DPChallenge: A Digital Photography Contest You are not logged in. (log in or register) 
 

DPChallenge Forums >> Challenge Results >> Score Graph
Pages:  
Showing posts 1 - 18 of 18, (reverse)
AuthorThread
 06/23/2003 05:17:30 AM · #1
mbardeen gave me a neat little perl script that logs my score over the week and I thought one or two of you might be interested in this graph it created of my Off-Centre entry.



Message edited by author 2003-06-23 05:17:49.
 06/23/2003 05:38:26 AM · #2
Ben... WOW...
The learning potential using this is huge. What people see, how they see it, what they respond to, and how we vote at different times of the week... or...or... or... This is cool... Graphs can tell a story!
 06/23/2003 05:46:01 AM · #3
Very cool. It would be interesting to see this trend analysis plotted against the number of votes and the voting distribution. Fun stuff.
 06/23/2003 08:18:05 AM · #4
Here is my own score graph from the off-center challenge..




I'm still working on the program, so it's not for general consumption yet. Additionally this type of data collection may run counter to the terms of use agreement of the site, so I won't release it until I get the OK from Drew and Langdon.

-Matt

Message edited by author 2003-06-23 08:19:52.
 06/23/2003 10:43:21 AM · #5
Could you add the days on the x axis?
 06/23/2003 10:53:41 AM · #6
At the moment no.. I would have liked that for my own use. I should have this fixed soon.

My graph is from June 18th and is logged in 10 minute intevals from that date.

Message edited by author 2003-06-23 10:55:18.
 06/23/2003 11:05:48 AM · #7
and can it chart the first derivative of the data too please :)

Would be good to track the rate of change as the world spins and different cultures take a look.
 06/23/2003 12:56:27 PM · #8
They have 12 step programs available...
 06/23/2003 03:55:50 PM · #9
The first step is admiting you have a problem. LOL!!!

Just kidding. You could get some interesting information out of that.
 06/23/2003 05:43:34 PM · #10
you guys are nuts :)

Other than for fun, how on earth could this be useful? Once you've submitted, you can't change any of it. It's like graphing the temperature...all it is is history.

Not to tell you how to spend your free time, but wouldn't you improve your photography faster by...say...taking pictures?

Just a thought.

Pedro

ps no, i didn't intend to offend anyone with this, I'm mostly kidding.
 06/23/2003 05:43:35 PM · #11
and it never double-posts my edits...weird.

maybe it's a netscape thing?

Message edited by author 2003-06-23 17:51:30.
 06/23/2003 07:13:43 PM · #12
Originally posted by pedromarlinez:
you guys are nuts :)
Not to tell you how to spend your free time, but wouldn't you improve your photography faster by...say...taking pictures?


See.. that's the beauty of being a computer geek. I spent two hours writing it, now I have all sorts of interesting data to look at and I can (hopefully) escape coming to the site every 10 minutes to check my score. The program does it for me! My main reason for writing it was just to see if I could do it, but I was also curious about the score trends.

-Matt

p.s. Pictures? why would we want to take pictures? I'm just here for the arguments in the forums!
 06/23/2003 07:57:36 PM · #13
Originally posted by mbardeen:
Originally posted by pedromarlinez:
you guys are nuts :)
Not to tell you how to spend your free time, but wouldn't you improve your photography faster by...say...taking pictures?


See.. that's the beauty of being a computer geek. I spent two hours writing it, now I have all sorts of interesting data to look at and I can (hopefully) escape coming to the site every 10 minutes to check my score. The program does it for me! My main reason for writing it was just to see if I could do it, but I was also curious about the score trends.

-Matt

p.s. Pictures? why would we want to take pictures? I'm just here for the arguments in the forums!


:)
 06/23/2003 10:56:30 PM · #14
Originally posted by mbardeen:
Originally posted by pedromarlinez:
you guys are nuts :)
Not to tell you how to spend your free time, but wouldn't you improve your photography faster by...say...taking pictures?


See.. that's the beauty of being a computer geek. I spent two hours writing it, now I have all sorts of interesting data to look at and I can (hopefully) escape coming to the site every 10 minutes to check my score. The program does it for me! My main reason for writing it was just to see if I could do it, but I was also curious about the score trends.

-Matt

p.s. Pictures? why would we want to take pictures? I'm just here for the arguments in the forums!


