You have to try multiple different legitimate design approaches in order to understand anything about the qualities of your final design.
In some sense, it's a lot like data science. The little-leaguers design according to simple gradient descent: They'll pick one basic approach and do loads of A/B testing on it in order to make minor improvements.
The pros will do something like simulated annealing: Pick a number of different approaches, and see if one has a larger starting advantage.
In some sense, it's a lot like data science. The little-leaguers design according to simple gradient descent: They'll pick one basic approach and do loads of A/B testing on it in order to make minor improvements.
The pros will do something like simulated annealing: Pick a number of different approaches, and see if one has a larger starting advantage.