Allegiance Ranking System

[Papsmear]

QUOTE (Adaven @ Feb 22 2012, 10:46 PM) Hmmm.... perhaps we could consider bundling a dataset and donating it to a data mining community like: UCI Machine Learning Repository.

1) More minds working on the problem. Someone might get lucky and come up with the perfect system for us.
2) Free publicity with a niche group that might yield us some players / contributors.

I'm almost kicking myself for not thinking of this when I was selecting my master's thesis topic.

What a fantastic idea!

[Icky]

QUOTE (RoyBrown @ Feb 17 2012, 07:01 PM) Here's the thing of it, let's toss everything aside and look at two perverse examples.

1) Babel is a rank 16, he can pod almost everything alive and is certainly more capable that the majority of the next 5 higher ranks.

2) Fuzzy is a rank 7, yet newbies without a clue what a nan is are ranked higher sometimes.

Two random examples of why the entire system is messed up, and there are many more anyone can see.

So even if the autobalance was perfect, GIGO.

The ranks do not (and are not MEANT to) measure how good of a dogfighter you are, or how well you nan.

They measure HOW MUCH YOU WIN GAMES. PERIOD.

Babel is a 16 and can whore really well, but sometimes he goes off and whores instead of defending miners maybe. It's possible (and not uncommon) to have a really good DMer with a bunch of kills lose, as you should know.

ALL THAT MATTERS is how often you win.

[Malicious Wraith]

I would love to see more uncertainty on a sliding window to make rankings more fluid, even if less accurate (but only slightly so).

[HSharp]

QUOTE (Bard @ Feb 23 2012, 02:09 AM) Is it just me or does RoyBrown remind anyone else of BJBaxtard?

Was that supposed to be a secret?

[Malicious Wraith]

QUOTE (HSharp @ Feb 28 2012, 01:48 PM) Was that supposed to be a secret?

Bard is slow.

[BillyBishop]

QUOTE (Icky @ Feb 28 2012, 05:53 PM) ALL THAT MATTERS is how often you win.

I'm well aware. All that matters is how often you win, your first 50-100 games.

[Icky]

QUOTE (RoyBrown @ Feb 28 2012, 05:39 PM) I'm well aware. All that matters is how often you win, your first 50-100 games.

Not true. We really don't need to rehash AS for the upteen hundredth time, but it's definitely possible for your rank to change after hundreds of games. Mine did quite a bit when I was more active.

MrC can clarify this in more detail if he has the patience, but I wouldn't count on it since this was all gone over ad nauseum years ago.

[Spunkmeyer]

QUOTE (Icky @ Feb 28 2012, 08:14 PM) Not true. We really don't need to rehash AS for the upteen hundredth time, but it's definitely possible for your rank to change after hundreds of games. Mine did quite a bit when I was more active.

MrC can clarify this in more detail if he has the patience, but I wouldn't count on it since this was all gone over ad nauseum years ago.

Go here and see for yourself:
http://atom.research.microsoft.com/trueski...calculator.aspx

Set up the opponent with whatever mu you want and a low sigma (team of vets you'll be facing when you start out). Set yourself up with default mu and (high) sigma. The change in mu will be huge.

Now lower your sigma as you gain experience. Create a large discrepancy in mu between you and the enemy team. The change in mu will be relatively insignificant no matter what you do.

There is no need to clarify anything. That's how it works. That's how it's supposed to work. Once it thinks it has you dialed in, it won't budge easily. What's being said here is Allegiance may not necessarily be working well with that pattern of quickly zeroing in on your skill level based on early results - that's why suggestions like eliminating early games via a sliding window, or converging more slowly, or allowing for greater variation in mu later on are coming up.

[vogue]

ALL I DO IS WIN WIN WIN NO MATTER WHAT

[SumVeritas]

and... why is this so important again¿?