POV-Ray : Newsgroups : irtc.stills : Website Update : Re: Website Update Server Time: 30 Sep 2020 09:30:13 GMT
  Re: Website Update  
From: Shay
Date: 23 Mar 2007 17:40:00
St. wrote:
>     Any other idea's, let me know here.
> 


Wow! I hadn't heard about the IRTC, and here you are starting another
one from the ground up! Some thoughts:

==> Voting

Three category voting is cool in concept but was a failure for the IRTC.
The average standard deviation between categories for each single entry
in 2006 was .59 or just under 3% of the maximum deviation of 20. Proof
that the system did not work.

Streamlining the voting process might encourage more participation.

Scaling the votes would provide a result more consistent with the true
opinions of the voters. By scaling, I mean that the standard deviation
of each voter's votes would be "corrected" to 1. This would be very easy
to automate. Here's an example:

UNSCALED:
        entry1 entry2 entry3 entry4 entry5 entry6 entry7 stan_dev
voter1     11     10      7      8      7      8      7     1.60
voter2     16     15     12     13     14     12      6     3.26
voter3     16     15     10     13      6     11      8     3.64
voter4     16     11      6     16     11      6     16     4.50
voter5     12      9      9     12     10      6      9     2.07
voter6     16      9     14     16     15     10      8     3.46
voter7      0     20     20     20     20     20     20     7.56
Total:     87     89     78     98     83     73     74
Rank:       3      2      5      1      4      7      6

See the problem? Six of the Seven voters preferred entry1, but because
of the "deviant" way in which voter7 ranked the images, entry1 is in
third place. Here's the "scaled" version:

SCALED:
        entry1 entry2 entry3 entry4 entry5 entry6 entry7 stan_dev
voter1  10.62  10.00   8.13   8.75   8.13   8.75   8.13     1.00
voter2  11.84  11.53  10.61  10.92  11.23  10.61   8.77     1.00
voter3  11.65  11.37  10.00  10.82   8.90  10.27   9.45     1.00
voter4  11.33  10.22   9.11  11.33  10.22   9.11  11.33     1.00
voter5  10.97   9.52   9.52  10.97  10.00   8.07   9.52     1.00
voter6  11.74   9.71  11.16  11.74  11.45  10.00   9.42     1.00
voter7   8.68  11.32  11.32  11.32  11.32  11.32  11.32     1.00
Total:  76.83  73.68  69.85  75.86  71.25  68.14  67.95
Rank:       1      3      5      2      4      6      7

Entry1 is now in first place, as most of the voters believe it should
be. This system isn't perfect, but it is an improvement. There is
nothing wrong about the way voter7 voted. There will always be those who
see the world in 0s and 20s. There will always be those who see the
world in 2s and 3s (a problem for everyone else, because 2s and 3s guy
can't vote for his own entry). All types of voting should be weighed
equally.

Here are the formulas for this scaling if you are interested:
SD = Standard Deviation for a set of values
A = average of all values in the set
N = number of values in the set
Set = the set

SD = pow(sum(pow(Set[0]-A, 2)+pow(Set[0]-A, 2)+...), 1/2)

Subtract the Average of all values from each single value and
Square the result. Then, add all of the squared values together and take
the square root of that.

To adjust the standard deviation of a set, just ...

10 + (value-10)/SD

for each value in the set. The 10 could be anything or nothing. It's
there to make the results look like IRTC numbers.

==> Participation

I've always though the IRTC should be a freeware only competition and
that a name should be chosen which invites Scan Line rendered as well as
Ray Traced entries. I would be nice to see the Pixie, Wings, etc.
communities encouraged to participate.

I like the idea of the continuous Tina Chep topic. Will there be a
rating system? I'm going to think on Tina Chep and try to come up with
something. Are you in need of $$ contributions?

 -Shay


Post a reply to this message

Copyright 2003-2008 Persistence of Vision Raytracer Pty. Ltd.