Every inch of grass at the southwest corner of the City of Palms Park in Fort Myers, Fla., lay flat after dozens of hard-core baseball fans had braved a night of sleeping bags and fire ants to snag the few remaining standing-room-only tickets for a spring training contest between the Boston Red Sox and the New York Yankees. Days earlier, a single ticket sold for $500 on eBay, an unheard-of sum for a spring training game.
The real baseball season wouldnt start for another month, but the Red Sox Nation couldnt wait another day. Since Sox owner Harry Frazee traded Babe Ruth to the Yankees in 1920 to finance his girlfriends Broadway play, the Sox havent won a World Series—an affliction known throughout baseball as the “Curse of the Bambino.” But damn the curse. This year is going to be different. The numbers say so.
Red Sox owners John Henry and Tom Werner have hired some of the best analytical minds in the business and have spared little effort in acquiring sophisticated scouting software, computerized video analysis and business intelligence tools for mining the stacks of statistics at their disposal. The goal: Identify the best talent available, get it before their rivals do, and then figure out just how long to keep it before it stops producing.
Its human capital management on steroids.
Baseball is a business where employee performance can be measured down to every swing, step or throw taken. Baselines are easy to establish, not just in chalk, but on every facet of the athletes physical characteristics, such as height, weight and medical condition, to minutely defined activities, including arm strength, hitting discipline and mental errors. If theres not an available statistic, one is created, through ratings.
Today, every season of every players career, from school play onward, is minutely chronicled. The Red Sox even require players in their farm system to keep a log of their every at-bat.
Imagine if your employees were required to log every significant phone call, presentation, report or other measurable “output” of their days.
And then imagine if you tied this data to some ultimate objective. In baseball, the goal is winning games. And an up-and-coming business executive, such as Red Sox general manager Theo Epstein, can figure out whether a player generates more than the average number of wins. If so, he will know whether to pay $12 million a year for the services of a star pitcher such as Curt Schilling, acquired over the winter from the Arizona Diamondbacks. Or pass on an all-star shortstop, such as Alex Rodriguez, when a threshold that works out to $20.25 million a year is reached.
Theres clearly a payoff for companies that can adapt their own information systems to keep increasing amounts of performance information—think “time spent with customers per day” or “trouble tickets resolved”—that can be delivered to managers and executives. Such measures help identify, recruit and retain the best talent, not necessarily the most expensive. But if a person is that much more productive, the higher salary can lower the overall cost of doing business.
Check out eWEEK.coms Enterprise Applications Center at http://enterpriseapps.eweek.com for the latest news, reviews and analysis about productivity and business solutions.
“If every single team adopted this management strategy with equal skill, the [low revenue] teams would forever be stuck in the cellar,” says Roger Noll, a Stanford University professor and respected authority on the economics of baseball. Because wealthy teams, like the Yankees, can always simply buy the best talent.
Instead, innovative means of statistically identifying the winning characteristics of “human capital” are the only way companies can get the players who will overcome financial constraints. “The same premise applies in business, but the effects are much more amplified in sports, says Noll.
Avid baseball fans already know this: Low-budget teams such as the Minnesota Twins, Oakland Athletics and Florida Marlins—who beat the Yankees in last years World Series— already have applied this premise to last deep into the playoffs.
Now the Red Sox hope to do the same.
Next Page: Sabermetrics: A whole new ball game.
Be sure to add our eWEEK.com enterprise applications news feed to your RSS newsreader or My Yahoo page
Sabermetrics
A Whole New Ballgame
The burden resting on Epsteins shoulders is impossibly heavy. He grew up barely a mile from Fenway Park, home of the Red Sox, and is just as rabid a fan as those second-guessing his decisions today.
Epstein was chosen general manager in 2002—becoming, at 28, the youngest GM in the majors—largely because he shared the Sox owners philosophy of putting numbers and analytics to work on the playing field. He is one of a growing number of general managers in major league baseball who are students and practitioners of Sabermetrics.
Sabermetrics is the mathematical analysis of player batting and pitching performances. Baseball is a sport already brimming with statistics, yet Sabermetrics—the term is derived from the acronym SABR, which stands for the Society for American Baseball Research, a community of baseball enthusiasts—is a departure from traditional player metrics such as runs batted in (RBI) and batting average.
RBI is a faulty gauge of talent, Sabermetricians say, because it is heavily dependent on where a player sits in the batting order. And batting average, they say, does not take into account how adept a hitter is at working a pitcher and drawing walks—a walk gets a player to first base just as well as a hit.
