Webinar June 18: Deep Learning Approaches to Forecasting and Planning Register

What is a Monte Carlo Simulation? (Part 3)

By Zachariah Miller • February 12, 2018

This is part 3 of a series by Metis Sr. Data Scientist Zach Miller dedicated to investigating how Monte Carlo can be a great tool. Part 1 introduces the concept of Monte Carlo, and in part 2, he goes over how to implement it using Python.
_____

Can you beat the casino in video poker?

Spoiler alert: No, you can't. If you could, the games wouldn't be in the casino. However, meta logic aside, how definite is it that the casino will always beat you? We're going to use Monte Carlo to explore that concept. The methodology will go a little bit like this: build out a Python representation of a card deck, then use MC methodology to determine the best possible play you can make given your hand, then make that play and track your money over time.

The MC part will come into play when we select which cards to hold in the game. (If you aren't familiar with video poker, try checking this out.) We'll see what cards we have, then simulate the results of choosing to hold the cards and see what method gives us the best possible payout. Since this is a lot of work, I've put together a fairly long notebook that shows you all the code it takes to build a card deck, set up the poker game (we have to do the hand scoring), and then actually run the Monte Carlo simulations to see what happens. Throughout the Video Poker Monte Carlo Notebook there are lots of comments and text blurbs to explain the methodology.

So what did we learn?

As you can see from the notebook, even if you play poker perfectly by always making the move that gives you the best-expected return, the casino pretty much always comes out on top if you play long enough. We built out some pretty nifty machinery to be able to let Python play cards with itself over and over to figure what that best move is and it's still all for naught. However, along the way we've learned about using combinations of cards, multiple randomized simulations, and expected returns to make better choices. We also see that by playing smartly, we can WAY outperform playing randomly. In our next installment, we'll investigate how we can use these same principles to figure out how particles interact with matter by building our own particle simulation package in Python (coming soon)!

_____

If you haven't already, read Part 1 and part 2 of Zach's series.


Similar Posts

data science
Sr. Data Scientist Roundup: Quantitative Drawbacks of Word Clouds; How to Solve Math Puzzles, & More

By Emily Wilson • January 24, 2020

When our Sr. Data Scientists aren't teaching the intensive 12-week bootcamps or corporate training courses, they're working on a variety of other projects. This monthly blog series tracks and discusses some of their recent activities and accomplishments.

data science
Data Scientist Roundup: Adopting Python in the Workplace, NLP with Straightforward Tools, & More

By Emily Wilson • April 30, 2020

When our Data Scientists aren't teaching the intensive 12-week bootcamps or corporate training courses, they're working on a variety of other projects. This monthly blog series tracks and discusses some of their recent activities and accomplishments. Check out what's been happening in April.

data science
Misleading Graphs: Manipulating the Y-Axis

By Roberto Reif • April 06, 2020

One of the most commonly used charts for data visualization is the bar chart. But too often, the starting value of the y-axis is intentionally modified to skew our interpretation of the chart and the data. In this post, see examples and learn how to readily identify this issue.