News for hackers. Stuff that matters.
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @01:01AM (#29366579)
> you are expecting a profit of $901

No, it would be stupid to "expect" a profit of $901 because there is no scenario in which you would gain $901. You would either have -$100 or +$100000 and never +$901 in your stated scenario.*

In your case since there is a 99% probability of -$100, my "expectation" would be -$100 and if "surprise" of $100000 happens that is "surprise", not part of the "expectation".

* I realize there is a term called "expectation value" and I hate that term with a passion because in any distribution that doesn't roughly resemble a bell curve it would be stupid to "expect" the mean value. Statistics like the mean are not a good way to establish a strategy for a game that you only get to play once.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by josephcsible on Sunday November 28, 2021 @01:05AM (#29366593)
"Expected value" means the mean of any distribution, not just ones shaped like bell curves. GP is correct.
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @01:13AM (#29366619)
Exactly, my point is it's stupid terminology. I use the word "mean", not "expectation". That brainwashes people actually "expecting" the mean to happen, which in this case, it never will. Treating the mean as an expectation of some sort is just not a good formulation to strategize about this game.

GGP said "you are expecting a profit of $901" -- I'm not, personally. I'm "expecting" -$100. How about we take a bet? I'll expect -$100, you expect $901. If I win you give me $100, if you win I give you $100, and if neither of us wins we split the $100K. Deal?

This is a game theory problem, NOT a statistics problem. In my case the utility of losing -$100 is zero, the utility of gaining $100K is very positive, the utility of missing out on the $100K is extremely negative.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by bobsmooth on Sunday November 28, 2021 @01:46AM (#29366738)
>This is a game theory problem, NOT a statistics problem. In my case the utility of losing -$100 is zero, the utility of gaining $100K is very positive, the utility of missing out on the $100K is extremely negative.

Calculating utility, or the expected value, is entirely statistics.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @03:13AM (#29367116)
utility != "expected value" "expected value" != MY expectation utility != statistics on dollars
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by mynameisvlad on Sunday November 28, 2021 @04:37AM (#29367541)
Just because you think so strongly that your definition is correct does not make it fact.

"Expected value" and its derivatives is a specific term, with a specific calculation. You can't just throw around words and expect others to understand you don't mean the actual definition when you say them.

https://www.e-education.psu.edu/eme460/node/730

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @11:46AM (#29370377)
I am not arguing over a definition, I'm just saying that the definition sucks and that GGGGGP was actually expecting the "expected value" which I wouldn't do in this case.
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by josephcsible on Sunday November 28, 2021 @01:32AM (#29366693)
Are you saying we should stop using the term "expected value" in cases where it isn't equal to the mode?
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @03:13AM (#29367117)
Yes.

I don't know about you but if you have a coin labeled 0 and 1 and flip it it would be mind-bogglingly stupid to expect a 0.5 out of this coin. Just like if you have a coin labeled A and C it would be stupid to expect a B to come out of it. You'd be better off picking either A or C and expecting that instead.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by Daishiman on Sunday November 28, 2021 @05:07AM (#29367703)
I guess you never took stat 101.
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Sunday November 28, 2021 @11:48AM (#29370397)
Personal attacks not appreciated.

No statistics class I took said that a coin with sides A and C can give you a B, or that you could "expect" 0.5 from a bimodal distribution with a delta function at 0 and a delta function at 1. The mean value may be 0.5 but it's not something you would expect to actually happen; if you wanted to maximize the probability that your prediction is correct you'd be better off picking 0 or 1.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by Daishiman on Sunday November 28, 2021 @01:30PM (#29371450)
You're making an entire thread about defending your position against using a term of art the way it's used by professional and academics all over the world. Judging by every reply you're getting, everyone know what they mean when they talk about the expected value of a probability distribution. Just drop it.
Re:Crypto enthusiasts want to buy an NBA team, after not pur... by kortilla on Sunday November 28, 2021 @04:32PM (#29373249)
It’s not a personal attack. “Expected value” is a very basic term of art introduced in stats 101.

It’s like having a huge argument about a physics problem claiming “force” isn’t mass times acceleration because nobody is “forcing” you to do anything.

Back to your 0 or 1 example. 0.5 is the meaningful value to know over any kind of large sampling because that is what matters for “expected value”. It tells me what I should expect overall if I play 1000 times.

What you want is called the “mode”. Given a single instance of an event, what is the most frequent value.

Re:Crypto enthusiasts want to buy an NBA team, after not pur... by dheera on Monday November 29, 2021 @01:36PM (#29383394)
> “Expected value” is a very basic term of art introduced in stats 101.

I already know that.

My point is it's shitty terminology, even though it's somehow become a standard.

Stats 101 also does NOT tell you to "expect" the expected value to happen.

GGGGGP however seems to actually use the value to formulate a strategy, and that's where I'm pointing out that this entire formulation of using stats to approach this problem is flawed. You don't have enough chances at the game such that the mean is a meaningful number to consider.

That's also why I always refer to it as the mean even if other people call it "expected value". I hate the term with a passion and I will not yield to the wise guys who wrote the books.

Honestly I'd rather not have this discussion in English because this isn't a problem in many other languages. In many languages mean is not called "expected" anything.