While I was testing our touring plan software in Walt Disney World last week, a computer named AlphaGo was beating one of the world’s best players of a game called Go. What made it remarkable is that Go is an enormously complicated game – there are something like 10^170 possible moves (that’s a 1 followed by 170 0’s). And that gave me the idea to express, in terms of games, how complicated it is to create an efficient theme park touring plan, and why we think computers can help.
The chart below shows some common games, an estimate of how many different ways there are to play that game, and what that complexity is similar to in terms of Magic Kingdom rides. For example, there are just under 32,000 ways to play the game tic-tac-toe. That’s in the same ballpark as the 40,320 ways to ride 8 attractions in the Magic Kingdom.
|Game||Ways to Play||Like a Touring Plan with|
|Connect 4||4.5 x 10^12||15-16 attractions|
|Checkers||5 x 10^20||21-22 attractions|
|Chess||10^40 to 10^50||35-42 attractions|
For comparison, our 1-Day Magic Kingdom Touring Plan for Adults has 26 steps, including lunch and dinner. Finding a good touring plan for that is “only” about 370,000 times more complicated than finding a good way to play checkers. I’ve helped write a fair number of books about getting the most out of your day at a theme park. But no book, no matter how good, can give you enough tips, tricks, and rules to sort through that kind of complexity. Put another way: Do I think I could beat a computer at checkers after reading one book on it? No. I might not make some common, rookie mistakes, but I won’t do well.
You’re probably thinking: Sure, there are a bajillion possible combinations of Magic Kingdom rides, but I can eliminate most of them pretty easily – don’t visit Country Bear Jamboree first, for example. But that’s misleading for a few reasons. For one thing, there plenty of scenarios – like touring plans that begin around noon – in which visiting Country Bear Jamboree would probably be a good idea.
Another is that people are generally bad at considering unorthodox strategies such as riding Country Bears first, even if they end up being good ideas. In the Go game referenced above, the computer’s 19th move was completely unexpected not only by the expert it was playing against, but by millions of Go players watching the match. It turned out to be a good move by the computer (it won the game) but widely acknowledged as a something no human would have done.
You might be tempted to say that all computer programs (ours included) have errors in their logic. Ask any parent of a teenager (ours included) whether people act logically all the time. The advantage to putting our touring plan software online (and free!) is that we’re able to get feedback if something looks amiss. We’re coming up on 11 million computer-optimized touring plans, and a person has looked at every one of them. Sometimes we get reports of things that are actual bugs; most of the time it’s that the computer has found an uncommon strategy.
A few other things add to the complexity of making a good touring plan: rides break down, the right FastPasses aren’t always available, and people run late. When a ride breaks down, it’s like a game referee telling you that you can’t make a move you thought you could make. Whatever strategy you had followed to that point, suddenly isn’t allowed. During my day at the Magic Kingdom last week, for example, 3 rides were temporarily closed when I attempted to ride them. A book can’t prepare for those kinds of events.
It’s the same with FastPasses – in our Unofficial Guide touring plan, we tell you to get morning FastPasses for Seven Dwarfs Mine Train, Big Thunder Mountain, and Peter Pan’s Flight. But those weren’t available by the time I made my plan, so the book’s strategy no longer made sense. And while I got to the park 45 minutes early, as the book suggested, that’s not always possible. People run late all the time, for lots of reason. In our game analogy, running late is like losing a turn. That calls for changing your strategy on the fly, in a few minutes, under less than ideal conditions. (I just clicked the “Optimize” button on the app.)
The last thing I’ll say is that having computers be better at games has actually improved how well people play games. In everything from poker to chess, the best human players improve their own games by playing against machines. The experts get to try hundreds of different, new strategies, while seeing what the computer does at the same time. It’s probably the same with touring plans – it doesn’t hurt to see what the computer would do in the same situation.