Search This Blog

Sunday, March 10, 2013

WAR and Fantasy Baseball: Any Connection?

So along with being one who blogs frequently and researches both important and obscure things within baseball, I fancy myself a bit of a fantasy baseball nut.  I'll play Roto, H2H, points, with divisions, without divisions, with BA, with OPS, with wins (reluctantly), and with just about anyone who wants to play.  Each year, I spend my fair share of time doing research on who I think should be taken and when they should be taken.  Up until last year, this research involved looking into home runs, averages, and roster construction.  However, as some of my readers may know, last year changed the way I think about baseball quite a bit.  In an extensive research project, I looked at how contact quality could be used to predict on-base percentage. I used the results of this study to make projections for 105 MLB regulars heading into the 2012 season.  This, combined with my knowledge of how to project power numbers, led me to predict OPS as well.  To stick true to my research, I went and drafted my 2012 fantasy baseball offense based on these projections.  Despite many injuries to both my offense and pitching staff that led me to moving players around and essentially blowing up my team, I finished exactly one thousandth of a point off the OPS crown (I finished with a team OPS of .814 for those that are curious). However, because of said injuries and unexpected disappointments in certain lineups (real gem of a job, Reds leadoff hitters) I finished with a lackluster overall offensive performance.

Why do I bring this up?  Well, my research into fantasy baseball never stops.  This year I'm doing the same league, and I'm going with the same approach.  Due to my reputation as being a bit of a stats nut, one of my co-players teased at the notion that I should rank players by WAR and do an auto draft.  While I would never do this for a slew of different reasons, it did get me thinking a bit.  First off, what is the purpose of WAR?  To approximate value based on overall performance.  What is the purpose of fantasy baseball rankings? To sort players based on an approximation of their overall projected performance.  Obviously the key issue here is that fantasy baseball is all about offense and WAR encompasses all parts of baseball.  This problem is quite easily solved by breaking apart WAR and simply using its offensive component.  So the question that comes up is: if you drafted a fantasy offense solely on oWAR, how would that compare to drafting based on Yahoo's rankings? I am here to figure that out.

Step One:  Setup

The first part of undertaking any research project is to understand what you are trying to find.  The thought here is that oWAR and fantasy baseball rankings should be looking to accomplish the same thing: telling the story of overall value relative to your raw performance, your position, and your park (and for fantasy baseball: the team around you).  To continue this understanding, here are a few key things to know:

1) oWAR hurts players for playing in a favorable environment whereas fantasy baseball rankings promote that.  It can't be denied that Carlos Gonzalez gets a boost in fantasy rankings because he plays half of his games in Coors Field.

2) Traditional fantasy baseball does not split the outfield into three positions.  Because of this, oWAR numbers for certain outfielders will be higher/lower than they *should* be, because the positional adjustments favor center fielders over left and right fielders.

3) oWAR cares not about the lineup around a player.  Part of the reason for creating oWAR and similar metrics was to get an idea of how a player would contribute given the "average" or "typical" environment.  Fantasy baseball cares a lot about the lineup around a player.  Giancarlo Stanton is hurt by the fact that Jose Reyes and Hanley Ramirez are no longer a part of the Marlins' organization.

4) Both oWAR and fantasy baseball are concerned with playing time.  They both rely on a player's ability to accumulate numbers over time.

With these things in mind, I created a table with several different categories.  I took into account games played in 2011 and 2012, oWAR in 2011 and 2012, oWAR/162 games in 2011 and 2012, and a projection of performance in 2013.  I wanted to incorporate overall performance, performance rates, and playing time so as to be able to account for injuries.  With this all set up, I was able to get into the data.

Step Two: Analysis

The following chart ranks Yahoo's top 100 players (the base column on the far left) and then lists how those rankings would look if they were based on other things:
 
