New "Top Photos" Feature

[root]

Marc, Here are some possible solutions; tell me what you think.

 

Maybe separating critiques from ratings makes sense. Requests for

critique often means only a request for votes. Some don't take kindly

to constructive criticism even if it includes a reasonably high

rating.

 

So separate the process. Upload images ONLY for the purposes of

critique. If the criticism produces an improved image or is mostly

positive to begin with, put it in the request for ratings folder and

then the clock starts. The game is rigged now for immediate positive

rating to get it on the two-day or one week pages. I now realize that

many of us have good 7.5 - 8 images that very few will ever see.

 

We might want to do away with the two-day list and replace it with

images that have finally reached a critical number of views regardless

of time. Allow for a detailed search, ie 100-200 views, 7 - 8.5

rating.

 

I think we also should limit uploads for rating and the number of

times you can rate one person to one per day. There, now I've

offended all the popular photographers, several of whom have actually

given me an occassional good rating. . . . .

[mg]

To me, all these ideas are actually good. I'd still like to keep the 2 days category up, though. But adding features could be a problem if it involves too much of an investment. I have no idea. Limiting the number of uploads per person is something I've proposed already somewhere, as well as limiting the total number of pictures a photographer can keep in his portfolio on PN... As for limiting the number of uploads a day to 1 at most per person, it would automatically prevent the front pages to be monopolized by the gangs, or by anyone else. Save bandwidth by getting rid of a lot of old forgotten images would help saving money as well, I guess.

<p>

At the moment, I would just want to thank Photo.net for the great new toy that this new search system is, but getting people back to critiquing and to recover their true freedom of speech would imo be the next goal. And it can only be achieved, I'm affraid by disconnecting a few gangs on the site. The retaliation thing kills critiquing.

<p>

A possibility to disconnect the gangs would be to allow us to give away only a certain number of 10/10s and 9/9s a month : 10 times 10 and 25 times 9 for each of the 2 rating categories. Enough to be honest, not enough to be dishonest...:-)

[mark_thomas3]

Sorry for the 'me too' post, but I would also like to express my appreciation for the efforts of Brian and crew to address what appear to be shared concerns. Marc's 'Regarding..' post above hit the nail squarely, as far as I am concerned. I am one of those who have refrained from issuing many comments/ratings because of my perception that the ratings game was severely influenced by the 'gangs'.

 

The sad part is that some/most of these groups are probably not even aware that they are causing problems, and will be highly offended if any action is taken to solve what they won't see as a problem. To me, the first step would be to get a statistical (and/or psychosocial?!) expert to come up with some ways of analysing the problem from the numbers, rather than from the complaints.

 

What would be the chances of s/electing a panel (elder elves?) that could investigate issues that came out of that analysis, and *then* perhaps look at complaints? I'm sure if you asked for people to dob in (perhaps by direct email rather than on a forum!) those who they think are in the 'gangs' you would get an interesting response, but I'm also sure that this is probably NOT the best way to get started. Although the names that turn up could then be checked against the stats as a further step..

 

One other small suggestion. In sports like diving and skating, the highest and lowest scores are often dropped from the analysis. To me, photography critiquing is very similar, in that the judges may miss something vital about the 'performance', or -heaven forbid- may be biased. Why not use that here, say on the basis that the top and bottom 5% of ratings on an image be ignored..? I am often astounded by the stupid, retaliatory or perhaps just ignorant ratings that are given to an image, and it takes very few of these to kick an image out of the race. This system would have the advantage of not requiring any investigation, and would be difficult to criticise as unfair or selective..

 

Anyway, very best of luck - I think this is the biggest challenge that the site has faced and if it can be improved.. Brian should be elected the next chairman of the United Nations!

 

mt

[imaginator]

I have never liked the "ratings game"... numbers are meaningless! Don't get me wrong, I like everything else about photo.net, just not the numerical nonsense! Give me real comments or give me nothing at all.

[rick vincent]

I like Mark Thomas' suggestion. I also have thought this to be a good idea to only base the rating on a weighted type average system. With my suggestion, however, I would remove both the highest and lowest ratings. For instance at 10 ratings you drop the one highest and one lowest, at 20 ratings you drop the two highest and two lowest, and so on. Then you could average the rest of the ratings for a more accurate calculation that removes some of the 10/10 and 1/1 that can tend to corrupt the ratings system. I think this is what they call a weighted average.

[root]

I just did a search on photographer's average, last month. The

number of images listed that are factored into the average

seems far fewer than the number of images they had that

recieved ratings so I'm guessing there's another filter. The

search brought up only 83 photographers; what's the cut-off?

[Jim_Tardio]

I've looked it over for 2 days now. It's a neat feature but all I get is virtually the same photos and photographers for every variation I apply. Now I know this is likely to happen because if you're top-rated in one thing, it's going to carry over to the other catagories.

