RPI Forecast for remaining regular season games

[AlanInNJ]

It can certainly add up; it's all about probabilities.

Let's say I have four games left and my chances of winning each is 60%. I'm the favorite in every game - yay! Am I then expected to go 4-0 over those four games? Hell no. The chances of me going 4-0 are actually tiny - .6 x .6 x .6 x .6 = .13 or 13% I have a 13% chance of going 4-0 over those four games in which I'm the favorite. So, the computers, looking at the most likely scenario for those four games, will likely say 3-1. I'm favored to win each of those games individually, I'm a serious underdog to win all four of those games in a row.

Now, if I was a 90% fav in each of those four games, then the computers would predict I'd be 4-0 at the end, as .9 x .9 x .9 x .9 = .66 or 66%. I'd have a 66% chance of going 4-0.

IMHO that's some flawed math. The events are mutually exclusive. They are independent events.

Either way, let's just win out and put all this to rest!

Alan

 

[DukeJim99]

IMHO that's some flawed math. The events are mutually exclusive. They are independent events.

Either way, let's just win out and put all this to rest!

Alan

I think you meant to say "The events aren't mutually exclusive," meaning, say, if Duke wins the first three games then they're almost certainly playing well and thus have a better than 60% chance to win the last of the four games. And I agree. The math I stated relied on the events being mutually exclusive, and they're not.

I considered adding that caveat. But seriously, I'd already typed up 500 words about math on a basketball forum. I thought that was plenty!!

Either way, though, even acknowledging the events aren't mutually exclusive, the point I made remains true. Just because you're favored to win in each of four games most certainly does not mean you're favored to go 4-0 over those four games. And that was GAAP's original question.



** Just in case anyone is curious about the issue Alan raised, the whole "mutually exclusive events" thing is what led to fivethirtyeight.com having a much higher prediction of Trump's chances of winning the election that other prediction sites. 538's predictive model recognized that many swing states, such as PA, OH, MI, and WI have similar demographics. And thus a small error in polling related to a key demographic could mean that all four of those states would move in the same direction. Whereas other predictive sites assumed that each state was independent of others and applied math accordingly, giving Trump only a single digit chance of winning.

 

[AlanInNJ]

Semantics DJ99

Math shmath, win out and it's a moot point!

Go Duke.

Alan

 

[Mac9192]

I think you meant to say "The events aren't mutually exclusive," meaning, say, if Duke wins the first three games then they're almost certainly playing well and thus have a better than 60% chance to win the last of the four games. And I agree. The math I stated relied on the events being mutually exclusive, and they're not.

I considered adding that caveat. But seriously, I'd already typed up 500 words about math on a basketball forum. I thought that was plenty!!

Either way, though, even acknowledging the events aren't mutually exclusive, the point I made remains true. Just because you're favored to win in each of four games most certainly does not mean you're favored to go 4-0 over those four games. And that was GAAP's original question.



** Just in case anyone is curious about the issue Alan raised, the whole "mutually exclusive events" thing is what led to fivethirtyeight.com having a much higher prediction of Trump's chances of winning the election that other prediction sites. 538's predictive model recognized that many swing states, such as PA, OH, MI, and WI have similar demographics. And thus a small error in polling related to a key demographic could mean that all four of those states would move in the same direction. Whereas other predictive sites assumed that each state was independent of others and applied math accordingly, giving Trump only a single digit chance of winning.

Who are you, Karl Rove?

 

[dukephysics]

I think you meant to say "The events aren't mutually exclusive," meaning, say, if Duke wins the first three games then they're almost certainly playing well and thus have a better than 60% chance to win the last of the four games. And I agree. The math I stated relied on the events being mutually exclusive, and they're not.
....

But for the calculation they are considered mutually exclusive. Your math was correct. When the calculation is done, there is no correction (at least, I would find it hard to believe there was one) to the 60% chance to win number based on the idea that we were also favored in the previous game, so now 60%->62% or something.

 

[AlanInNJ]

I wonder if they have these sorts of discussions over on the UNC boards??

 

[dukephysics]

I wonder if they have these sorts of discussions over on the UNC boards??

They do. It's over there that I learned that 40% of 3 is the same as 50% of 2. That kind of education is priceless.

 

[Quavarius]

I guess the computers were right.

 

[QC Dukie]

I wonder if they have these sorts of discussions over on the UNC boards??

Yes they do because the ceiling is da roof.