Rank Yahoo 2013 By 2012 oWAR By Expected 2013 oWAR By 2012 oWAR/162
1 Mike Trout Mike Trout Mike Trout Mike Trout
2 Ryan Braun Andrew McCutchen Miguel Cabrera Buster Posey
3 Miguel Cabrera Miguel Cabrera Andrew McCutchen Andrew McCutchen
4 Robinson Cano Buster Posey Jose Bautista Miguel Cabrera
5 Matt Kemp Robinson Cano Ryan Braun Robinson Cano
6 Albert Pujols Chase Headley Matt Kemp Melky Cabrera
7 Joey Votto Ryan Braun Buster Posey Joey Votto
8 Andrew McCutchen Adrian Beltre Robinson Cano Ryan Braun
9 Carlos Gonzalez Adam Jones Joey Votto Chase Headley
10 Prince Fielder Ben Zobrist Yoenis Cespedes Yadier Molina
11 Giancarlo Stanton Edwin Encarnacion Chase Headley Adrian Beltre
12 Troy Tulowitzki Prince Fielder Bryce Harper Austin Jackson
13 Jose Bautista David Wright Adrian Beltre David Ortiz
14 Buster Posey Joe Mauer Ben Zobrist Giancarlo Stanton
15 Josh Hamilton Yadier Molina Jose Reyes Matt Kemp
16 Adrian Beltre Aramis Ramirez Evan Longoria Edwin Encarnacion
17 Evan Longoria Austin Jackson Prince Fielder Joe Mauer
18 David Wright Aaron Hill Yadier Molina Ben Zobrist
19 Justin Upton Melky Cabrera Austin Jackson Jose Bautista
20 Dustin Pedroia Joey Votto David Ortiz Yoenis Cespedes
21 Hanley Ramirez Jose Reyes Joe Mauer Aramis Ramirez
22 Ian Kinsler Shin-Soo Choo Aramis Ramirez Adam Jones
23 Edwin Encarnacion Josh Hamilton Melky Cabrera David Wright
24 Jose Reyes Matt Holliday Giancarlo Stanton Evan Longoria
25 Jason Heyward Josh Willingham Adam Jones Prince Fielder
26 Bryce Harper Giancarlo Stanton Troy Tulowitzki Josh Willingham
27 Adam Jones Yoenis Cespedes Edwin Encarnacion Aaron Hill
28 Curtis Granderson Derek Jeter David Wright Josh Hamilton
29 Jacoby Ellsbury Curtis Granderson Josh Hamilton Ian Desmond
30 Jay Bruce Albert Pujols Dustin Pedroia Shin-Soo Choo
31 Adrian Gonzalez Ryan Zimmerman Matt Holliday Jose Reyes
32 Matt Holliday Ian Desmond Justin Upton Matt Holliday
33 Starlin Castro Alex Gordon Josh Willingham Dustin Pedroia
34 Billy Butler Dustin Pedroia Curtis Granderson Miguel Montero
35 Ryan Zimmerman Asdrubal Cabrera Asdrubal Cabrera Ryan Zimmerman
36 Yoenis Cespedes Miguel Montero Albert Pujols Asdrubal Cabrera
37 Paul Goldschmidt Matt Kemp Carlos Santana Derek Jeter
38 B.J. Upton Carlos Santana Pablo Sandoval Carlos Santana
39 Allen Craig Alex Rios Shin-Soo Choo Salvador Perez
40 Joe Mauer Jason Kipnis Alex Gordon Albert Pujols
41 Chase Headley Bryce Harper Jason Kipnis Bryce Harper
42 Jason Kipnis Billy Butler Chase Utley Curtis Granderson
43 Brett Lawrie Jason Heyward Mike Napoli Chase Utley
44 Brandon Phillips Michael Bourn Derek Jeter Alex Gordon
45 Ian Desmond Martin Prado Ryan Zimmerman Allen Craig
46 Ben Zobrist David Freese Miguel Montero Nick Markakis
47 Mark Teixeira B.J. Upton Allen Craig Jason Kipnis
48 Pablo Sandoval David Ortiz Ian Desmond Alex Rios
49 Elvis Andrus Jimmy Rollins Jason Heyward Pablo Sandoval
50 Aaron Hill Jose Altuve Salvador Perez David Freese
51 Yadier Molina Nick Swisher Aaron Hill B.J. Upton
52 David Ortiz Jose Bautista Carlos Beltran Billy Butler
53 Jimmy Rollins Starlin Castro Alex Rios Jose Altuve
54 Freddie Freeman Carlos Beltran B.J. Upton Nick Swisher
55 Carlos Santana Adam LaRoche Starlin Castro Michael Bourn
56 Alex Rios Pedro Alvarez Jacoby Ellsbury Alejandro De Aza
57 Desmond Jennings Hanley Ramirez David Freese Martin Prado
58 Aramis Ramirez Paul Goldschmidt Desmond Jennings Jason Heyward
59 Anthony Rizzo Allen Craig Adrian Gonzalez Pedro Alvarez
60 Michael Bourn Elvis Andrus Michael Bourn Paul Goldschmidt
61 Victor Martinez Matt Wieters Jimmy Rollins Will Middlebrooks
62 Matt Wieters Paul Konerko Matt Wieters Jimmy Rollins
63 Shin-Soo Choo Alejandro De Aza Carlos Gonzalez Carlos Beltran
64 Ryan Howard Mark Trumbo Paul Konerko Anthony Rizzo
65 Martin Prado Carlos Gonzalez Nick Markakis Adam LaRoche
66 Carlos Beltran Rickie Weeks Brett Lawrie Matt Wieters
67 Austin Jackson Pablo Sandoval Jose Altuve Troy Tulowitzki
68 Mark Trumbo Nick Markakis Nick Swisher Paul Konerko
69 Asdrubal Cabrera Evan Longoria Elvis Andrus Starlin Castro
70 Alex Gordon Brandon Phillips Victor Martinez Carlos Gonzalez
71 Wilin Rosario Desmond Jennings Alejandro De Aza Hanley Ramirez
72 Nelson Cruz Brett Lawrie Billy Butler Mark Trumbo
73 Jose Altuve Ian Kinsler Martin Prado Elvis Andrus
74 Paul Konerko Chase Utley Pedro Alvarez Mike Napoli
75 Shane Victorino Dan Uggla Paul Goldschmidt Brett Lawrie
76 David Freese Mark Teixeira Rickie Weeks Desmond Jennings
77 Melky Cabrera Wilin Rosario Brandon Phillips Wilin Rosario
78 Miguel Montero Carlos Gomez Wilin Rosario Rickie Weeks
79 Carl Crawford Mike Napoli Anthony Rizzo Mark Teixeira
80 Derek Jeter Salvador Perez Will Middlebrooks Brandon Phillips
81 Eric Hosmer Justin Upton Hanley Ramirez Carlos Gomez
82 Josh Willingham Adrian Gonzalez Mark Trumbo Ben Revere
83 Hunter Pence Shane Victorino Ian Kinsler Dan Uggla
84 Ike Davis Jay Bruce Mark Teixeira Carl Crawford
85 Rickie Weeks Anthony Rizzo Shane Victorino Ian Kinsler
86 Carlos Gomez Ben Revere Dan Uggla Justin Upton
87 Chase Utley Freddie Freeman Ben Revere Shane Victorino
88 Brian McCann Nelson Cruz Hunter Pence Adrian Gonzalez
89 Dan Uggla Hunter Pence Jay Bruce Jay Bruce
90 Mike Napoli Mike Moustakas Freddie Freeman Freddie Freeman
91 Salvador Perez Danny Espinosa Adam LaRoche Mike Moustakas
92 Adam LaRoche Will Middlebrooks Mike Moustakas Nelson Cruz
93 Nick Markakis Troy Tulowitzki Carl Crawford Hunter Pence
94 Mike Moustakas Ike Davis Carlos Gomez Danny Espinosa
95 Danny Espinosa Jacoby Ellsbury Nelson Cruz Jacoby Ellsbury
96 Nick Swisher Brian McCann Danny Espinosa Ike Davis
97 Pedro Alvarez Carl Crawford Brian McCann Brian McCann
98 Ben Revere Eric Hosmer Ryan Howard Eric Hosmer
99 Will Middlebrooks Victor Martinez Ike Davis Victor Martinez
100 Alejandro De Aza Ryan Howard Eric Hosmer Ryan Howard

