February 22nd, 2023
I have a routine in the JavaScript code that analyzes the overall behavior of the system using “A Million Monkeys”. This refers to the old saw that “A million monkeys with a million typewriters in a million years will produce Shakespeare’s plays.”
My routine substitutes random behavior (the monkeys) for player decisions. It runs through the entire storyworld just like any player, except that, where the player makes a decision and choses an option, the monkey randomly picks an option. It does this a hundred times, not a million times. It then compiles the ending values for each of the monkeys and calculates the mean and standard deviation for each of the critical variables used to evaluate the players performance.
My thinking is that a real player should be able to do better than, say, 95% of the monkeys, so I set the criterion for success in any given dimenstion at the mean plus two standard deviations. Therefore, in order to succeed in dimension A, the player must score better than 1.38. But it turns out that few players can exceed the required values. Now, not many players get all the way to the end of the story; over the last 14 days, only four players made it that far. Here are all the results for the monkeys and the players:
As you can see, the players did better than the monkey in three out of five variables, and worse in two. Does this mean that monkeys are smarter than people? I rather doubt it, although there are certainly subsets of the population for which this is true. No, what this means is that the choices I present to the player are either misleading or misworded. I’ll have to go through each and every encounter, rebalancing the weightings on the options. This will entail a lot of work.
I must also relax my standards for success. If I lowered my criterion for success to just one standard deviation above monkey results, the four players would have met the criterion on three of the five variables and just barely failed to qualify on the other two. That’s probably too generous. But at least I know that the correct criterion value is greater than one standard deviation and less than two.