Sports analytics is really sexy. Michael Lewis' 2004 best-seller Moneyball¹ put it on the map. A subsequent movie staring Brad Pitt didn't hurt things much, either. The MIT-Sloan School of Business even holds an annual conference on the topic.
We might expect — given all they hype — that analytics' reign in sports is unchallenged. The Moneyball movie after all places words to that effect in the mouth of Boston Red Sox' owner John Henry. The Henry character suggests that any organization not reinventing itself after the Billy Beane model would become a dinosaur.
Analytics does not in fact rule sports. The web page for sports network ESPN recently published a survey ranking major sports teams by their extent of analytics adoption.² Adoption was very non-uniform within major sports and between sports.
What explains this disparity? At least two factors come into play. Much has been written about the role of culture in analytics adoption. Some businesses just lend themselves more to analytics. I briefly explore both factors here.
Lavalle³ and Davenport⁴ consider extensively the role in culture to the adoption of analytics. Lavelle writes, "The adoption barriers that organizations face are managerial and cultural, rather than related to data and technology." Davenport's Analytics at Work devotes an entire chapter to "Building an analytics culture."
Baseball has a long, rich history of statistics. A college dormitory roommate comes to mind as an example. This young man often had three things going on at once as I entered the room. The television was on, with the volume turned all the way down. The radio was playing music. And he had a six-inch-thick encyclopedia of baseball statistics open on his lap. I saw no evidence of athletic participation in him. But he devoured baseball statistics.
A casual search of Amazon.com brings up volumes on baseball statistics. The legendary Bill James baseball abstract⁵ contains more than 1,000 pages of baseball history, rich with statistics. An annual update⁶ contains nearly 600 pages describing analytical methodologies. Baseball America publishes an annual edition to a baseball almanac⁷ presenting "a comprehensive statistical review from the majors all the way through to youth baseball." For the quant each of us knows and loves, there is even a book⁸ on using the R programming language to analyze baseball statistics.
Theories underlying basketball statistics have arrived more recently. Amazon yields fewer hits overall. A math professor and a software engineer collaborated on a recent book⁹ describing statistical methods for basketball. That this work was not picked up by a major commercial publisher might suggest something about the market for basketball statistics.
We might expect — given all they hype — that analytics' reign in sports is unchallenged. The Moneyball movie after all places words to that effect in the mouth of Boston Red Sox' owner John Henry. The Henry character suggests that any organization not reinventing itself after the Billy Beane model would become a dinosaur.
Analytics does not in fact rule sports. The web page for sports network ESPN recently published a survey ranking major sports teams by their extent of analytics adoption.² Adoption was very non-uniform within major sports and between sports.
 |
| Figure 1 — Sports website espn.com recently published a survey² on the adoption of analytics in sports. Not only did the extent of adoption between sports vary, but also within sports. |
What explains this disparity? At least two factors come into play. Much has been written about the role of culture in analytics adoption. Some businesses just lend themselves more to analytics. I briefly explore both factors here.
Cultures of data.
ESPN's survey shows that analytics plays a bigger role in baseball than other major U.S. sports. ESPN characterized each team in each of the four major U.S. professional sports leagues — Baseball, Basketball, Football, Hockey — in terms of five tiers of adoption:- All-in;
- Believers;
- One foot in;
- Skeptics; and
- Non-believers.
Lavalle³ and Davenport⁴ consider extensively the role in culture to the adoption of analytics. Lavelle writes, "The adoption barriers that organizations face are managerial and cultural, rather than related to data and technology." Davenport's Analytics at Work devotes an entire chapter to "Building an analytics culture."
Baseball has a long, rich history of statistics. A college dormitory roommate comes to mind as an example. This young man often had three things going on at once as I entered the room. The television was on, with the volume turned all the way down. The radio was playing music. And he had a six-inch-thick encyclopedia of baseball statistics open on his lap. I saw no evidence of athletic participation in him. But he devoured baseball statistics.
A casual search of Amazon.com brings up volumes on baseball statistics. The legendary Bill James baseball abstract⁵ contains more than 1,000 pages of baseball history, rich with statistics. An annual update⁶ contains nearly 600 pages describing analytical methodologies. Baseball America publishes an annual edition to a baseball almanac⁷ presenting "a comprehensive statistical review from the majors all the way through to youth baseball." For the quant each of us knows and loves, there is even a book⁸ on using the R programming language to analyze baseball statistics.
Theories underlying basketball statistics have arrived more recently. Amazon yields fewer hits overall. A math professor and a software engineer collaborated on a recent book⁹ describing statistical methods for basketball. That this work was not picked up by a major commercial publisher might suggest something about the market for basketball statistics.
Susceptibility to mathematical analysis.
Some activities just lend themselves better to mathematical analysis. The business literature recognizes that some business activities should remain more flexible and less structured.¹⁰ Human endeavors can be characterized by their locations along a continuum spanning from "pure art" to "pure science."
Activities that are more scientific easily lend themselves to mathematical specification. Hypothesis testing lies at the heart of the scientific method. I must specify an activity with mathematical precision before I apply analytics to it. Robin Williams' character in the 1989 movie Dead Poets Society gives us an illustrative mockery of the problems with applying scientific methods to artistic endeavors.
