If you were an NBA coach, how would you describe what a pick and roll is? Having trouble? Don’t worry, it’s not just you, it’s the NBA coaches too. It’s so difficult to describe, because there are many factors that go into determining if it is a pick and roll: timing, location, velocity, distance between players, and even the size of the players.
So, even if a coach could put into words what exactly makes a pick and roll, then how would we be able to take that description and turn it into an algorithm? That seems impossible.
But it’s not. We have created algorithms that can recognize even complex movements that only sports professionals would otherwise see.
It all started with the “Science of moving dots.” The algorithm began tracking the movement of players – represented as moving dots on a screen. Initially it just recognized the movement, but machine learning certainly is impressive; soon the algorithm could recognize a movement as something significant – such as a shot or rebound.
And soon after, the algorithm began recognizing even more complex movements, like down screens and pick and rolls.
Now machine learning has gone beyond our own ability to describe things. But this doesn’t take the human intuition out of things; we still need coaches. Because the coaches take the data, and the insight it provides and turn it into strategy.
But without the machine, the coaches likely would have never known that a pick and roll is actually a key to winning.
Watch the TED Talks video below to learn more: