Major league analytics
How data helps drive draft decisions for the St. Louis Cardinals
Sig Mejdal quotes Yogi Berra: “It’s tough to make predictions – especially about the future.”
Mejdal, Director of Amateur Draft Analytics for the world champion St. Louis Cardinals, can relate. Mejdal’s job – collecting data, cleaning it, analyzing it and then presenting it to the Cardinals’ front office – is an exacting undertaking. Complementing the eyeball analysis of the Cardinals’ seasoned scouting staff, Mejdal gathers every available statistic on a potential signee’s past performance.
Because humans are humans, and less than fully predictable, determining the outcomes will remain an imprecise science. But JMP statistical discovery software is making it less so. JMP is helping Mejdal maximize the use of data – and that’s an advantage a relatively small-market team like the Cardinals can’t afford not to have.
The margin for error is much narrower for teams with relatively limited budgets, says Mejdal, who joined the Cardinals in 2005. “The search for any sort of competitive advantage is never-ending.
“JMP gives me a lot of confidence that I’m not missing something important,” he says. “It helps give me the assurance that I’ve squeezed all I can out of the data.”
The results of taking full advantage of all resources available are there on the Cardinals’ bottom line: a premier organization, top to bottom, on the field and off, that has delivered to its fans four trips to the playoffs and two world championships since ’05.
The ‘misfit toys’
The release of Michael Lewis’ book Moneyball in 2003 brought the use of statistics in baseball in evaluating talent to the general public’s attention, but the practice goes back much
further. Just to name a few: Branch Rickey and the Brooklyn Dodgers relied on a quant named Allan Roth back in the late ’40s. Earnshaw Cook worked on the Manhattan Project before becoming obsessed with a quest for the perfect baseball statistic. And Bill James – who coined the term“sabermetrics,” referencing the Society for American Baseball Research (SABR), and is widely considered baseball’s No. 1 numbers guru – had been toiling over box scores, while working as a security guard, since the mid-’70s.
Mejdal has been at it for a while as well, having joined SABR in grade school, a time in which he spent most of his waking hours playing the board game All-Star Baseball. Mejdal says he would use regression analysis when playing the game – presumably not a common practice among 11-year-olds.
The quant bug had bit, Mejdal persisted, and a scant few decades later, in 2004, came the call he’d been waiting for – from St. Louis Cardinals’ VP of Scouting and Player Development Jeff Luhnow, saying, “I received your proposal, it looks interesting, and I’d like to talk to you about it.”
Mejdal’s job now is to complement the information gathered through scouting, to help discover the hidden gems, players at smaller colleges or those Lewis referred to as the “misfit toys,” guys who perhaps don’t look quite right – maybe they’re a little overweight, too short, don’t throw with the right release point or have an awkward stance at the plate.
There are sometimes players who have excelled at the high school or collegiate levels but are nonetheless overlooked, Mejdal says. He uses JMP to help discover them.
“Playing a role in bringing some of those guys into the organization has been a wonderful experience,” he says.
Mejdal has been working with Luhnow since he came on board in 2005. Since that time, the Cardinals have had more draft picks reach the major leagues than any other organization.
A 2011 Baseball America article wrote of Luhnow’s ascent to his current position: “He and his staff worked diligently to break down NCAA Division I stats, but also spoke of ‘cracking the code’ on numbers from smaller colleges.”
The last few innings of that incredible Game 6 of the 2011 World Series featured a few of the Cardinals’ draft coups.
Allen Craig – an eighth-round pick in 2006 with a reported $15,000 signing bonus out of a draft budget of millions – hit an eighth-inning home run to start the comeback. The 10th-inning comeback included leadoff singles from Daniel Descalso – a pickup from UC Davis, a school that had never had a major league position player – and from Jon Jay, both recent draftees.
While Mejdal is happy with the results, he’s quick to add, “It’s all about the process. In a probabilistic industry like we’re in, results are great, but you can’t get too carried away with them. We all know that there is much variability within results, and the attention should instead go to improving the process.
“No doubt, it’s so rewarding to see players that we’ve had something to do with succeed in the major leagues. Of course, I get a kick out of that. But really it’s improvements in the process that we’ve implemented under Jeff that I am most proud of. That’s the real measure of success. And, yes, JMP has been a big help throughout that.”
Mejdal hints at a process in which the players are treated as a collection of attributes, and that JMP has helped make sense of how those attributes combine to predict likelihoods of success.
Scouts gather an enormous amount of information, and Mejdal emphasizes that their expert evaluations play a huge role in driving draft decisions.
The Cardinals’ most recent No. 1 draft pick is a great example of the success of these combined efforts. Kolten Wong was the shortest player, at 5’9”, ever taken in the first round. He’s off to a great start, having hit .335 for the Quad Cities River Bandits, a Cardinals Single-A affiliate, in his first season.