Among major U.S. professional sports, baseball happens to lend itself to mathematical analysis. The game is highly structured. The flow of a baseball game is characterized by discrete states. These states are characterized by innings and outs, strikes and pitches, players on base, points scored.
Each play — commencing with the pitcher's release of the ball towards the batter — creates the the opportunity to move from one state to the next. Individual player's abilities — hitting, running, throwing, catching — determine their ability to contribute to transition to a next state that is more favorable to his team than the previous. Player statistics are largely based on how well their play moves the game to more-advantageous state.
Baseball enjoys an underlying structure that is inherently mathematical. A set of mathematical methods exists focused on characterizing transitions from one discrete to another. This set of methods is referred to as Markov Chains.¹¹ Baseball-statistics aficionados may not think about Markov chains. But they are there.
Tennis — not one of the "Big Four" U.S. professional sports — enjoys a similarly mathematical structure. The flows of tennis games progress through a discrete set of states. A statistical researcher recently described a model for tennis matches.¹²
The remaining three of the "Big Four" U.S. professional sports lack the underlying mathematical structure of baseball and tennis. Football is arguably the most structured. Beyond ball possession and points scored, scoring drives follow through discrete states characterized by down, yards to go, and location of line of scrimmage.
Three of these states are continuous. The line of scrimmage and yards to go are represented by continuous numbers. Transitions from one state to the next are also continuous in nature. These factors lend themselves to statistical analysis less easily than discrete states and transitions of baseball.
Basketball and hockey are even less structured. Game flows result from continuous, random interactions between players. ESPN's survey finds that analytics adoption in basketball is more advanced than in hockey. This may be because basketball teams are smaller. Contributions by individual players have a greater bearing on the flow of play.
Baseball, among major U.S. professional sports, enjoys a culture that is substantially analytical. Baseball statistics occupy a prominent place in the sport's fandom. Statistics are sufficiently important to the business of baseball for major commercial publishers to release large volumes.
Explaining the prominence of statistics in baseball in particular may be a "chicken-or-the-egg" question. That the structure of the game is fundamentally mathematical in nature certainly does not introduce any obstacles to a popular baseball-statistics sub-culture.
Organizational leaders seeking to inculcate analytics more deeply into their organizations must manage cultural change, first and foremost. Cultural change is perhaps the most formidable of management challenges. Management guru Peter Drucker's dictum "Culture eats strategy for breakfast" states the challenge well.
Leaders must also recognize the limitations of analytics. In order to apply analytics to a problem, we first must describe it with mathematical precision. Not everything does — or necessarily should — lend itself to this degree of precision.¹⁰
Assertions by exuberant advocates suggest that anything can be measured.¹³ Many such improvisational measurements require selection of proxies — substitutes for quantities that cannot be directly observed. Such proxies may lack precision — or sufficiently direct correlation to the desired quantity — to usefully reduce uncertainty.
Methods from the economics of information give us a rigorous approach to quantify the value of marginal uncertainty. Measurements and reporting based on proxies may not yield information-economic returns worthy of the required investment. Sports analytics gives us practical illustrations.
Activities that are more scientific easily lend themselves to mathematical specification. Hypothesis testing lies at the heart of the scientific method. I must specify an activity with mathematical precision before I apply analytics to it. Robin Williams' character in the 1989 movie Dead Poets Society gives us an illustrative mockery of the problems with applying scientific methods to artistic endeavors.
Among major U.S. professional sports, baseball happens to lend itself to mathematical analysis. The game is highly structured. The flow of a baseball game is characterized by discrete states. These states are characterized by innings and outs, strikes and pitches, players on base, points scored.
Each play — commencing with the pitcher's release of the ball towards the batter — creates the the opportunity to move from one state to the next. Individual player's abilities — hitting, running, throwing, catching — determine their ability to contribute to transition to a next state that is more favorable to his team than the previous. Player statistics are largely based on how well their play moves the game to more-advantageous state.
Baseball enjoys an underlying structure that is inherently mathematical. A set of mathematical methods exists focused on characterizing transitions from one discrete to another. This set of methods is referred to as Markov Chains.¹¹ Baseball-statistics aficionados may not think about Markov chains. But they are there.
Tennis — not one of the "Big Four" U.S. professional sports — enjoys a similarly mathematical structure. The flows of tennis games progress through a discrete set of states. A statistical researcher recently described a model for tennis matches.¹²
The remaining three of the "Big Four" U.S. professional sports lack the underlying mathematical structure of baseball and tennis. Football is arguably the most structured. Beyond ball possession and points scored, scoring drives follow through discrete states characterized by down, yards to go, and location of line of scrimmage.
Three of these states are continuous. The line of scrimmage and yards to go are represented by continuous numbers. Transitions from one state to the next are also continuous in nature. These factors lend themselves to statistical analysis less easily than discrete states and transitions of baseball.
