The problem though is that there will always be at least a few uninteresting screenshots in there from a general point of view. I'm just not certain if abstaining from voting helps though as really every matchup is only to differentiate between two images. Even if similar it comes down to making one decision with no adverse personal consequences. I just need that data to figure out the matchups.

The way it works now is that new images will be compared to high rated images first where possible in order to attempt to quickly figure out where the image falls. As it gets voted on more and more it gets compared to images that are closer and closer to it's existing rating.. hopefully narrowing in on a small window that represents it's actual accurate rating. But when that narrow windows for some screenshots it may in fact be pretty on par with other screenshots within the same window.

I'm using a modified ELO rating system with some severe tweaking of the K constant to grant large numbers of points initially for a win while not penalizing losses too badly against really good rated images.

Some things that come to mind:

• Perhaps some people who participate in #screenshotsaturday don't want to be in this (even if it is free publicity), perhaps there should be an opt-out tag (eg: "#screenshotsaturday-" seems like it could work, the punctuation doesn't seem to interfere with the tag and it's only 1 character lost).
• I feel it would be unfair and somewhat against the spirit of screenshot saturday to keep old images in the 'top images' for longer than say 1 or 2 weeks.

Despite my reservations, actually playing the game was somewhat entertaining, and a good way to quickly go through screenshots.

edit:

I noticed at least one person used '#screenshotsaturdays' (notice the extra 's'), so I guess that is something to keep in mind.

Some things that come to mind:

• Perhaps some people who participate in #screenshotsaturday don't want to be in this (even if it is free publicity), perhaps there should be an opt-out tag (eg: "#screenshotsaturday-" seems like it could work, the punctuation doesn't seem to interfere with the tag and it's only 1 character lost).
• I feel it would be unfair and somewhat against the spirit of screenshot saturday to keep old images in the 'top images' for longer than say 1 or 2 weeks.

Despite my reservations, actually playing the game was somewhat entertaining, and a good way to quickly go through screenshots.

edit:

I noticed at least one person used '#screenshotsaturdays' (notice the extra 's'), so I guess that is something to keep in mind.

I think your concerns were some of my biggest in doing this.. mainly because I don't want the competitive element to overshadow *anything*. The side by side comparison is purely to let you choose one that appeals to you more on some level so we can ultimately bubble up the really cool ones from the week. The top images are indeed one week only and you only are comparing images within the same seven day window.

The stream is a public one though, and there are even sites that go so far as to really scrutinize and analyze the individual projects being developed. The problem though is that it's really hard to sift through hundreds of screenshots even if they are dumped on the same page. I just wanted to find a way to weight some of them more than others.

The biggest and I mean biggest point of this is to put the screenshots themselves on stage. If you play the game you are guaranteed to never see a duplicate screenshot at least on the left hand side until you run through all of them from the week.

I don't know... I like regular IOTD. Isn't there some way of fixing it up or promoting it more?

The twitter thing is fine but I personally don't do twitter, don't have any desire to start, and the people posting in the #screenshotsaturday feed are not www,gamedev.net users so what's the point of aggregating their content here? Also don't know a lot about twitter but isn't the popularity of various hash tags extremely fickle? What if screenshot saturdays is a fad that everybody forgets about in a month or that only the same few people keep up with week after week?

The problem though is that the submission rate isn't there with our IOTD.. It's hard to compare 200+ screenshots per week of games in development versus the 1-2 we get per week. And the submissions we do get include a lot of commercial middleware rather than just regular joes making games. Clearly something isn't working. Using twitter though creates a strong feedback loop and that's something we need. What I added yesterday was the ability to add our own comments as a community to screenshots.. which is good.

I think we're in an age where game developers and twitter go pretty hand in hand. The game development community is *incredibly* active on twitter.

But in this case, #screenshotsaturday has been around for years and years and is actually growing in popularity. What I think I might do for users who don't have twitter is to post a link to their screenshots on our own twitter account and reference #screenshotsaturday. I'm still thinking about this.

> EAT ROPE, MMO hockey, MMO in HTML with variables, MP3 beating compression, various Boolean comics (especially asking medical advice on the forums), GDNet: The Trading Card Game... I'm just listing all the memes and often referred to threads/events that I can remember...

[quote name='Drew_Benton' timestamp='1358157469' post='5021328']
Anyways, Michael, I started clicking through the showdown and I don't particularly like the format. The best explanation I have to why I feel that way comes in the form of a question: "Do the screenshot necessarily need to compete with each other, or should a screenshot only compete with itself to the viewer?"
[/quote]

For me I think the format suffers when you have pretty uninteresting screenshots actively competing in matchups.. perhaps I can do a hybrid approach that also integrates up-down voting as I still want to at least allow the showdown to run as is for a bit. Or make it necessary that a screenshot receives at least one upvote to be entered into the matchups.

What I wonder is if there is some algorithm that could be written to determine if a screenshot might be interesting.. or at least have the potential to be interesting.

out of curiosity, could you use re-tweet count's to determine interest(although i'm not sure how much screenshotsaturday images are RT'd.

Good question. You could.. I just don't know how big of a percentage of images are retweeted. Though it could be enough to automatically have them enter into the matchup system. I guess after 1500 or so matchups that I've done I'm tired of seeing mostly black screens with a few simple sprites rendered onto them. When you see screenies that show stuff like this I find it pretty cool: http://www.gamedev.net/page/showdown/view.html/_/runevisioncom-r397

I was wondering if a histogram of the image could be used to determine much.. Looking at the bottom screenshots I do notice that there is a tendency to have very unsaturated colors, possibly use a very limited color palette, and contain relatively few image features of note.

http://www.gamedev.net/page/showdown/view.html/_/edg3za-r314

This isn't an exact science since this is purely aesthetics... I just need to be able to come up with some metric that quickly finds if an image is unlikely to suck. Maybe I can at least filter out some low hanging fruit.

On another path I have retweets of the image, the potential Klout score of the poster, the poster's number of followers, the ranks of all the author's previously submitted screenshots..