Instead, Sabermetricians have come up with measures that more accurately reflect a players value toward achieving a win, such as “runs created.” This statistic counts the number of times a batter gets on base, be it by walk or hit, and factors in an added value for the power of a hit, be it a single or a home run. The purpose is to determine what the batter does at the plate to create an opportunity for his team to score a run.
Done right, practitioners say, Sabermetrics can help teams more accurately find minor league prospects who will succeed in the big leagues. Similarly, Sabermetricians claim they can use the analysis to determine which major league players deserve a $10 million paycheck and which “stars” can be dumped.
Prior to the rise of Sabermetrics, teams primarily relied on their scouting systems for finding and drafting talent. The problem, as Oakland As general manager Billy Beane saw it, was the system was too much of a crapshoot, particularly when it came to drafting young players out of high school who had yet to mature physically and emotionally. “We didnt have the resources of the other clubs, so we couldnt afford to gamble,” Beane said recently at a conference in Chicago, referring to the teams low revenue, which is estimated at $110 million, the eighth lowest of all major league teams.
Beane, one of the best-known practitioners of Sabermetrics, has used baseball analytics to consistently field one of the best teams in baseball on one of the smallest payrolls—$50 million in 2003, the eighth lowest in the league and far below the Yankees $150 million payroll.
The As just missed making the playoffs in 1999 but have been there every year since. Theyve done it at an average cost per win of only $388,000 over that five-year period, the best in the league. The Yankees have been to the playoffs every year during that stretch, but at a cost of $1.23 million per win.
Beane set the bar; now the Sox are attempting to raise it. In addition to Epstein, they have hired Bill James, the so-called father of Sabermetrics, as a special advisor.
The object isnt to be cheap, says Epstein. After all, the Sox, one of the most lucrative franchises in the league with annual revenue of $200 million, open the season with the second highest salaried team in the majors—$125 million.
Its about making the best use of the teams resources so that when its time to re-sign Pedro Martinez ($17.5 million salary in 2004), Manny Ramirez ($20.5 million) or Curt Schilling ($12 million), the money will be there. “We are fiscally responsible because the alternative would be a disaster,” he says. “Fiscal irresponsibility is the single quickest way to hamstring a franchise for a decade.” The prime example is 2003s American League West cellar dwellers, the Texas Rangers, who broke the bank in 2000 on superstar Alex Rodriguez, signing him to a 10-year, $252-million deal, then lacked the funds to pay for a supporting cast.
As even a casual baseball observer knows, the Sox almost signed Rodriguez in the off-season. But the bottom line was that the team set a $20.25-million-a-year threshold it was not going to go over. After numerous failed attempts to structure a deal that was at or under that salary, the Sox walked away from the table. Rodriguez eventually signed with the Yankees after the Yankees got the Rangers to pay a large portion of the players salary.
But the ability to combine statistical smarts with the wealth to make the deals should give the Sox an incredible edge.
Next Page: Heavy-hitting software.
Heavy
-Hitting Software”>
Heavy-Hitting Software
It takes some serious number crunching for the Sox and other teams to decide which players to sign and how much to pay them.
To find new talent, for example, the teams have to sift through the stats on thousands of players in high schools and colleges and the hundreds of thousands of at-bats and pitches thrown in any given season. In the old days it was all done on paper: Scouts would send in reports from games in the field by mail or fax, and the reports would be filed into thick binders. When it came time to find a top-ranked left-handed pitcher, general managers would thumb through the binders and ask the scouts for their top choices.
One of the first moves Epstein made when he joined the Sox was to bring in a software package called ScoutAdvisor from E Solutions, a little-known Tampa, Fla., company. Epstein had used the software, which keeps track of player talent from the minors to the pros, while with the San Diego Padres, where he was director of player development. E Solutions now counts nine Major League Baseball teams on its client list. IBM has a similar package called PROS, in use by a similar number of teams. Both are based on the Lotus Domino platform (see Dossier, page 46).
Michael Morizio, co-owner of E Solutions, says the software was originally developed five years ago as a custom project for the New York Yankees. The team had begun building its own software to store and analyze scouting reports, but the project bogged down. It was simply too much of an undertaking for the teams small technology staff.
Through Yankee owner George Steinbrenners Tampa contacts (his primary residence is in the area), E Solutions was called in to finish the job. After completing the software, the company saw the potential to build and expand the offering. Rather than just focus on scouting, Morizio saw the potential to build a complete player management system that could track player development as well as provide a teams front office with a window into the rest of the league.
