Millions of NFL fans have completed their fantasy football drafts by now. The NFL schedule was set months ago and the season kicks off on Thursday night, with the Minnesota Vikings taking on the New Orleans Saints in prime time.
For many fantasy fans, the player-drafting strategies start with the best intentions: I'll do my research, check out some pro football websites and make a list of the players I really want.
However, when draft day finally happens and you've neglected to do any research, that "strategy" quickly degrades into player selections based on "gut feelings" that may or may not correspond to actual statistical evidence, current news or historical player trends. (The five beers don't help matters, either.)
In short: Most people wing it. Surveys have demonstrated an analogous situation inside businesses: Executives struggle to choose between relying on various spreadsheets (which might not be up to date or even correct) and gut feelings when trying to make strategic decisions. (See CIO.com's To Hell with Business Intelligence: 40 Percent of Execs Trust Gut.)
Hetal Thaker, a product manager at IBM and fantasy football enthusiast, is one woman who is definitely not winging it. Thaker uses widely available IBM predictive analytics software that taps into the large pool of qualitative and quantitative information available to fantasy team managers everywhere.
[ Read how the NFL uses off-the-shelf software to create its complex schedule ]
Her enviable record (three league wins in the last five years, since turning to the software) speaks for itself. Thaker recently spoke with CIO.com Senior Editor Thomas Wailgum (who's already been mathematically eliminated from winning his fantasy football league) about fantasy football strategies, trash talking, and her love of analytics and the woeful Detroit Lions.
CIO.com: Of course you've got me thinking about how horribly I've drafted my team for this year. Thanks. Are your fellow league members aware of your predictive analytics secret?
Thaker: They're quite aware that I'm using predictive analytics, not that they like it very much because it gives me an edge.
CIO.com: "Predictive analytics" might be a term that scares some people off. How do you describe it to people in your leagues?
Thaker: It really gets into taking qualitative information—that textual, historical and current player information—and using that with the numeric and quantitative information. The combination is incredibly valuable. And you're making your predictions that much stronger.
CIO.com: So did you write your own application, or do you use IBM software? How does it work?
Thaker: I'm lucky because we have wonderful tools and apps at SPSS and IBM to use. One lets me get all the information I need and allows me to prepare and clean [the data] to make sure it's correct. And then I use our statistics product, which is a Windows-based desktop application, and I just pull in the data. The nice thing is that it uses any format—whether it's text or Excel.
Then I take that historical information and dump it into our modeling tool. It has a text-analytics piece, which allows me to take all of that qualitative information—the news stories, the injury reports, the analysts briefings—and rather than reading each individual piece, it lets me automatically categorize information: Basically put it into buckets of things as simple as negative or positive comments, such as "Are they hurt? On suspension?" So I have all these "positives" and "negatives," and you can keep it as simple as that as if you want. But the tool allows me to dig into my details if I really want to get there.
So, I take that historical quantitative information, take my qualitative information, which I've now categorized, and put all that into my model. And when we go a step further—and I'm not a modeling guru whatsoever—it's really this application that allows me to take all this information in and "auto-model" it. Meaning, I don't know which model is best [for predicting the best fantasy football draft]. I simply say: Here's my data, give me a bunch of models, and then it tells me which ones have the best fit or predictive value. From that, I pick my model to use.
Now I can predict a number of different elements: Let's say I want to predict the number of touchdowns a receiver is going to get or number of total yards that a quarterback is going to throw. Or maybe I just want to know what my total fantasy football points are [going to be]. Because at the end of the day that's what I'm going after.
It's no different for a business, because at the end of the day the business is trying to increase the bottom line. It's all about profits. But if you can focus on the right people, the right kinds of customers or right decisions to increase the bottom line—without doing a lot of work—then that's exactly what you're looking for.
CIO.com: You buck a couple of common stereotypes regarding football viewership and fantasy leagues: 1. You are a woman. 2. You are a bit geeky—and I say that with all due respect. Do you feel like you've broken down some barriers for women but who might be reluctant to join a league?
Thaker: I hope so. When you hear "predictive analytics" or "fantasy football" it tends to be intimidating. When I first started at IBM [after IBM acquired SPSS], I had to give a presentation to all of these managers on any topic I chose. I ended up picking fantasy football [and the analytics application I created].
The room was split 50-50—half men, half women. The men were interested—actually, they were more interested to know who was on my team. And the women got it. I know at least two women who sat in that room listening to how fantasy football worked for 30 minutes, who ended up playing in a [fantasy] league the year after.