<p>

But I know that there are lots of high rated photos by various photographers that have yet to be seen...particularly anything that is more than 6 months old.

<p>

Is this a feature for only the new stuff?

[mottershead]

Carl, the one month view requires 5 images and 20 ratings. I've made the filter wider and I'm waiting for the next time I need to restart the servers to introduce the change. I don't want to bring down the servers just for that.

 

Jim, because of ratings inflation the ratings-based rankings are, unfortunately, dominated by images that received their ratings recently, even the longer periods, and "all". Considering the degree of inflation, it is amazing that some of the older photos show up at all.

 

However, "Highest of Day", "Number of Ratings", "Number of Comments", and "Number of Views" are more or less immune to the ratings inflation, and the "Sum" rankings are less susceptible to it than the "Average" views. (My personal favorite of the ranking methods is "Highest of Day", and of course this has nothing to do with the fact that a couple of my photos were "Highest of Day".)

 

I've solicited ideas for ways to normalize ratings for the filter feature.

[Jeremy Stein]

How about adding yet another view into the database: the highest-rated photos from some number of randomly-chosen photographers. That is, let the computer choose a number of photographers (who are not on any other one of the ways of listing) and display the highest-rated photo that each of them has. This will obviously be a very mixed lot of photos, but it might help us find some good stuff and good photographers in the chaff! If it is feasible to program, why not try it?

[mark_thomas3]

I really like Jeremy's idea. Even the worst photographer will occasionally nail one, and as it currently stands those photos will probably sink like stones..

[root]

I had thought 'photographer's highest of the week' would bring

up a whole host of currently active people who had at least one

image that had five ratings. It only brought up 157 people with

the lowest having an average rating below six. There must be a

filter on this that I can't figure out, but I would like to see it bring

up 500 people.

 

It would be useful if there was a field that recorded the upload

date and showed it as part of our records. Perhaps even include

it with thumbnails so everyone can distinguish between a new

release and an old forgotten image.

 

Keep up the good work. No need to respond to this post.

[mg]

The last time I checked, it seemed that the last 2 days top-rated shots filter showed - again - only pictures having more than 5 ratings. I saw you fixed that a few days back, but it seems to be back to 5 ratings and above. Is that a bug or is there a reason or have I missed something ?

<p>

Besides that, I too like Jeremy's suggestion above.

<p>

Finally, you wrote: "I've solicited ideas for ways to normalize ratings for the filter feature"...

<p>

My personal take on this is kind of rude, sorry for that. But here is is: if a picture gets more than 1 10/10 or 9/10 or 10/9 in the first 5 ratings, the other 10/10 won't be counted in. This simply because it has become a trend to exchange 10s and 9s during the first minutes or hours after an upload. A simple exchange strategy that has permitted to some people to hit the top of the default page on regular basis.

<p>

In general, I suggest that the 10% lowest and 10% highest ratings an image gets don't count anymore once an image got its 11 first ratings. That way, the double tens from the friends will be out, and so will be as well the nasty shots below the belt with 5/5s (or lower) on an image that had an 8 or such as an average after more than 10 votes.

<p>

Example of the kind of problems that would solve available at:

<p>

http://www.photo.net/photo/889369

<p>

Due to 1 person only which gave this a 3/3, I haven't found this image of mine anywhere with the filters we have - except, I think, via "numbers of comments" -, and therefore I haven't received much new comments since this nice hit. Out of 25 ratings, this is the only one below average, and 22 of the ratings are 7 and above in both originality and aesthetics. With the normalization I propose here, 1 or 2 of the 10/10s will go, and so will the 3/3, and as a result, people will see this image again. Countless numbers of good images have faced the same fate all over the site. Sorry for taking one of mine as an example.

[daniel flather]

I like it, seems to work fast too.

[nomade]

Brian, about this: <p>

 

<i>other people have rated your photos 895 times and written 257 comments on them. During the same period, you rated other people's photos 359 times and have written 350 comments. </i><p>

 

How about incluiding this kind of statistics on every members file, publicly visible as is the quantity and average rating today? That would be useful, to know how much each one of us have participated and get back.

[root]

The new ways of gaining access to weekly and monthly stats are

revealing some interesting trends. For example, earlier today I

was delighted to be able to page through 500 photographers

who had one or more images uploaded this week who had

managed 5 or more ratings. What I did not expect were three

people who had at least thirty images who met these criteria!

Ten more had at least ten images. Thirty-two had six or more.

Most of the images from these people have to be getting ratings

without going through the normal viewing process - paging

through images slowly one by one on the request for critique list

or on the ediotors' choice list. I've pretty much given up on that,

but would glady search new images if they could be viewed in

the same way as the current top photos pages.

 

Brian, is this in the works anytime soon?