E Solutions has since gone full circle with the Yankees, announcing plans in March to replace the teams custom system with ScoutAdvisor. The software is licensed for about $50,000 to $75,000 per year depending on the options.
The heart of the system is housed in a 15,000-square-foot data center in downtown Tampa. There, computers store raw data on baseball players, such as high school and college records, family backgrounds, psychological profiles and medical histories, culled from a wide range of sources.
E Solutions pulls in the feeds daily from a range of data sources, including the Major League Baseball Scouting Bureau, SportsTicker Enterprises and STATS Inc. The Major League Scouting Bureau has administered an Athletic Success Profile test to prospects every year since 1974, asking 110 questions designed to uncover their psychological profile. The questions measure 11 attributes including drive, endurance, leadership, self-confidence, emotional control, mental toughness, coachability and trust.
SportsTicker, on the other hand, has been serving up game reports since 1909, while STATS employs a small army of “reporters” at games to gather a wealth of statistics such as the pitch count at the end of an at-bat, types of pitches and where balls land in the field after being hit.
All of these statistics can be accessed and analyzed through ScoutAdvisor. The software can quickly retrieve, for example, the fact that Curt Schilling is one of only eight major league baseball players born in Alaska.
The ScoutAdvisor software package also is designed to help teams gather original data on players. ScoutAdvisor features a variety of modules tailored for amateur scouting, pro scouting, international scouting, medical injury tracking and major league player transaction reporting, among others.
The amateur scouting module allows scouts to file such stats as hitting power or fielding ability and commentary such as “exhibits a lot of hustle,” from the field using laptops or personal digital assistants and an Internet connection.
The software features standard evaluation forms, which may be customized by a team. “Were able to add evaluation lines like, How high can the player jump? without any difficulty,” says Sox assistant scouting director Amiel Sawdaye.
In their standard format, the forms are used to collect an exhaustive amount of data on a prospect, from his physical description (height, weight, build, etc.) to evaluations that explore each of a position players five main tools: running, fielding, throwing, hitting and hitting with power. Attributes that can be rated include hitting ability, plate discipline, power frequency, raw power, arm strength, arm accuracy, fielding ability, running speed and potential. For pitchers, the reports can record everything from fastball movement to ratings for curveballs, sliders, changeups, control, radar-gun readings (pitch speed), makeup (poise, habits, hustle, self-confidence, desire) and an opinion on potential.
The pro scouting module is similar to the amateur module, except it goes even deeper. For example, additional categories for pitchers include attributes such as arm angle, arm action, deception and release times. A look at the raw pitching statistics for Chad Bradford of the Oakland As would quickly lead most scouts to the conclusion that he isnt fit for the majors. His mid-80s mph fastball is minor league material. However, scouts looking at arm angle will discover Bradford uses a “submarine” style of delivery with his knuckles sometimes scraping the mound. Hitters have more trouble with his deceptive form than they do facing a 95 mph fastball.
Another popular feature is the player development module, which allows managers to track a players progress based on pre-set criteria. A manager can use the software, for example, to look for declining fastball velocity, in the case of a pitcher, or a tendency to swing at first pitches, in the case of a hitter.
The big advantage of ScoutAdvisor, says Josh Byrnes, the 32-year-old Haverford graduate Epstein hired to crunch numbers, is that it can slice and dice player data any way you ask. You can run a query report, for example, to find a catcher with a hitting ability greater than 60 (on a scale from 0 to 100), with an arm strength greater than 50, a speed rating from home to first base greater than 70, instincts greater than 60, and similar ratings for on-base percentage, aggressiveness, mental toughness and signability. Scouts generally rank players within a range of 20 to 80, with the assumption that no one worth rating is lower than a 20 and no one is better than an 80. Teams with a need for a multifaceted all-star—and money to spend—can search for players with 80s across the board. But teams on a budget need to trade off a high score in certain areas to find a lower-salaried player.
There are 32 pre-set fields that can be queried, and teams can add more of their own choosing. The software will then go through the universe of known players and pull up the top prospects according to the criteria. Its light-years away from the way teams used to drill through stacks of binders, says Byrnes.
