Know Your Final Score Now - DPChallenge Forums

Nikon AF-S NIKKOR 24-70mm f/2.8G ED

12/14/2005 12:19:38 PM · #1

Oh that painful and addictive little �update� button. Should I push it? No I can�t take seeing my score drop again. Wait, maybe it will go up. Should I? Should I? Just do it! Crap.

Well, that is my DPC life, but I really can�t say about yours. For me, being able to settle down my angst just a little is well worth a little number crunching. Many of us have heard �just wait until about 150 votes and then you will have your score plus or minus about 2 tenths.� You know, there is some real truth to that. Why wait that long, however.

I have been tracking my scores with the hope of being able to validate that little rule of thumb. Given that there are lots and lots of variables including but not limited to: different results for different quality of photographer and different results for photos in which you are taking a risk or running in the mainstream. So for these next challenges I will be not only tracking my scores, but my predictions as well. For a numbers geek like me this is excellent entertainment.

For those of you that would like to test my numbers on your own results during the week, here you go:

Simple Mode:
Votes---Final Score
<10-----Your score + 0.53
50------Your score + 0.33
100-----Your score + 0.22
150-----Your score + 0.18
200-----Your score + 0.16
250-----Your score + 0.15
>300----Your score + 0.12

More Refined and flexible mode:
Votes-----Final Score--------------------r2
<50-------y = 0.4418x + 3.2799----------0.5199
51-100----y = 1.0183x + 0.2364----------0.9513
101-150---y = 1.0691x - 0.1518----------0.9776
151-200---y = 1.056x - 0.1308-----------0.9890
201-250---y = 1.0803x - 0.2757----------0.9886
251-300---y = 1.1343x - 0.5837----------0.9861
>300------y = 1.1353x - 0.6278----------0.9932

Good luck and I would be interested in how these work for others. I of course will try to improve them with each new challenge.

Message edited by author 2005-12-14 12:24:26.

Dr.Confuser

Nikon D800

12/14/2005 12:26:38 PM · #2

I'm quivering with excitement!

With Love & Affection
-- The Math Guy

aeli2468

Olympus E-500 EVOLT

Olympus 14-45mm f/3.5-5.6 Zuiko Digital ED

12/14/2005 12:48:25 PM · #3

as an employed statistician myself, I LOVE that you included the r2. That is too too funny!

Message edited by author 2005-12-14 12:56:43.

Samsung Galaxy S21, S21+, S21 Ultra

12/14/2005 01:10:23 PM · #4

I was going to include percent error too but thought that was over the top :)

Bear_Music

Sony 24-70mm F4 Vario-Tessar T* FE OSS

12/14/2005 01:19:23 PM · #5

If I'm reading this correctly, the "short form" tells me that my image, at 78 votes and 5.4744, is predicted to finish between roghly 5.9 and 5.8? Just to see if I'm reading this correctly?

R.

seancgn68

Canon EOS-450D Rebel XSi

12/14/2005 01:55:33 PM · #6

LOL, I gotta say, I never thought of using regression analysis to predict my score. Neat idea!

Sigma 15mm f/2.8 EX Diagonal Fisheye for Nikon

12/14/2005 01:58:57 PM · #7

Not exactly. (Please note that these are developed from my results and there is lots of varibility between photogrpahers so your results could be way different.)

The short version requires the use of the speific number of votes (within say 2 either way). So you cannot use the short version again until you hit 100 votes.

An appropriate equation can be used for any number of votes within a range. So using the equation for 51-100 votes, I predict that if you were me your final score will be 5.8110.

Does that make sense, kind of?

muckpond

Nikon D200

12/14/2005 01:59:41 PM · #8

as someone who is wading into the SPSS waters for the first time for final projects this semester, i must say only this:

i hate you all.

