Welcome to the Forum Archive!

Years of conversation fill a ton of digital pages, and we've kept all of it accessible to browse or copy over. Whether you're looking for reveal articles for older champions, or the first time that Rammus rolled into an "OK" thread, or anything in between, you can find it here. When you're finished, check out the boards to join in the latest League of Legends discussions.

GO TO BOARDS


How many times do you use Strategy A?

0-10 1 2.17%
11-20 4 8.7%
21-30 1 2.17%
31-40 4 8.7%
41-50 4 8.7%
51-60 2 4.35%
61-70 2 4.35%
71-80 5 10.87%
81-90 6 13.04%
91-100 17 36.96%
Voters 46 .

Quiz time!

123
Comment below rating threshold, click here to show it.

KirbyHero

Recruiter

12-05-2010

Quote:
yourthevoiceyo:
Hello, you're in
League of Legends Community (http://www.leagueoflegends.com/board/index.php) > League of Legends (http://www.leagueoflegends.com/board/forumdisplay.php?f=1) > General Discussion (http://www.leagueoflegends.com/board/forumdisplay.php?f=2)

I believe you are seeking the off topic discussion board.

Reds have responded to GD threads that had absolutely nothing to do with LoL and not locked them. The feel is pretty lax. And I'm going to explain what this has to do with LoL after I get a couple more results.


Comment below rating threshold, click here to show it.

Polaritie

Recruiter

12-05-2010

The expected value of the result is .5A+.35B, with A+B=100.

Clearly, the maximum EV is 50, with 100 A.

Additionally, this method gives the greatest number of acceptable failures, with .6% tails rate being a success.

Hmm... percent of tails allowable... 40-.35B=points needed from A. A=100-B.

Allowable percent of tails=1-percent heads needed.
Percent heads needed=Needed heads (points from A)/A.

Therefore, the equation for maximum allowable percentage of tails is:
1-(40-.35B)/(100-B). Maximizing this function gives the ideal choice for B.

And of course, the maximum value is achieved for B=0, which proves my initial choice to maximize expected value.

(This is what happens when you give a math major a math problem. Expect a thorough analysis and/or formal proof!)


Comment below rating threshold, click here to show it.

Dieby

Senior Member

12-05-2010

Quote:
Polaritie:
(This is what happens when you give a math major a math problem.)
I have a problem I need solving.

What is the proability that flipping a coin 100 times will net at least 40 heads?


Comment below rating threshold, click here to show it.

Sergiu

Recruiter

12-05-2010

what happens if i change my mind after i vote


Comment below rating threshold, click here to show it.

KirbyHero

Recruiter

12-05-2010

It's about when I felt explaining the thread. Polaritie's proof is correct.
A formal proof isn't necessary if you understand what large numbers of trials do to probability. The more trials you have, the more likely it is that you receive a value very close to the expected return.

Say you used option B 20 times. Since Option B is actually less than the average you need to get, the remaining 80 coin flips have to reach a HIGHER average value to meet your minimum.

I asked the poll question in a very biased way. The reason I did so is because people look at elo in a very biased way, they all think they're amazing players. When confronted with something as abstract as 100 coinflips or 100 matchmaking games with 9 other players who leave on both teams, the bias can take over.

A lot of criticisms to the elo system come from the sort of people that would miss this question. There's nothing wrong with criticizing the elo system if you understand statistics and math pretty well, but using anecdotal evidence and bias about your own elo clogs the forums.


Interestingly, option B is VERY analogous to "stat tracking", the suggestion that players repeatedly make asking that stats like kills, assists, and tower destroys be used to determine someone's true elo. Stat tracking is more stable for a sample size of one game, just like option B, but since it's overall inaccurate on anything but carries (and probably still not 100% accurate for carries), it's missing the ideal amount the way option B does. Riot says they use it only for the first few games, very small sample sizes, where it makes sense.

Instead of stat tracking, it's better to allow a large number of random matchmaking games to occur and eventually arrive at an accurate elo.


Comment below rating threshold, click here to show it.

Seiuchi

Senior Member

12-05-2010

Quote:
KirbyHero:
It's about when I felt explaining the thread. Polaritie's proof is correct.
A formal proof isn't necessary if you understand what large numbers of trials do to probability. The more trials you have, the more likely it is that you receive a value very close to the expected return.

Say you used option B 20 times. Since Option B is actually less than the average you need to get, the remaining 80 coin flips have to reach a HIGHER average value to meet your minimum.

I asked the poll question in a very biased way. The reason I did so is because people look at elo in a very biased way, they all think they're amazing players. When confronted with something as abstract as 100 coinflips or 100 matchmaking games with 9 other players who leave on both teams, the bias can take over.

A lot of criticisms to the elo system come from the sort of people that would miss this question. There's nothing wrong with criticizing the elo system if you understand statistics and math pretty well, but using anecdotal evidence and bias about your own elo clogs the forums.


Interestingly, option B is VERY analogous to "stat tracking", the suggestion that players repeatedly make asking that stats like kills, assists, and tower destroys be used to determine someone's true elo. Stat tracking is more stable for a sample size of one game, just like option B, but since it's overall inaccurate on anything but carries (and probably still not 100% accurate for carries), it's missing the ideal amount the way option B does. Riot says they use it only for the first few games, very small sample sizes, where it makes sense.

Instead of stat tracking, it's better to allow a large number of random matchmaking games to occur and eventually arrive at an accurate elo.


Before this post, I thought the connection between this and Elo would be a stretch. Now I'm sure that there's no connection at all. You didn't even give the answer. Multiplying some amount by 0.35 is analogous to stat tracking? What? Coin flips represent playing LoL? Again, what? Do you even know how the Elo system works? This isn't analogous to it at all! I'M SO CONFUSED! Most importantly, give me the answer to the question, you jerk. Imagine if cereal boxes had puzzles on them but no answers on the bottom of the box: I'd never eat cereal again. Everyone knows that the most important part of any question is flipping to the answer page to see if you got it right, so give us the answer already.

EDIT: nvm, reread the OP and realized you're going for 40 minimum, not 40 exactly. Still makes no sense what this has to do with matchmaking.


Comment below rating threshold, click here to show it.

KirbyHero

Recruiter

12-05-2010

Quote:
KirbyHero:
It's about when I felt explaining the thread. Polaritie's proof is correct.

?


Comment below rating threshold, click here to show it.

Caprice

Member

12-05-2010

You still have a bias anyway... You're only going to get answers from people who read the forums, who read your post, who are willing to answer (even if they know the answer).


Comment below rating threshold, click here to show it.

Cryaotic

Senior Recruiter

12-05-2010

Quote:
skillz i haz the:
mind=go boom


^


Comment below rating threshold, click here to show it.

KirbyHero

Recruiter

12-05-2010

Quote:
Caprice:
You still have a bias anyway... You're only going to get answers from people who read the forums, who read your post, who are willing to answer (even if they know the answer).

Agreed, although the first part isn't an issue since I'm mainly concerned with the forumgoing community, the ones that complain that their elo is too low.


123