To illustrate his point, Byrnes calls up the player data on J.J. Davis, a big, power-hitting right fielder trying to crack the Pittsburgh Pirates starting lineup. On this day the Red Sox were getting a first-hand look at Davis during a spring training contest against the Pirates. Byrnes laptop tapped into a server back at Fenway Park and began flowing in page after page of data on Davis (6-foot-5, 250 pounds, number-one draft pick in 1997). The information included his minor league history (hit 26 home runs at Class AAA Nashville), to his psychological profile (has had a discipline problem, quitting Venezuela winter team after 15 games), through to his known injuries (recurring hamstring problem) and personal background (single, turned down baseball, basketball and football scholarships to sign with the Pirates).
Software companies are now looking to develop statistical analysis packages to evaluate corporate personnel. Vendors such as Unicru of Beaverton, Ore., are coming out with sophisticated software offerings that can develop profiles of potential employees using questionnaires and analysis of previous work experience. Its the business equivalent of a scouting report.
Next Page: Finding business all-stars.
Finding Business All
-Stars”>
Finding Business All-Stars
Unicru provides its clients, such as Albertsons, Target and The Sports Authority, with a kiosk-based system to accept job applications and screen candidates. In addition to general information and employment history, the kiosks ask applicants a variety of questions to determine if they have the right personality to sell on commission, or whether theyd be a courteous and attentive waiter. These include true-or-false questions such as, “I would rather sit around and read a book than go to a party with lots of people,” and, “You dont act polite when you dont want to.”
Within minutes after a candidate fills out an application, the system generates an assessment and also provides a hiring manager with a set of questions designed to probe a little further into areas of concern. If theres a worry that the candidate might not stay long with the company, the system prompts the interviewer to ask if the candidate remembers the name of his or her boss from two jobs ago, or three or four years ago. If the candidate does remember, the odds of retention are greater; if not, its another strike against.
Unicru chief scientist David Scarborough says the Internet, human-resources applications and services such as online job boards now make it possible to do the same kind of analysis that baseball uses. “Its not that there was a lack of data in the past,” says Scarborough. “There was a lack of data in a format that could be accessed and analyzed.”
Now, however, its not that hard to see how a Fortune 500 company could use its own sales or human capital management software to find the best heavy-hitting salesperson to fill a new opening. Criteria such as length of time to close a sale, size of average deal and number of deals closed per month are common metrics used in evaluations. Siebel Systems, which sells such metrics-based sales software, is known to be relentless in using such metrics, firing the bottom 5% of its salespeople each quarter.
Indeed, corporations are increasingly collecting vast amounts of data about their employees through sophisticated software offerings from vendors such as PeopleSoft, SAP and Siebel. Now these companies have to start using it to identify top performers, and tie it back to the training theyve had, where they went to school, employment history and personal characteristics. Then, says Jason Averbook, director of marketing for PeopleSofts human capital management division, corporations can start making smarter and more cost-effective decisions about hiring and managing people.
Some organizations are already making major league discoveries.
Charles Brooks, a senior human-resources officer with the Georgia Merit System, an agency that provides HR services to the State of Georgia, wondered, for example, why it was that some agents in the states Child Support Enforcement (CSE) program were much more successful at collecting money from deadbeat dads than others.
When he began his research in 1998, Brooks noted that CSE agents collected child support payments ranging from $172,107 to $1.68 million per year. The average amount was $802,237 per agent. Upon further study, he found that performers in the top 13.5% collected at least $1.14 million, or 42% more than the average, while low performers, those in the bottom 13.5%, collected at least 42%, or $341,436, less than the average.
Brooks began looking into why there was such a disparity. He was initially told not to go there, that it wasnt a factor that could be fairly measured—some dads, after all, were just harder to collect from than others.
But he pushed ahead and began studying the top performers in the agency, poring over computerized records and looking for common characteristics. By 2001 Brooks had developed a survey that could predict with 95% accuracy whether an agent was a top performer. In other words, he could ask the agent a series of questions, and based on the answers he could tell if that person had the right stuff. A top performer, for example, provided this answer to the question: How do you cope with failure? “I talk to my colleagues and try to determine what Im doing wrong, so I can fix it.” A low performer, on the other hand, provided this answer: “Its a numbers game; I just tell myself Ill get the next one.”
The results were impressive. The average collected per agent jumped 42%, from $802,237 in 1998 to $1.15 million by 2003. Not only could higher-performing agents be recruited, but the analysis could be used to identify underperforming agents and coach them in skills that could improve their success. The process resulted in an additional $24 million per year in increased collections for the agency. The state paid out $10 million less in welfare payments because it got more from the delinquent dads, and agent turnover has been decreased by 50%, resulting in further savings of $30 million over the four years the program has been in effect.
