Machine Learning Times
Machine Learning Times

This excerpt is from Asugnews. To view the whole article click here.  

6 years ago
Sameer Chopra’s Analytics Lessons from the Blackjack Table


Hit on 16 and stay on 17—that’s the extent of blackjack strategy for many players. But even basic strategies of the game say there are times when hitting on 18 is the smartest play—such as when the dealer is showing a 10. Yet most people wouldn’t do that, says Orbitz Chief Analytics Officer Sameer Chopra—because they are afraid to bust.

“Players will go with their gut feeling, even when it is wrong,” adds Chopra, who spoke this week at Predictive Analytics World in Chicago.

But the only way to come out ahead at the blackjack table, or at least with the most money possible, is to stick with the right plan over a long series of hands—and to avoid making short-term decisions that undermine a long-term strategy, he continues.

This is similar to the nature of predictive analytics in business, where data-driven decisions are often meant to be made over a longer period of time. Predictive models are expected to fail periodically—just like in blackjack, where no one is going to win every hand.

To fight the resistance to deviate from a long-term analytics strategy, companies must establish a culture of testing in which they learn and experiment—in other words, to find those situations where hitting on 18 is actually beneficial. In a testing environment, this can be compared to making “gut” decisions, without having an impact on the business.

And that type of testing doesn’t have to begin with advanced custom algorithms. Easy wins can be found from greater application of common tests, such as A/B and multivariate tests, Chopra adds.

Expanding Your Skillset

Even if a blackjack player applies the best basic strategy, the house will eventually win over time thanks to the game’s built-in house advantage, or “vig.” That’s why Chopra says the only way someone can really come out on top in blackjack is by counting the cards that come into play as a deck is dealt—a difficult skill most don’t take the time to learn (and which is highly frowned upon by casinos, who like their vig very much, thank you.)

So how does the card-counting analogy apply to analytics professionals? After years of a SAS and SQL-dominated market, the field is finally undergoing significant change, Chopra says. With the growth of new languages for data science, such as R and Python, as well as big data technologies like Hadoop—all of which are can be cheaper to implement than SAS—new skills are required to move forward in a career, to come away a winner.

“The bar is changing—SAS and SQL won’t cut it,” Chopra opines.

Along with a solid set of technical skills, he prefers his employees to have soft skills—the ability to work with a team.

At the office, this means collaborating with others that can complement your own capabilities, and also being able to explain analytical insights to those outside on the business side.

The Benefits of Being Human

While Chopra advocates sticking to long-term analytics plans, he admits there are times when human intuition is important. The same goes for the blackjack world, where card counting can get a player blacklisted from casinos or worse.That’s why it’s valuable to strategically deviate from the plan.

For example, splitting a pair of 10s isn’t something a novice blackjack player would do—that hand adds up to 20 and is a nice-looking total. However, there are situations where splitting 10s is really the money-making choice. But with casinos watching the tables, making that sort of move can be a red flag to a watchful pit boss or “eye in the sky.”

So, to avoid suspicion, it is wise in that situation for a player to simply leave the hand alone.

“Human judgment is essential to decision-making,” Chopra says. “At times it is wiser to pursue less profitable paths.”

He points out that TV advertising has a history of poor return on investment, but says sometimes it is something that a company has to do just get its name out to a new crowd.

This ability to make nuanced decisions is important for Chopra. He cautions against the current onslaught of self-service data tools on the market for this reason. These must be coupled with someone who can put the analytics in context, who understands their genesis and possible limitations, before the results are applied to business decisions.

This excerpt is from Asugnews. To view the whole article click here.

By: Craig Powers
Originally published at

Leave a Reply