Posts By: Owen Martin

Finding the “Whales” of Your Game

Author icon Owen Martin|Comments icon 3

Let’s say you’re developing a game and you want to see who your most profitable players are. Imagine you’ve got a power-law so heavy-tailed you don’t even have a mean. You want to estimate your shape parameter and your statistics are all over the place as expected (no pun intended). Looks like you’re going to have to chop off that long tail and treat it separately. Looks like you’re going whale hunting.

In the casino industry, so-called “high-rollers” are often referred to as whales. It’s easy to see why: while they might come in with an attitude of being able to beat the house by wagering enough money, ultimately they wind up providing the biggest payouts to the casino. A somewhat disparaging term, to be sure, but it’s easy to understand how casinos view their revenue stream: lots of small change with the occasional big kill. Continue Reading…

Why the Obsession with Big Data?

Author icon Owen Martin|Comments icon 2

There are so many articles on big data these days, and it has me wondering, ‘Why the obsession?’ And, does the current data deluge actually require a reconsideration of standard statistical methodology?

Having just completed my PhD in statistics, I came to Kontagent to work as a data scientist in hopes of exploring one of the richest, biggest data sets assembled in recent history. Deciding to come do this stems from very fundamental questions I started asking myself toward the end of writing my dissertation. Specifically I want to know:

  • What is the purpose of theoretical research in statistics?
  • Is statistical theory the art of making inference from data or is it the science of decision-making under uncertainty?
  • And finally, if data processing, data science, and so-called knowledge extraction from data truly form the next big technology that companies will require to stay competitive, as many pundits claim, what besides computing improvements really needs to be researched and developed? Are the traditional statistical methods of Fisher, de Finetti, and Tukey really going to go obsolete? Continue Reading…