New Reflexive Website

Discussion in 'Feedback Requests' started by Skinflint, Apr 28, 2006.

  1. Martoon

    Original Member

    Joined:
    Nov 5, 2004
    Messages:
    329
    Likes Received:
    0
    Sorry to necro an old thread, but I'm curious about how Reflexive's recommendation system works (I've had an interest in these systems and how they relate to games for several years).

    I'm assuming it's using some kind of collaborative filtering, where my profile is used to find "best-fit" matches with other profiles, then games that average high ratings amongst those profiles are recommended to me.

    I'm not asking for all of the indepth details, just a rough overview of how you calculate recommendations, and how effective you've found it to be. If you don't mind sharing (and I understand if you'd rather not).
     
  2. Skinflint

    Original Member

    Joined:
    May 26, 2005
    Messages:
    60
    Likes Received:
    0
    Currently, the system uses correlative sales data based on game preferences that you provide. When the player rates a game, the system sorts the games that it presents to you based on the games that it has and how the sales of other games conforms to these preferences.

    For example, if you love Alien Shooter, the system will look and see what people that bought that game bought also. It will then bump those games higher on your list to look at. Different game ratings will bump them up or down and to what amount. The list also takes into account how many other games of that type you like too.

    So, if you love Alien Shooter, but that's more of an anomoly and you typically love match-3 games, and rate a bunch of other match-3 games as "you love them" then the system should present you with more match-3 choices than it does shooters.

    Of course, the system isn't perfect...just because the sales data has a high correlation, it doesn't mean that you'll like the game. That's why we are going to be moving towards a ratings correlation system soon. Our new site has given us a TON of ratings for games, and now we can start using these to further help our players find the games that other people like too, not just buy. We think that, even if people don't buy a game, that what they play and love is pretty darn important. However, what they are willing to plunk down $ on is one of the most important votes of all...
     
  3. Martoon

    Original Member

    Joined:
    Nov 5, 2004
    Messages:
    329
    Likes Received:
    0
    Interesting. Thanks for the info.

    Now that you have a large pool of profiles with their ratings and/or sales data, you should be able to experiment with new ratings formulas by running simulations with the data to see how effective proposed formulas might be. For example, taking games that a user has already rated or bought, and seeing how high of a recommendation the proposed formula would give that game for that player. I think it would be interesting to run some genetic fuzzy-logic algorithms in the simulations to see if it discovers some formulas that you might not conceive explicitly.

    You make a good point about the most important thing being how likely the user will be to actually pay for a specific game. There are some games that I might try then give a fairly high rating, but not be as likely to buy. It could be because I don't see long-term play value (but it was really fun at first), or there's too many similar games from other vendors that I also like.
     

Share This Page

  • About Indie Gamer

    When the original Dexterity Forums closed in 2004, Indie Gamer was born and a diverse community has grown out of a passion for creating great games. Here you will find over 10 years of in-depth discussion on game design, the business of game development, and marketing/sales. Indie Gamer also provides a friendly place to meet up with other Developers, Artists, Composers and Writers.
  • Buy us a beer!

    Indie Gamer is delicately held together by a single poor bastard who thankfully gets help from various community volunteers. If you frequent this site or have found value in something you've learned here, help keep the site running by donating a few dollars (for beer of course)!

    Sure, I'll Buy You a Beer