STOP! Those with allergies to statistics, do not read this post, as it may cause a severe reaction... you have been warned.

I've been looking at the distributions of scores, and found that, surprisingly, many of the distributions for individual photos are pretty much gaussian (normally distributed). Some, however, are decidedly not normal. A co-worker and I were working on a similar problem, and I realized thaat the plot we had devised to deal with that was directly applicable here. I'll illustrate with an example of one of my submissions (primarily peppers). First, a histogram of the scores given:


Now you can see this particular distribution is not normal, specifically it's skewed right. What we've done is plot this on a cumulative normal probability scale, thus:



Why is this so valuable? First, the slope turns out to be the inverse of the standard deviation of the data. Second, the x intercept (the point at which the best-fit line crosses the x axis, where Z = 0) is roughly equivalent to the mean. Roughly, because it will be slightly different if the distribution is not normal. In this case, the votes are spread out widely, with a gap in the low middle that shows as a sag in the curve. This tells us that though most really liked the photo, there was a significant minority that thought otherwise! One of my other submissions produced this plot:



Here we see that the slope is much higher (lower standard deviation), fit is better (more normal), there are no "outloier" scores. Although in general the voters did not like this shot nearly as well (neither did I, to be truthful), they all agreed it was just OK.
I've found this to be a great tool for looking at the voters' reactions to particular shots, especially those that produce widely divergent reactions.
All that's necessary to construct such a plot are the vote totals for each score level (1-10). Paste 'em into Excel, let the spreadsheet do the work.
I think it would be cool to roll this together with Matt's plot, and see things develop in real time!
Damn, I am SUCH a geek.

06/24/2003 02:37:24 AM · #15
freak.
 06/24/2003 11:14:00 AM · #16
Originally posted by kirbic:


STOP! Those with allergies to statistics, do not read this post, as it may cause a severe reaction... you have been warned.

I've been looking at the distributions of scores, and found that, surprisingly, many of the distributions for individual photos are pretty much gaussian (normally distributed). Some, however, are decidedly not normal. A co-worker and I were working on a similar problem, and I realized thaat the plot we had devised to deal with that was directly applicable here. I'll illustrate with an example of one of my submissions (primarily peppers). First, a histogram of the scores given:


Now you can see this particular distribution is not normal, specifically it's skewed right.


Um, that distribution is clearly skewed left. Your analysis of it remains valid.
 06/24/2003 11:28:24 AM · #17
Hey Kirbic,

Can you run the stats on DrJones Magazine entry:
//www.dpchallenge.com/image.php?IMAGE_ID=24218

I'd like to see what happens with the non gaussian 1's piled on the bottom.
 06/24/2003 04:50:03 PM · #18
Originally posted by Seeker:
Hey Kirbic,

Can you run the stats on DrJones Magazine entry:
//www.dpchallenge.com/image.php?IMAGE_ID=24218

I'd like to see what happens with the non gaussian 1's piled on the bottom.


Joe, thanks for pointing out the wierd scoring on this one.
I ran the numbers, with very interesting results...
First, the histogram & cumulative plot as-voted:





The histogram looks nothing like normal, but the cumulative plot sez the problem is only in the bottom end (big surprise there). The fit is 0.97 (R Squared value), and the slope is 0.3001, which is very low, indicating "high disagreement" among voters.
Removing the 25 1's & re-plotting yeilds a very interesting result...





Now the histogram could be a chopped-off normal, with the exception of the "hump" at 6-7. The cumulative plot agrees. Slope rises to 0.4189, still low but much more toward an "average" value for DPC shots. The fit rises to 0.997. This is a very good fit to a normal distribution. Evidently those who voted "1" had a completely different mindset, and thus made up a statistically distinct distrribution from other voters. Removing the second "distribution" reveals the underlying normal one. This is the first time I've seen this so clearly here on DPC.
I would guess the 1-voters took offense to the sexual content, though that's pure conjecture (or would it be impure conjecture??) ;^)

Pages:  
Current Server Time: 04/24/2024 08:29:32 PM

Please log in or register to post to the forums.


Home - Challenges - Community - League - Photos - Cameras - Lenses - Learn - Prints! - Help - Terms of Use - Privacy - Top ^
DPChallenge, and website content and design, Copyright © 2001-2024 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 04/24/2024 08:29:32 PM EDT.