Basketball and hockey are even less structured. Game flows result from continuous, random interactions between players. ESPN's survey finds that analytics adoption in basketball is more advanced than in hockey. This may be because basketball teams are smaller. Contributions by individual players have a greater bearing on the flow of play.
The path to analytics adoption.
Becoming an analytics organization involves both cultural and practical aspects of change. Analytics-driven cultures enjoy the distinct quality of thinking quantitatively. Members of such organizations habitually measure aspects of their work. They benchmark measurements of the most important aspects of their activities. They also measure performance against those benchmarks.Baseball, among major U.S. professional sports, enjoys a culture that is substantially analytical. Baseball statistics occupy a prominent place in the sport's fandom. Statistics are sufficiently important to the business of baseball for major commercial publishers to release large volumes.
Explaining the prominence of statistics in baseball in particular may be a "chicken-or-the-egg" question. That the structure of the game is fundamentally mathematical in nature certainly does not introduce any obstacles to a popular baseball-statistics sub-culture.
Organizational leaders seeking to inculcate analytics more deeply into their organizations must manage cultural change, first and foremost. Cultural change is perhaps the most formidable of management challenges. Management guru Peter Drucker's dictum "Culture eats strategy for breakfast" states the challenge well.
Leaders must also recognize the limitations of analytics. In order to apply analytics to a problem, we first must describe it with mathematical precision. Not everything does — or necessarily should — lend itself to this degree of precision.¹⁰
Assertions by exuberant advocates suggest that anything can be measured.¹³ Many such improvisational measurements require selection of proxies — substitutes for quantities that cannot be directly observed. Such proxies may lack precision — or sufficiently direct correlation to the desired quantity — to usefully reduce uncertainty.
Methods from the economics of information give us a rigorous approach to quantify the value of marginal uncertainty. Measurements and reporting based on proxies may not yield information-economic returns worthy of the required investment. Sports analytics gives us practical illustrations.
References
¹ M. Lewis, Moneyball — The Art of Winning an Unfair Game, Norton, 2004, http://goo.gl/f7N2Yu.
² "The great analytics rankings," espn.com, February 25, 2015, http://espn.go.com/espn/feature/story/_/id/12331388/the-great-analytics-rankings.
³ S. Lavalle, et al, "Big data, analytics and the path from insights to value," MITSloan Management Review, Winter 2011, http://goo.gl/8RSn5H.
⁴ T. H. Davenport, Analytics at work, Boston: 2010, Harvard Business Review Press, http://goo.gl/olZkKm.
⁵ B. James, The New Bill James Historical Baseball Abstract, New York: Free Press, June 13, 2003, http://goo.gl/Q7a0iA.
⁶ B. James, The Bill James Handbook 2015, Chicago: ACTA Publications, October 31, 2014, http://goo.gl/OEHinJ.
⁷ Baseball America 2015 Almanac: A Comprehensive Review of the 2014 Season, Baseball America, January 6, 2015, http://goo.gl/xCenv9.
⁸ M. Marchi and J. Albert, Analyzing Baseball Data with R, Boca Raton, FL: October 29, 2013, http://goo.gl/gx4J7r.
⁹ S. M. Shea and C. E. Baker, Basketball Analytics: Objective and Efficient Strategies for Understanding How Teams Win, CreateSpace Independent Publishing Platform, November 5, 2013, http://goo.gl/rE36pI.
⁵ B. James, The New Bill James Historical Baseball Abstract, New York: Free Press, June 13, 2003, http://goo.gl/Q7a0iA.
⁶ B. James, The Bill James Handbook 2015, Chicago: ACTA Publications, October 31, 2014, http://goo.gl/OEHinJ.
⁷ Baseball America 2015 Almanac: A Comprehensive Review of the 2014 Season, Baseball America, January 6, 2015, http://goo.gl/xCenv9.
⁸ M. Marchi and J. Albert, Analyzing Baseball Data with R, Boca Raton, FL: October 29, 2013, http://goo.gl/gx4J7r.
⁹ S. M. Shea and C. E. Baker, Basketball Analytics: Objective and Efficient Strategies for Understanding How Teams Win, CreateSpace Independent Publishing Platform, November 5, 2013, http://goo.gl/rE36pI.
¹⁰ J. M. Hall and M. E. Johnson, "When should a process be art, not science?" Harvard Business Review, March 2009, http://goo.gl/q6rGKB.
¹¹ J. R. Norris, Markov Chains, London: Cambridge University Press, Jul 28, 1998, http://goo.gl/J0nV8S.
¹¹ J. R. Norris, Markov Chains, London: Cambridge University Press, Jul 28, 1998, http://goo.gl/J0nV8S.
¹² C. Gray, "Game, set and stats," Significance, Royal Statistical Society, February 3, 2015, http://goo.gl/8wdgH7.
¹³ D. W. Hubbard, How to Measure Anything: Finding the Value of Intangibles in Business, New York: Wiley, 2014, http://goo.gl/cFuFTb.
¹³ D. W. Hubbard, How to Measure Anything: Finding the Value of Intangibles in Business, New York: Wiley, 2014, http://goo.gl/cFuFTb.
No comments:
Post a Comment