AllegSkill Online

[hunkyhoney]

QUOTE (KofiMan @ Jan 19 2009, 03:29 PM) You win games, your rank goes up faster. Mu increases, and sigma decreases. Rank is mu - 3*sigma, so when sigma drops, your rank increases. Mu is going up too, so your rank increases faster.

over a long term it works, but in a short term it doesnt.
a 0 could stay a 0 for a very very very very very long time.

look, if u get better, your rank will first DROP!
and then it will increase.
by changing the equation, your rank will increase appropriately and then even out as your skills stop developing.

forexample, if your a 0, and you show a rank 12 skill level, you rank will be at 12, rather than sticking at 0.
however, if that rank 12 player shows a rank 6 skill level, then his rank will be decreased.

this removes the whole "trial period" to get your rank up to its appropriate number. isn't that what you guys wanted? a more accurate ratting system?

[hunkyhoney]

didn't anyone do pure math 30-1
or 1st/2nd year university math (if your in the US)?

[sgt_baker]

What on earth are you on about?

"if u get better, your rank will first DROP!"

Eh? Please explain that to me.

"forexample, if your a 0, and you show a rank 12 skill level, you rank will be at 12, rather than sticking at 0."

Now you're off in to a world of fantasy. Are you saying that (12)s are... no. You expain what you're on about.

[juckto]

Do they speak English in What?

[hunkyhoney]

QUOTE (sgt_baker @ Jan 19 2009, 03:45 PM) What on earth are you on about?

"if u get better, your rank will first DROP!"

Eh? Please explain that to me.

"forexample, if your a 0, and you show a rank 12 skill level, you rank will be at 12, rather than sticking at 0."

Now you're off in to a world of fantasy. Are you saying that (12)s are... no. You expain what you're on about.

*sighs*

sigma is the standard devation right?
lets say its based off of points...

one game u get 100 points. than ur mu becomes 100 and sigma 0
the next game u get 200 points. than ur mu becomes 150, and sigma 50!

based off of the equation, because you went up 100 points, ur rank goes down 150 points.
therefore your rank is based off of having a mu of 50, rather than 150.

for my modified equation.
if you have one game 100 points, ur mu becomes 100, and sigma will become -100
thus your rank will be based off of 200 points to account for the sudden increase of skill.
the next game you get 200 points, than ur mu becomes 150 and siga 50.
and this time, your rank will be based off of 200 points! including the sudden increase of skill!
---

simple enough? or do you have to go to university?

[badpazzword]

Going to university isn't enough to understand those mere ten pages, trust me.

[MrChaos]

QUOTE (hotnoob @ Jan 19 2009, 06:39 PM) over a long term it works, but in a short term it doesnt.
a 0 could stay a 0 for a very very very very very long time.

look, if u get better, your rank will first DROP!
and then it will increase.
by changing the equation, your rank will increase appropriately and then even out as your skills stop developing.

forexample, if your a 0, and you show a rank 12 skill level, you rank will be at 12, rather than sticking at 0.
however, if that rank 12 player shows a rank 6 skill level, then his rank will be decreased.

this removes the whole "trial period" to get your rank up to its appropriate number. isn't that what you guys wanted? a more accurate ratting system?

Your talking to one of the people that worked with Baker early on in this endeavor. Baker has been eating, breathing, and living this subject matter for a long time now (over a year now easily). I respectfully call him The Professor since he knows knows knows this stuff now inside and out.

Don't take this unkindly:
Please take a deep breath and consider the fact that conservatively well over a man year of effort was put into this being implemented. Tests run, code checked, logic hashed and rehashed, even some consultations with MSResearch (the makers of this fine game AND TrueSkills which this is based around). So for it to get to this stage with that kind of oh @#(! is simply looney. Somehow you've got your knickers in a twist, about what exactly I can't follow, since at the base of it your conclusions are based on a skewed view of things. I promised myself to not discuss anything with someone who gets all worked up in this matter without a basic understanding of statistics and probability. There is no shame since the theory behind this thing ( go read the research papers published on it) itself was a bitch kitty to get to the bottom of due to all the subtle bits and pieces. That's with my schooling, plus real world experience of the broader subject matter.

So in closing before you set yourself on fire and get the bashers merrily bashing you then please relax a bit.

PS Having negative and positive values for sigma is, of course, not an issue. If you think it is then all I can tell you is to crack open a textbook and begin reading.

edit: oh hotnob hotnob hotnob most of us are educated peeps but honestly your college degree means well you have a college degree and not much more. Ok you've reached the won't listen and off to the races stage. Best of luck with that one

[juckto]

QUOTE simple enough? or do you have to go to university?[/quote]
No, I think you'd have to attend, errh, I'm not entirely sure what it's called. That place where they train the people that work at mental asylums.

[hunkyhoney]

QUOTE (Badp @ Jan 19 2009, 04:08 PM) Going to university isn't enough to understand those mere ten pages, trust me.

i took a look at the link... and according to it, trueskill works in the long run!
But so does the modified equation! the only difference is that the modified equation REMOVES the need for it to be LONG TERM.

[badpazzword]

Nevermind me