Now, I must admit that these projections aren't based on any system I would typically use.  I simply looked at the 2011 and 2012 rates and gave a bit more weight to the more recent performance.  As you can probably tell, the rankings came out quite different.  I'll let you look into the rankings to find quirks if you want to, but after breaking down the rankings, I thought, "Well, the point of this was to find out what the teams would look like if you put it on auto draft.  So that's what I did!  Let's say you have a 12 team league that only uses offense (fantasy baseball players, this will seem weird to you) and a team has the #1 overall pick both times.  With 100 players, Team 1 (boring team name, I know) picks at 1, 24, 25, 48, 49, 72, 73, 96, and 97.  Here are what the teams look like if the drafts went perfectly in order (in cases where a position is filled, I will move down to the best player at an unfilled position):

Yahoo Rankings

1- Mike Trout, OF
24- Jose Reyes, SS
25- Jason Heyward, OF
48- Pablo Sandoval, 3B
49- Aaron Hill, 2B (Elvis Andrus doesn't fit)
72- Nelson Cruz, OF
73- Paul Konerko, 1B (Jose Altuve doesn't fit)
96- Nick Swisher, UTIL
97- Pedro Alvarez, UTIL (all players 96-100 are 3B/OF/1B)

oWAR Rankings (using expected in 2013)

