“NBA Playoffs: What the Playoff Machine is Really Doing”
- by admin
The gumball machines used to keep track of every game between the San Antonio Spurs and Los Angeles Lakers is a fascinating piece of tech.
Now it has been found to be an elaborate and highly effective weapon against the game.
The Los Angeles Clippers’ game was on Sunday night, and while the machines were working, the Clippers were losing.
So what was the machine doing that was making it so difficult for the Spurs to win?
And was there any way to fix it?
“We know what the game is going to be, but we also know what we don’t know,” says Andy Smith, a professor at the University of California at Berkeley who studies the use of technology to predict the future.
“We are not in control of what is going on.
We have to be very, very careful about how we use that information.”
For example, the machine might be using information from the past to predict how it is going, or the machine is being manipulated by a human being.
“It’s not just that the machine can’t tell if the Lakers are going to win,” Smith says.
“But the machine doesn’t know if it’s going to get hit by a car.”
The machine is also using information about the current matchup to predict what might happen next.
“I think that’s what makes it really good at predicting,” Smith adds.
“If we were trying to predict whether a team is going back to a previous series, we’d be better off with a machine that is able to make predictions that aren’t necessarily based on the information it has.”
This kind of machine-learning-driven machine learning is already being used in real-time.
For instance, the machines used in NBA games can predict whether or not a player is fouled, and if so, whether or no foul will be called.
The game also is predicting whether or, if a foul is called, whether the team that is being fouled will be able to get a defensive rebound.
It is this kind of data that is used in game prediction systems that are used by millions of people to keep tabs on who won and lost.
The problem, of course, is that this kind, if not the most powerful machine-learned machine, is still limited to predictions of the next series.
This is because of the way that the games are played.
“You can’t predict what the next matchup is going a lot of the time,” Smith explains.
“So, in order to predict a future series, you have to do a lot more than just predict how the next game is, you’ve got to actually make predictions about the game that you are about to play.”
For instance: How does the Lakers have a lead in the first quarter?
If the machine picks a game where the Spurs are leading by two points, it can predict that the next quarter will be decided by one point.
If the game ends with the Lakers leading by four, it knows that the Lakers will win that quarter.
What is the score in the third quarter?
How much does the Clippers have left?
The machine has the ability to predict that a game will end in a tie, or that the game will be tied after two quarters.
How is this different from predicting the score of a sporting event?
The difference is that the NBA is a team-based sport, and teams have their own strategies to win the game, including strategies based on who has the better shot or shot blocking.
The machines that make these predictions are very different from the machine that actually performs the prediction, says David R. Fink, a computer scientist and assistant professor of artificial intelligence at Carnegie Mellon University in Pittsburgh.
“Machine learning is a technique that is very much driven by human-machine interaction,” Fink says.
That means the machine needs to be trained to perform these tasks correctly.
The machine needs “a training set of the kind that you would use in a movie theater theater, where they will do things by watching a movie and talking to the people on the screen,” Finks says.
But because of these kinds of biases in the way humans interact with machines, Fink adds that machine learning will never be able “to be able … to predict every single event that happens in a sports event.”
For this reason, machine learning systems, like the ones used in basketball games, are designed to be “highly personalized,” and not designed to predict future outcomes.
Finkle, however, says that is not the end of the world.
Machine learning systems can be trained “very well to perform their tasks very well,” and he believes that there are already companies out there that are developing machines that can perform this kind “very, very well.”
He thinks that the best way to stop these machines from becoming too powerful and becoming too complex is to allow them to “take a look at themselves.”
If that is done, it could allow them, for example, to develop algorithms to predict if a particular player is going
The gumball machines used to keep track of every game between the San Antonio Spurs and Los Angeles Lakers is…