replacement (WAR)
• Cross-referenced against top 80 pitchers in 2012 & 2013o Added top prospects (Taijuan Walker, Gerrit Cole, etc.)o Kept relievers (Craig Kimbrel, Aroldis Chapman, etc.) in order to test
values
1 2 3 4 5 6 7 8 9$0.00
$5.00
$10.00
$15.00
$20.00
$25.00
$30.00
2014 Salary v 2013 WAR
Population Comparables Contracts
• Select comparables for each pitcher in order to project future performance, based on:o Age at time of comparisono Performanceo Volume of work that yearo Handedness
• Comparables were based on Bill James’ Similarity Scores
7
Population Comparables Contracts
• Pulled contract data from Cot’s Baseball Contracts
• Projected arbitration salaries using:o Previous year’s performance / projectiono Recent compso 5% WAR Inflation
8
Analyze Contract Terms
Arbitration Projection Example
Important Note: Our arbitration salaries are calculated dynamically. When a player performs better in the simulation, his pay increases!
9
• Generated projected performance for each pitcher based on the performance of the 10 most comparable pitchers. Projected statistics informed FIP, which was then park-adjusted to calculate WAR
5.0% 90.0% 5.0%4.3% 87.6% 8.0%
-0.523 0.101
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Fit Comparison for Fernandez IP Y4RiskExtValueMin(-0.042391,0.15455)
Input
Minimum-0.52349Maximum0.10077Mean -0.13924Std Dev 0.22444Values 9
ExtValueMin
Minimum −∞Maximum +∞Mean -0.13160Std Dev 0.19822
@RISK Student VersionFor Academic Use Only
5.0% 90.0% 5.0%
107.32 169.42
40
60
80
10
01
20
14
01
60
18
02
00
Values in Millions ($)
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
Fernandez Total Surplus
Fernandez Total Surplus
Minimum $49,570,278.60Maximum$194,997,229.57Mean $139,734,502.10Std Dev $18,804,478.32Values 9999 / 10000Errors 1
@RISK Student VersionFor Academic Use Only
o IPo IP/GSo BB/9
o HBP/9o K/9o HR/9
• Ran 10,000 iterations per pitcher
Statistical Analysis
10
• Example: WAR and Value Surplus simulation results
1 2 3 4 50
50
100
150
Jose Fernandez100 = WAR of 5.6 and
2014 Surplus of $32.8m
Mean WAR Mean Surplus (PV)
Control Years
1 2 3 4 50
50
100
150
200
Jose Fernandez100 = 2014 Surplus of
$32.8m
Mean Surplus (PV) 0.95 Percentile0.05 Percentile
Control Years
Statistical Analysis
11
• WAR dollars using agreed-upon ($6 M/win with 5% inflation)
• Less that season’s compensation (known or estimated)
• Equals excess value• Used risk-adjusted discounted rates to calculate
present value of excess…
• BUT – What about Return on Investment?o If two players generated the same excess, wouldn’t we place higher value
on the lower cost player? YES!o But where’s the trade-off between Gross Surplus and ROI?
• What are assumptions made about return on other salary commitments?
• Portfolio theory is an incomplete framework- only 5 rotation slots available!
• Need to look at multiple measures of value and layer in team-specific situation
Statistical Analysis
12
Statistical Analysis
Moo
re
Stras
burg
Bumga
rner
Wai
nwrig
htRyu
Mill
er
Darvi
sh
Teher
an
Kersh
aw Sale
Ferna
ndez
0
40
80
120
160
Surplus Value ($M)
Moo
re
Stras
burg
Bumga
rner
Wai
nwrig
htRyu
Mill
er
Darvi
sh
Teher
an
Kersh
aw Sale
Ferna
ndez
0.0
2.0
4.0
6.0
Ratio of Surplus Value to Expected Salary
13
• Sourced mechanics ranks from Doug Thorburn (Baseball Prospectus) and DL assessments from Jeff Zimmerman (FanGraphs) to support qualitative research
• Mechanics ratings helped inform player discount rates used in PV calculationo Hurt: Miller and Saleo Helped: Bumgarner, Fernandez and Darvish
Scouting & Risk Analysis
