From Fantasy Sports to analytics on the court, we sat down with Cory Jez, Coordinator, Basketball Analytics at the Utah Jazz to chat about his career in data. In this interview, you’ll learn about his career path to finding his dream job and how he helps keep the team competitive by using data to get the entire organization closer to their North Star Metric: winning the NBA championship. You can follow Cory on Twitter.
Tell us about your path to becoming a data scientist. How did you get involved with data?
My path to analytics and the data science part of it is probably like that if a lot of people in this space. I graduated school in 2011 with a degree in economics and started looking for my first job. I was entertaining things from applying to law school to possibly working in sales, but ultimately, I landed in an entry-level analyst role. I was hooked from there and it got me to learning SQL, becoming proficient in data visualization and testing out different BI tools. I started doing side projects that turned me onto Python and R and I continued to take graduate-level coursework and progressed through jobs into advanced analytics, machine learning, simulations, etc.
I spent the first 3-4 years of my career figuring out what I wanted to do within the broader analytics scope. In 2015, I had a quarter life crisis where I took a leap of faith and left a stereotypical corporate analytics job and started freelancing for the Pac-12 Conference. Up until then it had always been about building a tool to support my business user, but I had never been passionate about the data I was using or building the tool for. I really wanted to get myself into an area where I was passionate about the decisions that were being made and that I could play a role in the decision-making process. Sports was the next logical step for me.
Consulting for the Pac-12 Conference led me to FanDuel where I worked for almost 2 years as an Analytics Manager. However, my end goal has always been to be in sports analytics within a front office for a professional franchise. So when the opportunity with the Jazz came up, I jumped at it.
What helped me land my dream job so early in my career was developing what I was good at within the first few years of my career and then applying it to something I was passionate about
What is a data project that has been particularly inspiring to you?
The cool thing about sports analytics in general, regardless of what field you are in, is that probably 95% of the data you need is public information. So even before I had a job in sports, I was able to start working on projects that would help advance my career. One in particular was creating a model to pretty accurately estimate the shots a player would take and the success of each shot.
From scraping data from the web and getting it from an API, cleaning the data, sorting a database, doing analysis, creating a machine learning model and finally presenting it in a dashboard, I was able to run the whole lifecycle of an analytics project. This really helped me when it was time to interview. I was able to speak to some of the specific pain points of sports analytics.
What are some ways you keep ahead of the curve, education wise, when it comes to data science?
I try to read for an hour or so each day, both in the analytics field, as well as in the basketball world, which is my domain now. I’ve found it really practical to set up RSS feeds from blogs that I link into a Slack channel, so I’ve basically got a list of articles / interesting things waiting for me each day.
My top reads are:
Ben Falk. He runs a blog called Cleaning the Glass. He was the VP of Basketball Strategy with the Sixers and Basketball Analytics Manager with the Blazers, so he provides a perspective into what's going on in the NBA that is relevant to my day-to-day.
In your opinion, is there a “right” or “wrong” way to use data in modern business? Please explain.
Do you want to be transactional or transformational? A stakeholder might come to you asking for a simple question but in reality, their question might have more layers than what they send in the initial request.
The stakeholder might think they are helping by asking x, y, z and serving it up in an easy way for their analyst, but really they are narrowing the scope. While it can be very quick to answer someone’s question, it falls on you, the analyst, to ask more questions.
This is not always easy, it takes a good working relationship, a stakeholder who is open to new ideas and people who are open to being told they are not using the best metric or approach.
Whenever it’s possible to take the transformational approach of probing the stakeholder what they are really trying to ask, I go that route. But there are also certain cases when you are down to a deadline, an hour before a game, and you just need to get them the numbers they need.
Neither approach is right or wrong, but you need to understand when it is best to dig deeper. Long term you want to become less about counting the number of widgets you have and be more about decision support. Analytics is a service-based decision support field and you can’t do that without having conversations with people.
What are some of your core challenges as a Coordinator, Basketball Analytics?
The biggest challenge is balancing the complexity of the newest tools, technologies and cutting-edge analytics with the needs of your stakeholders. The adage I live by is: simpler is better.
While we always want to say we are good at the hottest new thing in analytics, that can create a dilemma in adoption and understanding of our findings.
As you move into tree-based models or network based models you lose that interpretability of that model. While your model might be significantly more accurate, the loss of the interpretability hurts the adoption and the understanding of it. For example, predicting draft players, the first question is always “why”.
So if you can reach an accuracy threshold with a simpler model, do it. Don’t make things complex to show off your code - that isn’t the point. Your goal should always be to try to make things as actionable and easy to understand as possible.
At what point do you value a scout’s ‘gut feeling’ over statistical analysis?
I think the paradigm of “scouts vs. analysis” is over and done with. Both of these things are vital pieces of information which are considered in decision making. I think the other misconception is that all scouts just have a “gut feeling” - most people who are scouts at the NBA level are there because they’re the best, and they have a very analytical framework for how they break down a player’s skills. This allows us to compare players to each other in quantitative ways - even if the assessments are inherently subjective.
How is data being used by Utah Jazz? Who is involved in the organization and how?
We use both qualitative and quantitative data, like any corporation, to reach specific goals throughout the year. However, we all have a common goal, our North Star Metric -- we are trying to win a championship.
So while our overarching goal for the year is to win a championship, we have smaller objectives we need to achieve along the way to get there. For example, finding players that will help us get to that goal (better shooters, a more veteran player to bring more leadership, etc.). So, we break the overall goal down into smaller, measurable achievable tasks.
We are constantly balancing quantitative data, like usage rate, or their number of assists or True Shooting percentage, with qualitative data like their relationship with their prior coach.
What is most interesting to me, is when your intuition and the objective analytics don’t match up. That is where you have a real opportunity for growth and allows you to challenge your beliefs and come to new conclusions to better reach that end goal.
What types of ‘Moneyball’ principles are applied to Basketball as compared to Baseball?
There’s a handful of stats out there in the public forum that have their origins in baseball. For example, Bill James’ original method for estimating if a team will win a game was amended to the NBA by Daryl Morey to predict won-lost percentages. There is also WARP, wins above replacement player, created by Kevin Pelton, who took the idea from baseball’s WAR, to measure a player's total contributions to his team.
Stats like these are used very often to predict an outcome of a season. In general, there is a concept of exploiting inefficiencies to gain a competitive edge -- Can I find a player who is undervalued in the market based on their stats, but make a solid projection that he will get better and overvalue him relative in the market while not making a gamble? That is the core of sports analytics, looking for points of value where the general market doesn’t see them. Combining data and qualitative information from what the scouts saw can get help give you the competitive edge you need to find a key player that can work with your team to get to the championship.
What is your view on data analytics tools? What are some new innovations and what still remains a challenge from a user perspective?
Analytics in the NBA have come to center stage. Teams are quickly building entire departments and we are trying to catch up to where the Facebooks and Ubers of the world are.
In comparison, our analytics team is still small so I am always looking to make our lives easier and do things more efficiently. Because of that, I try to stay up-to-date and understand the new best practices in analytics to create better forecasts and reports.
In general, a good way to learn about new tools or implementations is by talking with colleagues, going to conferences and learning how others handle problems or overcome hurdles within your industry. However, the NBA is a highly competitive industry, and so there is less collaboration with people outside of your organization. Because of this, you have to stay hyper-focused on where the analytics industry is at large in order to maintain your advantages.