:P

thank you...back to writing papers.

idnic

Canon EOS-60D

Canon EF-S 17-55mm f/2.8 IS USM

12/17/2005 09:52:49 AM · #9

Bumping this thread cause I think it rocks. Yes, I'm a geek. :P

ralph

Nikon Z9

12/17/2005 10:08:01 AM · #10

interesting numbers ..
but i'm wondering if there is a signif. diff. of boost numbers between
Member Challenges & Open Challenges ?

i would expect there to be more drop out with open vs. member ...

Nikon AF Nikkor 50mm f/1.8D

12/19/2005 11:23:40 AM · #11

Some additional results are now available for me to play with and there is a distinct difference between latest results and previous results. I may be able to make a distinction between open and member challenges, but possibly this recent difference is due something else entirely. Interestingly, using the regresion equations, from 100 votes on, I was within 0.05 units of my final socre for the Holiday Catalog challenge.

jpeters

Nikon D200

12/19/2005 11:43:13 AM · #12

Ok, i'm supposed to get a 5.8507 on the 4-5am challenge. We'll see how it turns out!

jpeters

Nikon D200

Nikon AF Nikkor 50mm f/1.8D

12/21/2005 03:02:11 AM · #13

Originally posted by jpeters:

Ok, i'm supposed to get a 5.8507 on the 4-5am challenge. We'll see how it turns out!

Well, it turns out I got a 5.695. Apparently the voters didn't pay attention to your equation when voting on my photo.

Nikon NIKKOR AF-S DX VR 55-200 f4-5.6 G IF-ED

12/21/2005 03:32:53 AM · #14

More data, more data!

One of mine was very close from 100 votes on and one was off nearly the exact same as yours. I shall recalculate with the additional data and for those interested.

rayz1

Nikon D300

12/26/2005 12:33:10 AM · #15

Out of interest for you, when I read your post I checked out my projected score, at 148 votes, the projected score by your formula was 6.2273
Eventual score for my entry, Deferred, was 6.226

Bud

Canon EOS-550D Rebel T2i

Sigma 18-250 f/3.5 - 6.3 DC OS HSM

12/26/2005 03:46:46 AM · #16

This was at 6.9xxx after 30 votes in the "Visual Puns" challenge - final score 6.476 a drop of nearly .5000 with 225 final votes.

This was at 6.7xxx after 120 in "Your corner of the World II" but drifted to 6.628 with 180 votes. I remember this because it is the first time I saw my score dropping like that.

But, that being said, the normal for most of my photos is to bounce up a bit at the end of the scoring, which is what your numbers would indicate.

zxaar

12/26/2005 04:22:56 AM · #17

Simple Mode:
Votes---Final Score
<10-----Your score + 0.53
50------Your score + 0.33
100-----Your score + 0.22
150-----Your score + 0.18
200-----Your score + 0.16
250-----Your score + 0.15
>300----Your score + 0.12

In this formula, there is a distinct possiblity of failure when the votes are less than 150,
biggest failure would be fore less than 10 votes, the assumption that your score increases after 10 votes , could fail with high probablity.

after 150 votes the score is pretty much stablished and is seen that it increases little bit, 0.15 to 0.18 is fair judgement.
anyway good effort.

mycelium

Canon EOS-350D Rebel XT

Canon EF-S 10-22mm f/3.5-4.5 USM

12/26/2005 04:32:36 AM · #18

Originally posted by aguapreta:

This was at 6.9xxx after 30 votes in the "Visual Puns" challenge - final score 6.476 a drop of nearly .5000 with 225 final votes.

I think what's happening here - and what's hard to account for in a predictive model - is the dropping of votes, at the end of the voting period, of people that didn't vote on at least 20% of images. You might notice that sometimes your final vote count is actually lower than the vote count shown right near the end of the voting phase. In my own experience, this can change scores (edit: up to) +/- ~.3.

Message edited by author 2005-12-26 04:33:49.

zxaar

12/26/2005 04:44:09 AM · #19

yes, the final score shall not vary more +/- 0.3 if we are using score of >200 votes as base.

But then almost everybody understands this and we do not need a model for it. ;-)

ShutterPug

Canon EOS-6D