Pitching Mechanics & DL Risk AssessmentPitcher Balance Momentum Torque Posture Release Distance Consistency Overall 2014 DL Risk 1
Bumgarner 70 45 65 80 65 70 A 26.2%Fernandez 65 65 70 65 65 60 A 28.6%Darvish 60 60 60 80 65 60 A- 37.3%Teheran 55 65 60 65 65 60 B+ 30.2%Moore 60 55 70 60 60 35 B+ 35.1%Strasburg 65 55 70 50 55 55 B 39.9%Kershaw 55 55 60 50 50 70 B 28.0%Ryu 55 60 55 55 60 60 B 35.1%Wainwright 65 45 50 55 55 70 B 42.9%Sale 30 50 60 70 60 65 B 30.3%Miller 45 55 65 55 60 55 B- 30.4%
1 League Avg DL Risk = 38.1%
14
• Kershaw and Wainwright paced the field in terms of both endurance and durability
• Though projected to be a solid return on investment for the Cardinals, Miller has yet to establish himself pitching late into games. Additionally, his low ground ball rate and high FB reliance raise questions
Pitcher ProfilesPitcher Age Height Weight Throws FIP IP IP/GS GB%Bumgarner 24 6'5" 235 Left 3.05 201.1 6.5 46.8%Fernandez 21 6'2" 240 Right 2.86 172.2 6.2 45.1%Darvish 27 6'5" 225 Right 3.17 209.2 6.6 41.0%Teheran 23 6'2" 175 Right 3.69 185.2 6.2 37.8%Moore 25 6'3" 210 Left 3.95 150.1 5.6 39.4%Strasburg 25 6'4" 200 Right 3.21 183.0 6.1 51.5%Kershaw 25 6'3" 220 Left 2.39 236.0 7.2 46.0%Ryu 27 6'2" 255 Left 3.24 192.0 6.4 50.6%Wainwright 32 6'7" 235 Right 2.55 241.2 7.1 49.1%Sale 25 6'6" 180 Left 3.17 214.1 7.1 46.6%Miller 23 6'3" 215 Right 3.67 173.1 5.6 38.4%
2013 Headline StatsProfile
15
• Fernandez’s ability to attack the zone early on in his career bodes well for confidence and projections of stuff
• Darvish shows elite strikeout ability but struggled with HRs in ’13
• Ryu relies on command, not velocity to achieve results
Pitcher ProfilesPitcher Avg FB Velo K% BB% HR% SwStr% Fastball % Breaking Ball % Off Speed % Zone %Bumgarner 91.2 24.8% 7.7% 1.9% 11.1% 39.6% 50.1% 9.7% 51.4%Fernandez 94.9 27.5% 8.5% 1.5% 10.1% 57.3% 34.0% 8.7% 55.0%Darvish 92.9 32.9% 9.5% 3.1% 12.6% 38.2% 59.9% 1.9% 46.8%Teheran 91.5 22.0% 5.8% 2.8% 10.5% 63.8% 31.5% 5.3% 53.2%Moore 92.4 22.3% 11.8% 2.2% 9.5% 62.2% 19.6% 18.4% 44.3%Strasburg 95.3 26.1% 7.7% 2.2% 10.6% 61.0% 22.9% 16.1% 49.4%Kershaw 92.6 25.6% 5.7% 1.2% 11.4% 60.7% 36.9% 2.4% 50.2%Ryu 90.3 19.7% 6.3% 1.9% 8.1% 54.2% 23.9% 22.3% 51.1%Wainwright 91.1 22.9% 3.7% 1.6% 9.6% 40.5% 58.0% 3.8% 48.9%Sale 93.1 26.1% 5.3% 2.7% 10.8% 51.4% 29.8% 19.0% 52.4%Miller 93.7 23.4% 7.9% 2.8% 9.0% 71.3% 18.8% 6.3% 53.1%
Stuff Approach
The Three Most Valuable Pitchers in
Baseball
16
#3
17
Pitcher #3
Pitcher #2
Pitcher #1
18
Yu Darvish – SP Texas RangersBats: R Height: 6’ 5”Throws: R Weight: 225lbAge: 27 Service Time: 2 yrAcquired: post ‘12 Awards: 2x ASG Contract Terms
2014 2015 2016 2017 2018 2019
Yu Darvish $10.00 $10.00 $10.00 $11.00 FA FA
YearW L G IP ERA WHIP CG SHO H R ER HR BB K K/BB K/9 GB/FB ERA+ FIP WAR
2012 16 9 29 191.1 3.9 1.28 0 0 156 89 83 14 89 221 2.48 10.4 1.46 112 3.29 3.9
2013 13 9 32 209.2 2.83 1.073 0 0 145 68 66 26 80 277 3.46 12 1.08 145 3.28 5.8
2 Yrs 29 18 61 401 3.34 1.17 0 0 301 157 149 40 169 498 2.95 11.2 1.25 127 3.28 9.6
Elite Stuff: Three of the top Seven Whiff Rate pitches in our sample belong to Yu and his 32 K% paces the field
Pitcher #3
Pitcher #2
Pitcher #1
19
Yu Darvish – SP Texas Rangers
- Very consistent velocity despite increased usage in ML year 2
- Showcased elite production in ’13 despite an elevated HR rate
- Reduced fastball effectiveness and usage in ’13 is trend to monitor though no visible signs of velo or movement loss
- Sheer variety of offerings is a unique attribute
Pitcher #3
Pitcher #2
Pitcher #1
21
