Free Allegiance Forums

Just to re-iterate; I don't endorse that system, I just think that you could built a self-policing 'reputation' system, if you wanted to. I don't think it would be as good as an automated one, or any faster to build, but it could probably work well enough, and might give people the warm fuzzies about their involvement.

But I want to see AllegSkill have its shot, the theory is sound. I'll wait for the presentation of the data.

It's easy, just use the FAO member rating system to determine your ranks. Easy as that. /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />

Speaking completely hypothetically because I'm not sure the popularity contest method is the best...

I would have to go with spidey's method here. A handful of respected individuals doing the ranking rather than everyone is better in my opinion. The only reason I say this is because I may not feel like looking at how everyone ranked everyone which would be very time intensive. Any of these systems would be time intensive for those involved. Even if you only do the ranks quarterly it would take an assload of time, not to mention many Americans wouldn't even know where to begin to rank people who play on off-peak US hours or vice-versa.

factoid wrote:QUOTE (factoid @ Dec 17 2007, 04:12 PM) Just to re-iterate; I don't endorse that system, I just think that you could built a self-policing 'reputation' system, if you wanted to. I don't think it would be as good as an automated one, or any faster to build, but it could probably work well enough, and might give people the warm fuzzies about their involvement.

But I want to see AllegSkill have its shot, the theory is sound. I'll wait for the presentation of the data.

Well let me make it clear I am NOT proposing it as an actual system (although it could work but I am not sure how well) just throwing out an idea. However if it were to actually happen it would have to be a select group otherwise it turns into a popularity contest. As a further example even people who HATE BoG will admit he is a semi decent player. Its all about being objective

Think of all the little voobs who probably think of McW as a lovely whitewashed figure...they would make him one of the top 20 players. Now those of us in the know will remember that McW should ALWAYS be #1 no ifs ands or buts due to the actions of Mrs. Meridth McWarren!

I think Thalgor should just rank us all and that should be binding for eternity /cool.gif" style="vertical-align:middle" emoid="" border="0" alt="cool.gif" />

The problem with this sort of peer review would be newbies. You're never going to accurately rate and keep on top of changing newbie ranks as they learn and get better.

spideycw wrote:QUOTE (spideycw @ Dec 17 2007, 08:18 PM) With such a system a truly accurate system of ranking could be developed - ranked by your knowledgeable peers instead of unthinking machinery.

This line actually made me laugh.

Let's assume we have exactly 16,000 players to rank, and that our hypothetical human ranking system involves answering the question "Is player X better than player Y? Yes/No". To compare the entire community to itself in this fashion once would require 127,992,000 votes. Now lets assume we have a team of ten 'trusted peers' working eight hours a day generating the votes at a rate of one vote per second. It'd take them 444.42 days to generate the ~128M votes. It took some effort, but now we've got raw data. Given that 'unthinking machinery' doesn't enter into the equation, it's time to bust out the old pencil and paper! Need I go on?

/wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

Edit: Missed an order of magnitude when calculating time required. Fixed.

128 million votes? Sure... if you're using bubble sort, which is O(n²). If you used quicksort, which has an average case of O(n log n), you'd get something on the order of 150 000 votes, if you choose your pivot players well /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

if we were seriously proposing to do it this way

which we aren't

we wouldn't rate the 16000 people that log in every year, we would rate the (generous estimate) 500 current vets and then update once very few months. you also wouldn't need to do every single relationship because we're human and we can quickly sort into broad categories. it could be done in an afternoon-look how quickly spidey come up with a rough list of commanders.

it's a dumb idea for reasons we've already stated several times. but it isn't mechanically impossible.

Ksero wrote:QUOTE (Ksero @ Dec 18 2007, 11:46 AM) 128 million votes? Sure... if you're using bubble sort, which is O(n²). If you used quicksort, which has an average case of O(n log n), you'd get something on the order of 150 000 votes, if you choose your pivot players well /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

Hehe. I wasn't exactly trying to make the case for efficiency. ~128M is the number of non-repeating two player combinations possible from a set of 16,000 players, calculated using C = n! / k!.(n-k)! where n = 16,000 and k = 2. It is the required number of comparisons if one were to compare every player to every other player. I freely admit that it's quite possibly the least efficient solution I could think of, but I was attempting to demonstrate how 'nice ideas' can rapidly grow to unmanageable proportions.

B