Page 20 of 32
Posted: Thu Dec 13, 2007 10:52 pm
by jgbaxter
Terralthra wrote:QUOTE (Terralthra @ Dec 13 2007, 03:39 PM) If you'd actually worked seriously with statistics for 10+ years, you'd not evade simple and rational questions about your dubious idea. You'd just answer them, as they'd be easy questions to answer if you had over a decade of statistical experience to draw on, and your system has any real viability.
Perhaps if every other words wasn't a direct or veiled insult...
Posted: Thu Dec 13, 2007 11:10 pm
by Kltplzyxm
LOL. I love trolls. /rolleyes.gif" style="vertical-align:middle" emoid=":roll:" border="0" alt="rolleyes.gif" />
Posted: Thu Dec 13, 2007 11:16 pm
by sgt_baker
This thread is exactly the reason I chose to develop behind closed doors.
JG, given your ten years of experience you should be able to digest this:
ftp://ftp.research.microsoft.com/pub/tr/TR-2006-80.pdf
Read, mark, learn and inwardly digest. Once you're done I suggest you contact Ralf Herbrich (he's a nice guy) to point out the errors in his ways. When MSR have entered your desired state of realisation and corrected the obvious flaws inherent in their system in accordance with JGWorldâ„¢,
and have published another paper, I'll gladly implement said system.
Posted: Thu Dec 13, 2007 11:24 pm
by jgbaxter
Look more (un)veiled insults. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
I think I already read that last when Pook was working on helo, I'll give it another look though.
Posted: Thu Dec 13, 2007 11:29 pm
by Evincar
i just have three questions JGB:
1)why do you try to talk us into a system that will not measure skill?
2)how do you guarantee EVERYONE'S rank?
3)since you have been working with statistics for about 10 years, explain me how do you plan to build a statistical ranking with a non statistic criteria?
Posted: Thu Dec 13, 2007 11:36 pm
by jgbaxter
I'm sorry, if you want a test, my paypal account is open for donations, I'll say, $99 is fair.
No system can guarantee anyones rank to be accurate.
I'm conveying my opinion, do with it as you will.
/smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />
The main issue I find with any system will be one of experience, which does not equate skill.
Let's say we have;
Aarm
spidey
Weed
noob(0)
noob(0)
I think that team will beat;
Lyk
Shiz
Terra
Baker
Factoid
9 times out of 10.
Posted: Thu Dec 13, 2007 11:40 pm
by Evincar
QUOTE The main issue I find with any system will be one of experience, which does not equate skill.[/quote]
lol?
Posted: Thu Dec 13, 2007 11:42 pm
by sgt_baker
I'm serious. If there is something awry in TrueSkill the developers will be happy to discuss the matter.
Do you know what is an insult? Spending countless hours pouring over the ranking problem only to have countless upstarts rubbish the effort. This is an open community. If anyone has a better system they're welcome to engage their grey matter in the implementation and testing of their system. I strongly suspect that the reason our group was granted access to the ASGS data just might have had something to do with the suspicious abscence of hot air and bluster.
For the record, when this project was started we considered numerous systems. TrueSkill was the only system that survived the numerous checks and balances imposed by the stats group. If there were a better system you could place a hefty bet on our using it.
Anyhow, I'll leave you lot to your arguments, since this really isn't my domain.
B
Posted: Thu Dec 13, 2007 11:43 pm
by Cadillac
jgbaxter wrote:QUOTE (jgbaxter @ Dec 13 2007, 11:36 PM) Let's say we have;
Aarm
spidey
Weed
noob(0)
noob(0)
I think that team will beat;
Lyk
Shiz
Terra
Baker
Factoid
9 times out of 10.
BS.
Posted: Thu Dec 13, 2007 11:43 pm
by Kltplzyxm
LOL. Thanks for joining my ignore list Bax. Life has gotten a lot less annoying. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
I'm glad you'll know you're the only one on it.