-
Not Synced
My colleagues and I are fascinated
by the science of moving dots.
-
Not Synced
So what are these dots?
-
Not Synced
Well, it's all of us.
-
Not Synced
And we're moving in our homes,
in our offices, as we shop and travel
-
Not Synced
throughout our cities
and around the world.
-
Not Synced
And wouldn't it be great if
we could understand all this movement?
-
Not Synced
If we could find patterns and meaning
and insight in it.
-
Not Synced
And luckily for us, we live in a time
where we're incredibly good
-
Not Synced
at capturing information about ourselves.
-
Not Synced
So whether it's through
sensor,s or videos, or apps,
-
Not Synced
we can track our movement
with incredibly fine detail.
-
Not Synced
So it turns out one of the places
where we have the best data
-
Not Synced
about movement is sports.
-
Not Synced
So whether it's basketball or baseball,
or football or the other football,
-
Not Synced
we're instrumenting our stadiums
and our players to track their movements
-
Not Synced
every fraction of a second.
-
Not Synced
So what we're doing is turning our
athletes into -- you probably guessed it
-
Not Synced
moving dots.
-
Not Synced
So we've got mountains of moving dots
and like most raw data,
-
Not Synced
it's hard to deal with
and not that interesting.
-
Not Synced
But there are things that -- for example
basketball coaches want to know.
-
Not Synced
And the problem is they can't know them
because they'd have to watch every second
-
Not Synced
of every game, remember it
and process it.
-
Not Synced
And a person can't do that...
but a machine can.
-
Not Synced
The problem is a machine can't see
the game with the eye of a coach.
-
Not Synced
At least they couldn't until now.
-
Not Synced
So what have we taught the machine to see?
-
Not Synced
So, we started simply.
We taught it things like passes,
-
Not Synced
shots and rebounds.
-
Not Synced
Things that most casual fans would know.
And then we moved on to things
-
Not Synced
slightly more complicated.
-
Not Synced
Events like post-ups,
and pick-and-rolls, and isolations.
-
Not Synced
And if you don't know them, that's okay.
Most casual players probably do.
-
Not Synced
Now, we've gotten to a point where today,
the machine understands complex events
-
Not Synced
like down screens and wide pins.
Basically things only professionals know.
-
Not Synced
So we have taught a machine to see
with the eyes of a coach.
-
Not Synced
So how have we been able to do this?
If I asked a coach to describe something
-
Not Synced
like a pick-and-roll, they would
give me a description and
-
Not Synced
if I encoded that as an algorithm,
it would be terrible.
-
Not Synced
The pick-and-roll happens to be the stance
in basketball between four players,
-
Not Synced
two on offense and two on defense.
-
Not Synced
And here's kind of how it goes.
-
Not Synced
So there's the guy on offense without
the ball and he goes next to the guy
-
Not Synced
guarding the guy with the ball,
and he kind of stays there
-
Not Synced
and they both move and stuff happens,
and ta-da, it's a pick-and-roll.
-
Not Synced
(Laughter)
-
Not Synced
So that is also an example of
a terrible algorithm.
-
Not Synced
So, if the player who's the interferer
-- he's called the screener,
-
Not Synced
you know, goes close by,
but he doesn't stop.
-
Not Synced
It's probably not a pick-and-roll.
-
Not Synced
Or if he does stop,
but he doesn't stop close enough,
-
Not Synced
it's probably not a pick-and-roll.
-
Not Synced
Or, if he does go close by
and he does stop but they do it
-
Not Synced
under the basket,
it's probably not a pick-and-roll.
-
Not Synced
Or I could be wrong.
They could all be pick-and-rolls.
-
Not Synced
It really depends on the exact timing,
the distances, the locations
-
Not Synced
and that 's what makes it hard.
-
Not Synced
So, luckily with machine learning
we can go beyond our own ability
-
Not Synced
to describe the things we know.
-
Not Synced
So how does this work?
Well, it's by example.