"Chuting" for Speed: A Dad's Quest for Answers

Posted by Raif Majeed on November 26, 2007

Filed under: games QA

Chutes and Ladders is a popular game in our family; it's great for evening "together time" without the need for deep strategizing. If you played the game as a kid, you probably enjoyed the back-and-forth of advancing up the board and then sliding back down. And if you ever played it as an adult, you've probably wondered when the game is going to end! I've got some answers and advice.

After about an hour and a half of one particularly tedious game, I realized that I could write a program to simulate Chutes and Ladders and answer the key question for parents: How long does a typical game take?

I figured I would write my own quick-and-dirty simulation, save the results of every move by every player in every game, and feed those results into Tableau to get the answers I wanted. Those of you who have used the Tableau Software product before won't be surprised to hear that I got much more information than I initially expected.

(By the way, I'll defer discussion of the technical details of Chutes and Ladders -- and my simulations of it -- to a second post.)

First, a quick recap of the game rules: The board has 100 squares; the first player to advance to square #100 is the winner. Normally players move 1-6 squares at a time, based on the result of spinning a spinner; but landing on certain squares forces the player to move upward (via a "ladder") or downward (via a "chute"). One other twist: if your "spin" takes you past the end of the board (eg you're on square #99 and you spin a 3), you have to stay where you are.

So after playing around with my game-simulation data in Tableau, here's what I found. First off:

• If you're budgeting time for a Chutes and Ladders game, plan on about 30-40 moves per player for a 3-player game. So if there's less than an hour until bedtime, consider bringing out a different game.

To see why, take a look at the statistics that Tableau came up with, using data from 32,000 3-player game simulations. To get this result, I filtered out all moves except the ones that had endpoint 100 (the winning square), then took statistics on my database records.

The average game is about 20 moves -- much shorter than I expected. But the large standard deviation (~ 10 moves) shows that it's really difficult to talk about a "typical" game because they vary so widely. As a parent, it’s best to plan for the longer games and play it safe. About 95% of 3-player games will finish within 40 moves (mean + 2 SD’s).

To see more details on this, take a look at the histogram of game length -- that is, the number of games that took (n) turns, plotted against (n) -- which Tableau generated for me:

As you can see, there are a lot of relatively short games, but there's a nontrivial "tail" of rather long games. So to reinforce the point above:

• Once you get past 40 moves in a game, don’t assume you’re "almost" done. There is a substantial probability that you’re not. Instead, just call it a tie and pack it up.

So now we have an idea of how long it takes to win. That brings up another question: Who usually wins? Is there an advantage to going first? I’ll put it this way:

• The official Chutes and Ladders rules from Hasbro specify that the youngest player goes first -- make sure you follow that rule.

Why? Because there is a slight advantage to the first mover. Here's a plot of winning percentage for each player in a 3-player game:

Now that I had the information I wanted about game lengths and winners, I was curious about how the game (on average) evolves over time. Using the Pages shelf in Tableau is ideal for visualizing time-dependent processes like this. I used it to make a movie of the distribution of player position.

Here, each bar represents a square, and the height of the bar represents how likely it is for a player to be on that square. (Disclosure notice: I’ve used a moving average and filter to smooth the animation easier to interpret; it makes things a little smoother, but doesn’t change the conclusion.)

As you can see, this distribution actually reaches a sort of steady state. Even though players are moving back and forth across the board, the likelihood of a square being occupied (given that the game hasn’t ended yet) is fairly constant. Why does this happen? It’s an interesting linear-algebra question which I’ll leave for my next post.

My final advice for parents:

• If your kids are insisting on Chutes and Ladders and it’s close to bedtime, show them the movie above. They may forget about playing a real game.

Like any interesting analysis, this brings up at least as many questions as it answers. What happens if you modify the board? What if you modify the rules? What if you add occasional random moves (representing accidental piece knockdowns or toddler intervention) – do positive and negative random events cancel out?

I'll talk about all that in my next post, as well as how this Chutes and Ladders simulation turned out to be related to software testing, conditional probabilities, Markov Chains, search-engine rankings, leaky water buckets, and other such things.

And as for the children in our family? Well, they're not sure why I did all this work. They enjoyed watching the Pages shelf animation of Chutes and Ladders and asked me to play it many times. But now they're into Connect Four -- a much harder simulation!

If you'd like to play around with a smaller version of this data set, you can download the packaged workbook. You'll need Tableau Desktop 3.5 or higher, or the free Tableau Reader. Your views will look a little different and your statistics won’t be quite so clean – with the full data set the workbook would have been a 2 GB download.

Why is community so important? Our customers aren’t just using Tableau Software to more quickly and easily access their data to generate their existing reports or perform existing analysis processes. They are also discovering that they can question their data in ways that weren’t possible before. Users that used to only view static reports can now have intuitive Q&A sessions with their data. Fast visual analytics for everyone.

With these new abilities come questions: What are the best practices of visual analytics? What types of questions can I ask of warranty claim data or investment data or network utilization data or survey results? How can I most effectively approach a Q&A session with my data? How do I best equip my user community for visual analytics? How do I get help completing my tasks in Tableau? We hope our web site will help users answers many of these questions. The forums and sample gallery will provide an opportunity to connect with other users and to provide access to examples of how people are using Tableau in many verticals and for many tasks.

One of the things I am really excited about is that our new packaged workbooks and Tableau Reader have given us the technology to not only provide screenshots and write ups of examples but to also allow you to click and download that workbook and data to your desktop and interact with it yourself or repurpose it to work on your own data.

Many of our users are also passionate evangelists within their organization for visual analysis and for our products. And we are extremely grateful for that. But they want to know more about our mission, our values, our product roadmap, and about the science of visual analysis. One way we are working to support that is through our new corporate blog and “Tableau Letters”. In our blog, we will be providing frequent examples of how we are using the product internally and we will be talking about exciting ways our customers are using the product. Tableau Letters are longer articles that our team will be authoring on the science of visual analysis and our unique way of connecting you to your data. Our CEO, Christian, has authored the first letter. In it, he debunks the two major myths of visual analysis and shows why visual analysis is so important regardless if you have 200 rows of data in Excel or 2 TB of data in Netezza. It’s a great introduction to visual analysis and our unique perspective.

I am really excited about our growing community and enabling deeper connections between our users and ourselves.

Cheers
Chris

We had that happen yesterday. Stephen Few posted a very positive review of our latest releases, Tableau Desktop 3.5, Tableau Server, and Tableau Reader in his blog, Visual Business Intelligence.

The thing I’m most proud of is Stephen’s recognition of Tableau Software’s work to keep you the customer at the center of everything we do.

Stephen said “The team at Tableau succeeds better than any other software design and development team I know in identifying the most important next steps in the product’s evolution and proceeding through those steps expertly and thoughtfully.”

We make a huge effort in this area and it’s great to see one of the visualization industry’s most respected thought-leaders recognize it. So thank you, Stephen. I promise you and our customers - we’ll keep doing what we’re doing and improving.

As a firm that lives and dies by helping more people and companies have "analytics everywhere", Tableau Software can appreciate those hurdles. But using on-demand platforms like Salesforce with analytics that anyone can use truly does mean organizations can be stunningly smarter and far more effective. Tableau itself is a case-in-point. I’m still pretty new to the company so I’ve seen it first-hand with a bit of an outsider’s perspective.

Peter had a chance to meet some of Tableau's internal team whose jobs are virtually defined by Salesforce.com. What he may not have seen is how deeply many more people at Tableau live and die by Salesforce.com with Tableau analytics laid on top. Tableau eats its own dog food.

The marketing team and I don’t have to make marketing decisions without knowing how our pipeline is developing and how our marketing decisions affect it. Because of Tableau's "analytics everywhere" software on top of Salesforce.com, day 2 on the job here I was already seeing and understanding how Tableau brings in new leads and nurtures them. I can rapidly look at detailed Salesforce.com data the way I want to – no predefined reports, no manipulating in Excel. And the amazing thing is that every executive at Tableau accesses the same Salesforce.com data using Tableau. So we always have meaningful conversations based on the same set of facts. No way have I had that kind of visibility or level of discussion in any previous job – and it wasn’t from lack of trying or wishing.

So, Peter, amen to your post. It’s way past time to make more of the capabilities of on-demand platforms by layering in analytics everywhere.

• Desktop users can publish dashboards and workbooks securely in real-time with just a few clicks
• Server users can search, filter, sort and view data with just a browser
• Everyone can post comments, tag views and participate

We released v3.5 of Tableau Desktop. New features include improved visual styles and even faster performance.

We have made available a FREE application called Tableau Reader. We like to think of it as “Acrobat for Data” and its release is based in part on the influence of our newer Board Member, Adobe Chairman and Founder Chuck Geschke. Once installed, this application allows anyone to immediately start viewing and interacting with Tableau packaged workbooks. Check out our Visual Examples for download information and several packaged workbooks.

We revamped our brand identity, with a new logo and look-and-feel. This new identity better expresses Tableau’s attributes: authentic, personal (not corporate), easy-to-use, clean, smart, and personal. Response so far has been enthusiastic. Your thoughts? I'd love to hear them.

In keeping with the new brand, we upgraded our website. You’ll find deeper content and more surfacing of the content that users care about. This includes a new Learning Center and v1 Community Center. Again, feedback is welcomed!

Please let me know if you have any questions or comments. As always, we welcome your feedback.