“For the analysis of large data files, it’s all JMP,” Mejdal says. “It’s made things so much easier.
“Just the speed with which you can really explore the data and the relationships with JMP made analysis a lot easier and gave me more confidence, especially with large data sets. I didn’t previously have software that was even close to the flexibility to handle that.
“So much of what I’m doing is regressions – multiple linear or multiple nonlinear regressions – and the flexibility and the capabilities with JMP, of forcing in an attribute or variable, or removing that variable with just the click of a button, seeing how it affects other variables – it’s the best I’ve seen, especially, again, when the data file is huge.”
Mejdal is a frequent user of the JMP Profiler: “It’s great that you can allow each predictor variable to be unconstrained, to get a sense of whether you’re missing something.
“I also love the heat maps. They’re such a wonderful visualization tool, so much better than a simple scattergram.”
He was also an early adopter of JMP 10, and raves about the sigmoidal curve-fitting features: “I’ve never seen software that can fit the monotonic curves into such a large data set so quickly. That’s been a great time saver.
“But what I honestly love most about JMP is how it can access an industrial-size database and put it into an intuitive interface, and in a way that you can so easily manipulate the data.”
Applying the visuals
While the recently released movie version of Moneyball depicts the relationship between scouts and quants as an adversarial one, Mejdal underscores that that’s certainly not the case with the Cardinals.
“In my opinion,” he says, “Moneyball framed the question as an ‘either/or’ proposition. But it’s very much a collaborative process. It’s all information, and we want any information we can gather. You’re not going to handicap yourself with the ‘either/or’ question. It’s really the ‘and’ question.”
The Cardinals’ philosophy is to apply both their scouts’ expertise and quantitative analysis of data from multiple sources for maximum predictive ability.
Mejdal says Cardinals GM John Mozeliak has always been a strong advocate for “breaking down the artificial walls” between traditional scouting and analytical analysis. Scouting reports, medical evaluations, contract data and statistical analysis all go into the mix.
“I was hired by someone [Luhnow] who believes that analytics has something to add,” Mejdal says. “So there is certainly an embracing of curiosity about it all. The quantitative analysis has been introduced in a thoughtful, gradual way.”
Still, decision makers often remain at least a little skeptical of analytics, “and rightly so,” Mejdal says, especially if the data appears to come from a black box. The biggest hurdle, he says, is getting decision makers to sometimes step back and rethink what they initially see and feel.
JMP plays an important role in this process. Mejdal uses its visualization tools to present his findings as effectively as possible.
“That’s been really very helpful. P-values and regressions aren’t always a daily part of the decision makers’ lives, and illustrating findings with, say, a visualization like a heat map so often really drives home the point – especially when you have JMP live, and you can select different areas and find the particular players who are representing that hot spot. I think that makes for a more compelling illustration.”
Using that illustration to augment anecdotal information – for example, comparing a prospect to a current player and saying, “This is what we can likely expect” – provides a more complete understanding of a player’s potential.
“The use of data visualization, in combination with compelling storytelling,” Mejdal says, “can often help direct decision makers to optimal baseball decisions.”
World championships notwithstanding, the search for new talent must go on. For Mejdal, that means continually collecting data, cleaning it, turning it this way and that, and brainstorming on additional attributes that should be collected. He’s always looking for improvements and honing his models.
“So much of the job is trying to find relationships between the data you have on players and what then becomes of them. What variables came most into play? I’d describe all those relationships as insights to some degree. JMP has such wonderful regression capabilities – the ability to lock variables in or out, lots of permutations – and that really greases the skids to discover and analyze those relationships – to gain those insights.”
A player’s evaluation depends on an organization’s methodologies, Mejdal says, “and to be a part of this system and see them in the major leagues is amazingly rewarding – especially a guy who might not have been as appreciated without a combination of those methodologies.
“The use of a tool like JMP helps ensure that players are evaluated using all information available. It helps identify players who might otherwise have been overlooked. That inspires me.”
Note: In December 2011, the Cardinals’ VP of Scouting and Player Development, Jeff Luhnow, was hired as General Manager of the Houston Astros. Mejdal subsequently joined him as the Astros’ Director of Decision Sciences.
“JMP gives me a lot of confidence that I’m not missing something important. It helps give me the assurance that I’ve squeezed all I can out of the data.”
Director of Amateur Draft Analytics,
St. Louis Cardinals
To complement traditional scouting with analytics, evaluating and predicting the performance and value of baseball players, and thereby making more-informed acquisitions.
The world champion St. Louis Cardinals are using JMP® statistical discovery software from SAS to help make data-driven decisions in the yearly amateur draft.
The Cardinals front office has powerful decision-making tools that incorporate both savvy scouting and leading-edge data analysis.
Download a PDF of this story.
Contact JMP Sales