1- Mike Trout, OF
24- Giancarlo Stanton, OF
25- Adam Jones, OF
48- Ian Desmond, SS
49- Aaron Hill (Jason Heyward doesn't fit)
72- Billy Butler, 1B
73- Martin Prado, 3B
96- Danny Espinosa, 2B
97-Brian McCann, C

One thing that comes to mind in looking at how these teams panned out is that both systems seem to agree that taking outfielders is important.  After all, other than third base, the OF position had a highest average expected oWAR/162 games in 2013 at 4.09.  Where the oWAR rankings fail is in accounting for the actual number of players available at each position. This is why the Yahoo team had no available key position players left in the later rounds and had to settle for OF/1B types.  Yahoo understands that additional value is given to players based on how rare quality players are to find at those positions (something WAR does not account for, as it only cares about how hard the position is to play overall).  To show what I'm saying, here's a breakdown of top 100 position players by position:

Number of Players Position
11 C
15 1B
12 2B
8 SS
14 3B
38 OF

As you can see, there is a plethora of outfield talent available in the top 100. Even though I listed players by primary positions instead of by Yahoo! available positions, there are more outfielders than actual available outfield spots (36 total OF spots in a 12 team league).  Now, this listing technicality throws the other numbers off just a little bit as well (Ben Zobrist and Martin Prado would be available to play SS, etc) but I think the idea is clear: other positions are too thin for OF to be taken with the first three picks.  Therefore, Yahoo! bumps up positions like SS/2B/C in their overall rankings.  They understand your chances of finding value late at OF are better than doing so at these other positions.

Now, I don't want to imply that Yahoo! is right at all, because it's quite possible that their philosophy is wrong.  Before I jump to conclusions, let's check some things out.  First, it's important to know what kind of expected value each position brings.  For that, here's a chart showing average expected oWAR by position:

Row Labels Average of Exp oWAR 2013/162
1B 3.10
2B 3.68
3B 4.24
C 3.94
DH 4.00
OF 4.09
SS 4.05
Grand Total 3.89


When you look at this, the numbers suggest that certain positions have a better "average" top 100 player.  First baseman are hurt in Yahoo! rankings by the same thing they are hurt by in oWAR: the typical first baseman provides more offense relative to other positions.  This chart suggests that, relative to position, the other positions are pretty balanced.  Here's how the standard deviation breaks down:

Row Labels StdDev of Exp oWAR 2013/162
1B 1.53
2B 1.36
3B 1.52
C 1.31
OF 1.62
SS 0.82
Grand Total 1.48

This chart reveals a few things when paired with the first chart.  Number one: shortstops are both valuable and pretty similar in the top 100.  However, when you add Zobrist and Prado, there are only ten of them.  This suggests taking a shortstop should be a priority.  Number two: there's quite a bit of difference between outfielders, first basemen, and third basemen. This suggests that the top-line talent is a lot better than the lesser talent.  Both Yahoo! and oWAR agree that this is the case, with oWAR having eight outfielders in the top 20 players and Yahoo! having nine.  Knowing this, perhaps the oWAR team was right in drafting three OF early...or not.  While it is nice to know that OF is top-heavy in terms of ability, it also has to be made known that "lesser tier" outfielders are still being ranked in the top 100 while lesser tier players at other positions do not appear.  The moral of the story after looking at average talent and distribution?  It's all about the players available. There are fewer 2B/SS/C available that will contribute value than there are OF.

Step Three: Conclusions

In short, ranking players based on oWAR is probably a really bad idea.  Position-wise, both ranking systems have 19 players who are primarily outfielders in the top 50, both have a similar number of shortstops, third basemen, etc.  The problem isn't the distribution of players within the rankings.  Both formats agree on that concept.  Where things get dicey is where those players get ranked.  Below is a simple list of some classic examples where you would be torched by using oWAR to autodraft (first number is Yahoo! ranking, second number is oWAR ranking):

Robinson Cano: 4, 8
Albert Pujols: 6, 36
Carlos Gonzalez: 9, 63
Giancarlo Stanton: 11, 24
Troy Tulowitzki: 12, 26
Jose Bautista: 13, 4 (overdraft city)
Buster Posey: 14, 7
David Wright: 18, 28

I could keep going on and on with differences, but in the end there just isn't enough in common between what oWAR wants to accomplish and what Yahoo! wants to accomplish.  While I don't recommend auto drafting either way, I would say it's probably better overall to let Yahoo! do the work.  Things like lineup context and ballpark are important in fantasy baseball, because we're talking about producing runs and RBI. oWAR is really good at what it does, but it should stay out of fantasy baseball.

No comments:

Post a Comment