Interview: Pallas Horwitz, Blue Shell Games on Why Gaming Analytics is Not a Piece of Cake
Tags: Advice, Blue Shell Games, Challenges, Machine Learning, Missing Values, Pallas Horwitz, Predictive Analytics, Video Games
We discuss the challenges of gaming analytics, most desired missing data, current trends, career advice, important soft skills in data science and more.
Pallas Horwitz is a Senior Data Scientist at Blue Shell Games, where she built out their cross-platform data infrastructure and oversees the statistical techniques used for analyzing features and optimizing revenue. Blue Shell Games is a casino gaming studio best known for Lucky Slots, one of the top grossing casino games of 2012 and 2013. Prior to Blue Shell Games, Pallas was on TinyCo's Data Team, where she worked on in-game economy optimization and monetization for Tiny Zoo, one of the top grossing mobile games of 2011.
First part of interview.
Here is second and last part of my interview with her:
Anmol Rajpurohit: Q5. What are the biggest challenges in deriving actionable insights from gaming data?
Pallas Horwitz: Whales. Our highest value customers are whales, but they skew the data and their spending behavior is erratic. You can exclude them from an analysis, but then you aren't accounting for one of the most profitable segments of your user base. If you analyze and cater only to whale preferences, you risk alienating the bulk of your user base and turning off players that may become whales next month. The other problem with whales is that statistically, you need a lot of them before the central limit theorem applies and you can use classical statistical techniques. I’m happy if I can get a p-value of 0.2. Data Scientists in most fields would find this preposterous.
AR: Q6. As a data scientist in gaming industry, is there any kind of data that you wish was available and accessible (besides what is available currently)? How do you bridge the gap due to non-availability of such data?
Since I started in the gaming industry, I've always wanted uninstall information.
If a user uninstalls your game, they are essentially terminating their relationship with the game. That’s very different than leaving an app installed, but untouched for several months. A user that still has the game installed just needs the right enticement for re-engagement. Trying to reactivate a user that has uninstalled the app is probably a lost cause. The best we can do, currently, is assume every user is waiting to be re-engaged. To some extent this is true, but our marketing tactics would be much more effective if we had uninstall information.
AR: Q7. Which of the current trends in Big Data arena are of great interest to you?
PH: I’m really interested in Machine Learning and predictive analytics. The gaming industry is very young. Retail, finance, and pharmaceuticals have been using predictive analytics for decades. The gaming industry is still trying to figure out how to appropriately apply these techniques. Our games evolve weekly. It can be hard to justify investing time in predictive solutions when the conclusions may be no more informative than business intuition. Furthermore, by the time a solution is found, the game may have evolved to a point that the original problem is moot.
Identifying which problems are best solved with predictive analytics is a growing pain that all young industries face and eventually solve.
I’m really excited to see where these techniques will lead and what types of questions I’ll be able to answer a year from now.
AR: Q8. What is the best advice you have got in your career?
PH: It’s a freaking video game! (The actual language was more colorful.) When I started working on my first game, it was six weeks away from losing its engineering resources due to under-performing revenue. In my mind the game was a dying patient, six weeks away from having the life support pulled and it was my job to save the patient. Lives were at stake! I’m proud to say that we did turn the revenue numbers around and the game kept a full-time engineering team for another year. Despite extending the life of the game, it’s important to maintain perspective and not take work too seriously. I've seen people burn out very quickly in this industry and I think many people would benefit from hearing it’s just a game, not heart surgery.
AR: Q9. What soft skills do you think are the most important for practitioners in the field of Data Science?
PH: A statistics, programming, or math background is very useful. However, none of those backgrounds teach you how to be both stubborn and creative. The data will always be imperfect.
You need to find creative proxies for the data you don’t have access to and the stubbornness, patience, and perseverance to make sense of these proxies. Sometimes it takes dozens of algorithms and data transformations before a meaningful insight begins to take shape.
When your data looks like a Jackson Pollock painting, you can’t be deterred. You also need a really thick skin. Nobody else really understands what data scientists work on. Even with a p-value < 0.05, game designers and executives will still disagree because the conclusion “feels” wrong. This doesn't mean that the analysis wasn't sound, it only means that some solutions have both a qualitative and quantitative element. The magic happens when you can find the solution that marries the two.
AR: Q10. On a personal note, we are curious to know what keeps you busy when you are away from work?
PH: On a weekday, I’ll probably watch a movie with my cat. On the weekends, I’ll take short trips to wine country or go to the theater. I like to keep things laid back and pamper myself a bit in my down time.