With allegskill...

[Malicious Wraith]

This is not personal. I ranked better under AS then I did under HELO.

However, "Joining" the better team, and being on the better team are two entirely different things that AS, again, can not deal with sufficiently. I was a big player in 360 games, and I know how Ms Skill was applied, and how effectively it worked.

It worked because players were forced to join match made teams, and had no say in what their team mates or opponents would be, and usually on teams between one to four players.

You are trying to apply that same system to AGS, and sure, you can determine victories, but the fact of the matter is that those victories reflect a teams performance (a teams victory), more than any individual contribution.

The numbers are clear, and they reflect those flaws.

[Broodwich]

this thread was rather amusing ^^
and that video was teh awesome

[omegahero]

i have no qualms with brood, kap and nuke's ranks, they seem about right. I have not played with H_Mallow in any recent memory but crazebomber and whorez are not better players than them (although whorez AS is a bit low)

[sgt_baker]

QUOTE (Picobozo @ Jun 17 2009, 07:34 PM) Picbozo stuff

Nothing is going to be perfect. You're correct in that it'd take a load of code-change to even begin to contemplate a points-based system realistically. I could go on about this in great detail, however I'll say just this:

We've done everything we can to expose in-game stats to players. You now have things like base kills and drone kills. Given the current system, little more is possible. Sure a fleet of genius developers could descend upon this game and reveal a veritable smorgasbord of new in-game stats. This is simply not happening.

I'm not the person to modify the whole of alleg to please three/five/eight people. I'll make a fuss when something affects everyone.

What you currently have as a ranking system is the cutting edge of that area of maths, understanding and research. Literally the cutting edge. Microsoft have been in touch with me to inform me that we're the first game to use time-relevant updates to 'skill'. Why is this important? In the old days, people used to stack the last two minutes of a won game in order to gain rank. Dunno if anybody has noticed, but that doesn't work any more.

I've sent all of the code we use in Alleg to the clever people over at Microsoft. The clever people at MS are fine with it.

Until the clever people have a problem, we're unlikely to make sweeping changes.

If you have a better idea for ranking, please notify myself and the community at large as to your breakthrough.

[MrChaos]

QUOTE (Picobozo @ Jun 17 2009, 01:34 PM) Mr Chaos, I know that MS ranking wasn't perfect, but it did a pretty good job of showing you who was skilled.

Let me more blunt the old MS ranking was Ba AD, horrible, @#(!, way to rank and was based mostly, just like WOW, on time in game.

What is needed are somethings akin to strange attractors to help determine ranks. Some ingame action(s) that are pointers to a person's skill in the game overall. It may be obvious as hell or totally obstificated to us all atm. Most likely it is some combination of ingame actions and interactions. I can not emphasis this one enough, NO one knows what they are and without analysis you will be wasting your time. It would be a bit like trying to put a 100,000 piece puzzle together by trying each piece individuals until it fit, trying to put two different pieces together and guess the picture. A bit like the monkeys and the typewriter old saw.

Since we need to find this/these strange attractor like objects we need to data mine a tremendous amount of things. In order to do this one would need all of the actions of every player in every game and when/where/who they performed then on, every core and all of it's changes, every faction and all of their changes, every game commander, every fellow player on each team AND all of the interactions between every game action, core, players, commanders, team mates, and a bunch of stuff I'm sure I forgot. Add in the timing factor; people dropping, and joining, commanders coming and going and now your matrix is a multi-dimensional monster.

IIRC We have none of this data atm since the game event interactions are not captured. So we would be starting from scratch and would need to comply a data base of thousand and thousands of interactions, events etc across as large a range of games as we have currently IMHO . The amount of effort required to be able to use individual performance to rank everyone is well beyond not only the scope of the total man hours of the community but the game does not give enough information in it's current form to make it happen. It's a herculean effort which IS doable given a team of people who devoted a @#(! load time, software we do not own, and resources we do not have, nor will we in the forseeable future to make it happen.

All to get a better accuracy on a system that can determine game outcome in the 80 percent range IIRC. You are, of course welcome to undertake the effort but I advise you to give it a strong pass.

With respect
MrChaos

edit: MW in order to predict game outcome in the 80 err low 90 percent range you MUST know how the player ranks against the community. Game play should improve if one was forced on a team but that bud is a controversial topic given the nature of the community.

[Picobozo]

QUOTE (MrChaos @ Jun 17 2009, 02:02 PM) Stuff

That wasn't the opinion that I offered, but like I said i'll refrain from further input.

Sorry.

[takingarms1]

QUOTE (Malicious Wraith @ Jun 17 2009, 02:41 PM) However, "Joining" the better team, and being on the better team are two entirely different things that AS, again, can not deal with sufficiently.

I cannot begin to contemplate what this is supposed to mean

I'm not sure your critiques are really relevant, or even make sense. You seem to be saying that because AS works by using the outcome of the team (win/lose), it is inherently limited/flawed. But frankly that seems to be why it works - what better measure of skill is there than whether you help your team win?

Admittedly it is possible for you to ride on other people's skill in individual instances, but even the most determined stacker can't do that forever, and eventually AS finds the times that you don't and clobbers your rank for it (see Baker's post above).

[MrChaos]

QUOTE (Picobozo @ Jun 17 2009, 01:34 PM) Mr Chaos, I know that MS ranking wasn't perfect, but it did a pretty good job of showing you who was skilled. I'm sure that if MS supported the game for longer then a year we would have had many iterations of a ranking system that would have hopefully addressed the balance issues. Current developers have the opportunity to modify these stats with some of the work they have already done to help account for issues (like time played, or events performed per hour) to help get a more accurate ranking.

Pico

Actually MSResearch, the very makers of this game, are also the originators of TrueSkills and it is my completely random opinion this game was why they went in an entirely new direction in the matter on how to do ranks accurately rather then haphazardly

[Grimmwolf_GB]

sgt_baker, how much influence does it have on the rank of old vets, that their learning phase (1999/2000-200x) was not recorded, while the learning phase of new vets, voobs and noobs is on file? My rank is 16, my command rank is ridiculously high, but just because I mainly won the games, because I did not have to learn while being "recorded".

The second item I have some problems with is the stack rating. I know that I do stack from time to time, but I also draw/drew a stack as well (those days are probably over). It is quite difficult to give it a better name, but stack rating is inaccurate.

[takingarms1]

I believe this question was asked and answered previously... short answer is, AS does account for this, and does it by raising the uncertainty rating of the rank until the rank stops fluctuating, then the rank stabilizes.