Message Detail
Board: ARCHIVE - SpC2k5 to SC2k6Topic: SC2k7 Oracle Challenge Discussion Topic
From: King Morgoth
Posted: 9/7/2007 10:42:31 AM
War is ON! >_>
Haste_2 | Posted 9/7/2007 5:35:38 AM | message detail
KM, remember on AIM how we were talking about the validity of your 4-way poll formula (and mine)? You know, how I was trying to convince you that if, say, Kefka gets 20% on Link, and Edgeworth and Bidoof each got 10% on Link, Kefka would get get 50% on Edgeworth and Bidoof? Anyway, the following stuff suggests it's all true.
Well, I skipped testing that idea with creative's formula, thinking it wouldn't work... until a minute ago. IT WORKED PERFECTLY (the other methods did not). So, maybe his was mathematically correct, after all (despite an apparently obvious flaw)... however, the results still don't make sense, so maybe multiple factors besides raw strength influence the results.
The way I tested them was this: use the amazing excel file King Morgoth provided, and then use only three characters. Then, I'd set Link to 50%, and set both Edgeworth and Bidoof to X%. I'd optimize their percentage until the "X-stats" results suggested 80% for Link, and 10% for the others. Then, I'd set Kefka to 20% (20% on Link), and keep the x-stats of Edgeworth and Bidoof the same. The result ended up 50%, 25%, 25%, exactly as they "should" be. I did a similar thing with other 4-way result calculators, and it wouldn't work out.
The whole problem lies in that you can't shake away the idea that if A, B and C get the same total as D against E, D will necessarily get exactly 50% against them. It's cute, it's instinctive, it's still false until you can prove it wrong. And saying that's what "should" happen isn't proving it =P
I've had a class once where we proved for a certain stochastic process T (I think it was a martingale, too) with stopping times, the expectancy of the time elapsed before the process stops is infinite. The teacher then went on to prove that, despite that, the probability that a finite time elapses before the process stops is ....1. I didn't like it either the first time. And despite being the one who just proved it, I still don't really like it. But I know it's right is an interesting thought coming from a teacher.... The first thing you should learn is that just because it doesn't make sense doesn't mean it's wrong...
---
Kaxon is a figurehead. Morgoth has all the real power. - Janus5000
SpC2k4 Triple Crown WINNER (Oracle, Betting, Spread Betting)
Haste_2 | Posted 9/7/2007 5:35:38 AM | message detail
KM, remember on AIM how we were talking about the validity of your 4-way poll formula (and mine)? You know, how I was trying to convince you that if, say, Kefka gets 20% on Link, and Edgeworth and Bidoof each got 10% on Link, Kefka would get get 50% on Edgeworth and Bidoof? Anyway, the following stuff suggests it's all true.
Well, I skipped testing that idea with creative's formula, thinking it wouldn't work... until a minute ago. IT WORKED PERFECTLY (the other methods did not). So, maybe his was mathematically correct, after all (despite an apparently obvious flaw)... however, the results still don't make sense, so maybe multiple factors besides raw strength influence the results.
The way I tested them was this: use the amazing excel file King Morgoth provided, and then use only three characters. Then, I'd set Link to 50%, and set both Edgeworth and Bidoof to X%. I'd optimize their percentage until the "X-stats" results suggested 80% for Link, and 10% for the others. Then, I'd set Kefka to 20% (20% on Link), and keep the x-stats of Edgeworth and Bidoof the same. The result ended up 50%, 25%, 25%, exactly as they "should" be. I did a similar thing with other 4-way result calculators, and it wouldn't work out.
The whole problem lies in that you can't shake away the idea that if A, B and C get the same total as D against E, D will necessarily get exactly 50% against them. It's cute, it's instinctive, it's still false until you can prove it wrong. And saying that's what "should" happen isn't proving it =P
I've had a class once where we proved for a certain stochastic process T (I think it was a martingale, too) with stopping times, the expectancy of the time elapsed before the process stops is infinite. The teacher then went on to prove that, despite that, the probability that a finite time elapses before the process stops is ....1. I didn't like it either the first time. And despite being the one who just proved it, I still don't really like it. But I know it's right is an interesting thought coming from a teacher.... The first thing you should learn is that just because it doesn't make sense doesn't mean it's wrong...
---
Kaxon is a figurehead. Morgoth has all the real power. - Janus5000
SpC2k4 Triple Crown WINNER (Oracle, Betting, Spread Betting)
Or, you may just return to the Message List without reporting this message.