Other companies are asking the same types of questions of their human-resources data to find, hire and keep talent.
For years Dow Chemical Co., based in Midland, Mich., sent recruiters to both Ivy League and second-tier schools to find top M.B.A. prospects. But then it started sifting through its human-resources records, in this case kept in PeopleSoft, to find out exactly which schools most of its recruits had come from, which schools produced employees who stayed with the company longer and which produced the most executives. The company also keeps detailed records on such factors as how many on-site interviews were held and what offers were made and accepted.
The results were eye-opening, says Bill McNeill, head of the companys M.B.A. recruitment program. The study showed that Dow was basically wasting its time at the Ivy League schools. M.B.A. grads from the likes of Harvard and Yale typically didnt want to move to Midland, asked for more money than Dow was willing to pay, and didnt stay in the job as long. (The company would not release percentages.) Dow will typically send recruiters to campuses several times a year to woo top M.B.A.s—McNeill estimates it costs as much as $80,000 to reel in a new hire. After looking at the numbers, the company shifted its resources to universities such as Michigan State, Brigham Young and Purdue where Dow had greater success in attracting and keeping M.B.A.s and other highly sought-after graduates.
The Payoff Pitch
The Payoff Pitch
In addition to finding and developing talent, one of the key uses of a statistics system for personnel is figuring out how much to pay people.
The Red Sox, and for that matter any major league team, will not reveal exactly how they decide a players worth. Market forces certainly play a major part, but teams must also decide whether paying tens of millions of dollars to a player will generate appropriate returns. Oakland As GM Beane said recently at the conference in Chicago that he knows exactly what a player is worth to him based on how many wins hell contribute to the team in a season. “But Im not going to tell you it,” he laughed.
However, the fact that the Sox hired Bill James provides some insights into how they may be tying their analysis to their checkbook. James has created several methods to estimate the value of a player, including a formula he published in 2002 called Win Shares. Win Shares takes a mathematical approach to evaluating the contribution of an individual player to a teams overall performance. It considers a variety of statistics, including pitching, hitting, walks and defensive contributions, and adjusts them for a particular ballpark and the league. Some parks are more hitter-friendly, such as Coors Field, home of the Colorado Rockies, where the thin air allows batted balls to carry farther.
At its basic level, a single Win Share is one-third of the credit for winning a game. So if the Red Sox win 95 games this year, there will be 285 Win Shares to spread among players. Spread across the 25-man roster, an average player might wind up with 10 Win Shares.
But consider the Sox acquisition in the off-season of star pitcher Curt Schilling. According to James system, Schilling achieved 24 Win Shares in the 2002 season and 15 in 2003, for an average of 19.5. To get Schilling, the Sox had to assume his $12 million salary. Other pitchers with similar Win Shares in 2003 were the Sox Pedro Martinez (20 Win Shares, $17.5 million salary), the Yankees Mike Mussina (19, $16 million) and Seattles Jamie Moyer (18, $7 million). Schillings $12 million price tag is at the low end of the range.
Alex Rodriguez led the league with 32 Win Shares in 2003 and the Sox would have loved to sign him, admits GM Epstein. The bottom line, however, is that the team would not cross its threshold of $20.25 million a year, and had to walk away from the table.
Corporations play out their own version of Win Shares every day. A new salesperson isnt hired unless it is calculated that he or she will bring in more revenue than the cost of that persons compensation. Boards of directors must determine whether paying $28 million for a new chief executive will provide an appropriate return in shareholder value. There just isnt the level of sophistication with business statistics as there is in baseball.
And the Sox are using all the data at their disposal and applying statistical analysis to find the best and brightest. Its how they decided to cut Jose Offerman and John Valentin, two underperforming infielders with combined salaries of $6 million, and were able to pick up sluggers David Ortiz and Kevin Millar for a combined $3.25 million. Together, Ortiz and Millar contributed 162 runs in 2003, while Offerman and Valentin had contributed 66 runs in 2002.
Indeed, stereotypes and conventional thinking are being tossed aside, unless the numbers back it up. As James states, the goal is to let the data take on the power of language— to let it determine what is true and what isnt.
The guidelines are being set because the data says its the key to winning.
The baseball season is underway, and most Red Sox fans dont know or care what kind of computer or software Epstein and his crew are using behind the scenes. Its the results that matter. And only one is truly acceptable: a World Series title.
