Free Allegiance Forums

Posted: **Mon Jan 19, 2009 6:36 pm**

Just wanted to say again nice work Baker.

I notice I have 0 Command Draws listed but I can think of at least 4 off the top of my head. What gives?

Posted: **Mon Jan 19, 2009 6:37 pm**

sgt_baker wrote:QUOTE (sgt_baker @ Jan 19 2009, 11:21 AM) The maths behind kills/hour aren't really up for debate. Hint: It's kills/hours played!

I have planned to expand the non-Trueskill stats envisaging someting along the lines of kill-rating/bomber-rating etc etc. There are certain constraints preventing me from doing so at this time. Hopefully they will resolve in the near future.

I'm not done with this project. Today is Day 1.

okays...

btw, i took a look at the wiki...
n if u just look at the formula that was provided...

there is a massive flaw! *im not so sure cause it dont say much*
but what i see is that if you suddenly improve a TON! than your rank will go down... instead of up... because the sigma is an absolute value.
perhaphs you could include a negative and positive "sigma" into the equation...

Posted: **Mon Jan 19, 2009 7:02 pm**

Edit: Do you mean the "conservative rank" thing? From what I understand, that's supposed to provide the 'lowest' value you probably have, to err in favour of under-estimating rather than over-estimating players (so you know you're "at least" this good), right? Also, wouldn't "improving a lot" mean your mu goes up, so unless your sigma also goes up (which from what I understand it never does), your rank can't possibly fall?

Posted: **Mon Jan 19, 2009 7:03 pm**

Nice work baker! And nice work Tigereye for helping implement it and all other guys behind it too!

Forgot this in my previous post

Posted: **Mon Jan 19, 2009 7:16 pm**

spideycw wrote:QUOTE (spideycw @ Jan 19 2009, 01:36 PM) Just wanted to say again nice work Baker.

I notice I have 0 Command Draws listed but I can think of at least 4 off the top of my head. What gives?

read the wiki!

oops

Most likely the games were not used do to some anomaly in the data Spidey

Posted: **Mon Jan 19, 2009 7:16 pm**

Thanks for eventual positive attitude

Good luck to you in your further work.
QUOTE but what i see is that if you suddenly improve a TON! than your rank will go down... instead of up... because the sigma is an absolute value.
perhaphs you could include a negative and positive "sigma" into the equation...[/quote]
no. As you can see, sigma is present in all the equations as sigma^2, so making it negative has no effect.
And, as sigma is essentially a possible deviation of your 'true' rating from Mu, there's no sense in it being negative at all.
And, I don't follow your logic about 'suddenly' improving a TON. Do you mean that player starts winning all the games? I don't see how the rating goes down in this case...

Posted: **Mon Jan 19, 2009 9:47 pm**

pkk wrote:QUOTE (pkk @ Jan 19 2009, 12:26 PM) Somehow Opera $#@!s up the rank column, if you sort it...

I only tested with firefox and IE, and had to handle that column differently for both
(FireFox's columns are 0-based, whereas IE's are 1-based)
Looks like Opera is also 0-based but I was treating it like IE.

I'll fix that when I get home.

--TE

Posted: **Mon Jan 19, 2009 10:27 pm**

Zruty wrote:QUOTE (Zruty @ Jan 19 2009, 12:16 PM) Thanks for eventual positive attitude Good luck to you in your further work.

no. As you can see, sigma is present in all the equations as sigma^2, so making it negative has no effect.
And, as sigma is essentially a possible deviation of your 'true' rating from Mu, there's no sense in it being negative at all.
And, I don't follow your logic about 'suddenly' improving a TON. Do you mean that player starts winning all the games? I don't see how the rating goes down in this case...

do u no wat absolute values are?
NEVER NEGATIVE!
according to the equation the higher ur standard deveation aka sigma, the lower your rank will be.
Therefore, if you suddenly get better your sigma will increase, as well as if you suddenly get worse, then ur sigma will increase.
thus leading to a decrease in rank.

Rather than lazily squaring the sigma, which really isn't needed... ya put |var| to make it an absolute value... depending on the programing language... but there should be an absolute function, if not, than one can be made.
anyway... rather than squaring it, you have to seperate the negatives and the positives.
then you need to add them, thus negative+positive = +-
once that has been completed, the equation needs to be fixed...
however, idealy, you wouldn't use only one equation, you would implement some of those programming skills.

here is how ya would do it for those who can program...
btw, it says the site was done in php, so im gonna asume it was made using php and mysql.

Code: Select all

<?php
    //connection  junk here.
    $positiveSigma=0;
    $negativeSigma=0;
    $rank = 0;
    $mu = 0;
    $query="select * from `player`';
    $result=mysql_query($query);
    if($row=mysql_fetch_array($result))
     {
        $mu = $row['mu']; // define the players current mu
      }
    $query="select * from `games`";
    $result=mysql_query($query);
    while($row=mysql_fetch_array($result))
      {
         $count+=0;
         if($mu < $row['points'])// im not sure what sigma is based on... so ill use points.
          {
             $positiveSigma += $row['points'];
           }
          else if($row['mu']< $row['points'])
          {
            $negativeSigma += $row['points'];
           }
          else 
           {
              $count-=1; // gonna ignore the values with the same as the mu, to increase processing speed
            }
       }
      $sigma = $positiveSigma - $negativeSigma;
      $sigma = $sigma/$count; // there is no need to use absolute values now, because there abilities are decreasing if the negative is greater than the sigma.
      // forumula time
      $rank = 0.6($mu +3($sigma)) // the numbers may need to be adjusted slightly to keep the same ranks.

?>

my appoligizes if i left a couple typos or bugs... i didnt use my editor to write this...
if i did, well, ya get the point, it should show you almost exactly how.

Posted: **Mon Jan 19, 2009 10:29 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.

Posted: **Mon Jan 19, 2009 10:31 pm**

KofiMan wrote:QUOTE (KofiMan @ Jan 19 2009, 06:29 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.

you sir are correct!

edit: wellll

if you now add in the bit where when you win against those the system thinks you should you get less "bang for the buck" faster sigma decrease but slower mu increase. When you win against those with a higher rank your Mu goes faster but your sigma decreases slower however overall your ranks goes up. The bottom line is the longer you play the more sure the system gets in your skill.

Free Allegiance Forums

AllegSkill Online