[sinanspd]

Yeah the forum has a certain number of problems. As you said it isn't anything to do with the staff but more fundamental stuff. One being, the software isn't fully controlled by the staff, I am sure this is built on an existing forum engine. The second came from the fact that this was built very early on. If you look at some of the very early posts, there are a lot of posts that don't belong in the section they are in. The forum opened new boards as the hobby evolved but never rearranged and moved the older stuff.

But perhaps the biggest issue I have is thale fact that images aren't hosted by the forum. A lot of the image hosting services that were used early on went out of business and now those images are not accessible, the members are no longer active, making a good 30-40%, maybe more of the posts here rather useless. For example, one person posted about a release event for TLOU they went to in Argentina back in the day and posted the items they got there. The pics are gone so is he and I have no idea what they were.

Following that is the lack of a proper search functionality. Any time I need to find an older thread I just google the keywords along with collectorsedition.org to find what I am looking for.


In terms of the amount of work needed.

1) It is very important to highlight that this is never meant to be a comprehensive catalog of all games, not even all AAA games. Even if an expert volunteer steps forward, I will be very selective in terms of what I let through and what not. In the Request section for example I make it clear that I am staying away from Japanese studios with a few exceptions like From Soft. Companies like Square Enix, I am convinced have released more merch than the number of lines of code in their games, and I just can't be fucked with. It is neither worth running the core algorithm for something like that nor worth hosting all those images. Same goes for Ubisoft.


2) for everything else, the algorithm does a pretty good so far. 3 games advertised + 3 others I haven't put up yet (for reasons I will explain below) it had very good accuracy. Even for a very large title with 1K + items like last of us, if I remember correctly it returned 40-50 "unsure" items it couldn't determine the origin of. I was able to manually track down 13 of them. The rest even I couldn't find anything definitive on. I will put those under a "help needed" section later and try to crowd source the info. Most of the work I had to do, came from work I created for myself. Adding measurements of items, shooting videos and pics etc.

I can give a brief explanation of how this works:.

I built a company a while back with the Grammy winning musician Chris Adler and Jason, who used to run digital marketing of Sony music. The platform essentially micro optimizes a musician's career using data from the entire industry as a whole, to stay ahead of trends etc. It also evaluates audience response by constantly monitoring social media, forums etc. and performing sentiment analysis and topic association.

In addition, it constantly scans sites like Etsy, red bubble and other "on demand" bootleg sites to automatically issue cease and desist notices for copy right infringement.

The core of the system is pretty much borrowed from those systems. I combined it with a fake news/quote misattribution/deep fake detection algorithm I build back in 2015. (it had 92% accuracy in 2016 elections and was adapted by all the big socials at the time) to do another step of fine grained filtering and to deal with things those automated bot accounts that try to dump a scammer shirt site on Twitter as soon as you utter the words "where can I get this?" Or "that's a cool design".

The video part is still experimental. It searches YouTube for expo walkthroughs and I also compiled and maintain a comprehensive list of influencers from all regions so their socials and channels are constantly scanned for new content. And this was the reasoning behind not including anything pre 2010. Videos shot before that simply doesn't have enough resolution for accurate results from our detection algorithm.

3) I was GDC last month, and I sat down with quite a few companies. Can't announce who yet but you will see it soon enough. We are working on partnerships which allows me to get an already compiled list of everything they printed then I only have to deal with the regional stuff local publishers made. I will be adapting the algorithm we use for the musicians for them, and using the data of how many views something gets, how many people add to wishlist. I will help them adapt their merchandising strategy to speak better to their customer base. All the company acknowledged their existing merch sucks and sometimes sits in the warehouse for years so after seeing the results we got in the music industry, they were excited to get onboard.