Chris Sale – SP Chicago White SoxBats: L Height: 6’ 6”Throws: L Weight: 180lbAge: 25 Service Time: 3.06yrDrafted: 13th – 2010 Awards: 2x ASG2014 2015 2016 2017 2018 2019
Chris Sale $3.50 $6.00 $9.15 $12.00 $12.50* $13.50*
* Denotes team options
**Denotes time spent as a reliever
Contract Terms
Unique Movement: In 18 pitcher, 87 pitch sample, Sale has three of the top ten pitches with the most horizontal movement
YearW L G IP ERA WHIP CG SHO H R ER HR BB K K/BB K/9 GB/FB ERA+ FIP WAR
2010 2 1 21** 23.1 1.93 1.071 0 0 15 5 5 2 10 32 3.2 12.47 1.39 225 2.74 1.2
2011 2 2 58** 71 2.79 1.113 0 0 52 22 22 6 27 79 2.926 10.01 1.55 156 3.12 2.3
2012 17 8 30 192 3.05 1.135 1 0 167 66 65 19 51 192 3.765 9 1.40 140 3.27 5.9
2013 11 14 30 214.1 3.07 1.073 4 1 184 81 73 23 46 226 4.913 9.5 1.46 140 3.17 6.9
4 Yrs 32 25 139 500 2.97 1.1 5 1 418 174 165 50 134 529 3.95 9.52 1.44 144 3.12 16.3
Pitcher #3
Pitcher #2
Pitcher #1
22
Chris Sale – SP Chicago White Sox
Sale, at age 20, after being drafted, asked about what other pitchers he models himself after:"You can't really pitch like anyone," Sale said. "Everyone has own style of pitching, as they do hitting. I don't try to pitch like [Cole] Hamels or [Randy] Johnson, throwing 100 mph or the nastiest breaking ball ever. I pitch my game."
- Sale actually slightly increased velo from year 1 to year 2 as a starter and saw gains in already lethal slider’s effectiveness as well.
- Concern remains about “Inverted ‘W’” delivery:
Makeup and Personality
#1
23
Pitcher #3
Pitcher #2
Pitcher #1
24
Jose Fernandez – SP Miami MarlinsBats: R Height: 6’ 2”Throws: R Weight: 240lbAge: 21 Service Time: 1yrDrafted: 14th – 2011 Awards: ‘13 RoY, ASContract Terms
2014 2015 2016 2017 2018 2019Jose Fernandez $0.64 $0.80 Arb 1 Arb 2 Arb 3 FA
Attacks Hitters: 55% of pitches in strike zone, represents highest % in our top-tier sample
Pitch Breakdown
Year W L G IP ERA WHIP CG SHO H R ER HR BB K K/BB K/9 GB/FB ERA+ FIP WAR
2013 12 6 28 172.2 2.19 0.979 0 0 111 47 42 10 58 187 3.224 9.774 1.36 176 2.73 6.3
Pitcher #3
Pitcher #2
Pitcher #1
25
Qualitative Analysis:• Prototypical workhorse build
(6’2”, 240)• Outstanding makeup – Escaped
Cuba as a teen, rescued his mother after she fell overboard
• Fernandez, after admiring his first career home run from the batter’s box in a game against the Atlanta Braves: “This is a professional game, and we should be professional players. I think that never should happen. I'm embarrassed, and hopefully that will never happen again."
• Fearless mound presence
Jose Fernandez – SP Miami Marlins
Scouting Analysis:• Pounds the zone• Highly ffective against both
Righties and Lefties• Development of secondary
pitches could sustain current level
Risk Factors & Pitfalls
26
Potential flaws in our approach:• Scope of comp set
• Model ran with 10 comps per player; it was challenging to identify good comps for such young, inexperienced pitchers, and this introduced variation
• Risk assessment• Captured projection confidence and injury probability in two places:
comps, and discount rate. For example, used higher discount rate on Chris Sale because of his inverted “W” mechanics. Also used higher rate on Jose Fernandez because of his brief, 1 year track record
• Definition of value• We defined value in multiple ways, first by gross excess return over
contractual obligations, then by return on investment. These approaches yielded somewhat different results, and so the answer can change based on the desired outcome
Thank you – Questions?
27
Joe Puccio
Mike Velcich
Sean McCluskey
Andrew Ungerer