Breaking the Computer Series: Why bother?

(Note: This is the first installment on my series Breaking the Computer.  For other articles in this series, please click here.)

Man vs. Machine. It’s become cliché at this point. Few doubt that the computer has been one of the most significant inventions in human history. Yet still, many resist to use the computer: they don’t believe that computers can help them play better Scrabble.

Skeptics have their reasons. Some reasons are pure bias: a disdain for thought that is more advanced than their own. “Computers are robotic, but we are not robots.” I heard someone say. “I’ll never be able to think like a computer, so why does it matter?” Such logic is misguided and fatalistic as best, and represents a lack of willingness to adapt more than an actual argument.

But these skeptics sometimes do have a point. “Computers can’t think,” they argue. “Computers are slaves to their basic algorithms”. This is true: computers do have their limitations. Computers are great at manipulating data: taking input data, applying formulas and spitting back an answer. But they are not good at interpreting data: they need humans to think for them.

For man, that data manipulation is the hard part. How good is that leave? How much does this play reduce bingo percentage? These are the questions at which computers excel. Computers can also use objective logic: they are not subject to the various biases and tendencies that plague humans. While man cannot calculate better than the computer, it can interpret computer data.

When we evaluate and interpret the data provided by computers, we are looking for two things: a computer that computes a large amount of data, and for the data it computes to be relevant for us to figure out which play is best. A simple computation is not what we are looking for: we want computers to “show its work” so we can understand the numbers that led to its solution.

The program that does this best is called Quackle. Quackle was created by two extremely strong Scrabble players: Jason Katz-Brown and John O’Laughlin. Quackle uses Monte Carlo simulation based primarily on points and leave to make decisions, and allows you to input any position and then simulate, providing an analysis of which play is best as well as key statistics, such as bingo percentage, average score, and standard deviation.

The primary mistake that is made with programs such as Quackle is that people ask for and get a single solution (they’ll simulate and wait for a play to “win the simulation”) and trust that solution blindly and ignore the data: the very subject in which computers excel.

In the end, our goals when we use the computer are twofold. One goal is to figure out the best play, and using the knowledge of “best play” to develop and refine our intuition and heuristics, improving our decision making process. Using computer data to ultimately deduce the best play will help us ultimately become better players, since it will add to our intuition and heuristics.

Second, our goal should be to develop our intuition so we can approximate the data themselves. In the middle of a game, we don’t have access to computers: therefore, we can’t estimate bingo percentage, entropy, etc. But if we look at enough positions and we use the computer to estimate that data, we can develop our intuition so that our estimate is “close enough”. Although we’ll never be as good as computers at data analysis, computers can still improve our intuition in this important area.

This series will use Quackle, but the principles will be the same on any program: it will show you how to use all of that data, and focus on which data to use, which data to ignore, and most importantly, why. Understanding how to use computers, its strengths, its pitfalls, and how it can be used to understand and improve your Scrabble will prove invaluable as you improve as a player.