Not signed in (Sign In)
    • CommentAuthorog
    • CommentTimeDec 29th 2007
     
    In my small brain, one of the biggest differences between Auctioneer Classic and Advanced is the switch from using the median to the average of auctions as the foundation for it's analysis. Out of curiousity, what was the primary reason for this switch? Also why weren't both averages and medians included with Advanced?

    I'm sure this is somewhere in the forums and/or site, but I couldn't find it.

    Og
    Og's Ledger

    •  
      CommentAuthordinesh
    • CommentTimeDec 29th 2007
     
    in my head, the change of stats wasn't one of the primary changes. we could easily add a median-based stat back in at any time. heck, any user could contribute one as well. the change to a framework which allows this was the major change.

    in any case, why means instead of medians? norgs would have to answer, since he wrote the original AA stats. but here are some reasons:
    * We wanted a stat which had a time component, and decided that EMAs would work well. These are mean based statistics.
    * Medians are generally used to minimize the effects of outliers, but in our case, given the small number of data points we stored per item (35), it was fairly easy to abuse and poison them.
    * We were able to come up with some anti-scam algorithms that worked for means.
    • CommentAuthorKinesia
    • CommentTimeDec 30th 2007
     
    Your blog is pretty good Og, nice balanced point of view on things showing both pluses and minuses.
    • CommentAuthorog
    • CommentTimeDec 30th 2007
     
    Dinesh - Thanks for the great response. I figured that the under-the-covers code base was probably one of the biggest changes, but from a end-user perspective what get's displayed is usually what people see as changes. I like the averages. But I like the medians too. It would be nice to not have both AC and AA to get both.

    But I'm not complaining. In RL I've run the gambit of roles in the software development lifecycle and understand the challenges you guys and gals face. And the auctioneer team always steps up to the plate and knocks it out of the park.

    Kinesia - Thanks for the feedback and stopping by. If I can keep an even keel and help people make some extra coin then I'll be happy. Keep me honest on my posts about the Auctioneer suite! After this response by Dinesh I know I have to make at least one update :)

    Og
    Og's Ledger

    • CommentAuthorWyz
    • CommentTimeJul 7th 2008
     
    The problem with mean is that large, outlandish prices have a large effect.

    Say something is posted 5 times at 2.1, 2.3, 2.4, 2.5, and 20. The median price would be a realistic 2.4G, but the mean would be 4.8G. Likewise a few very low prices may skew the mean downwards.

    Good descriptive statistics, and this is what Auctioneer is all about, should report median, mean, interquartile range (1/4-3/4), and sd. Alternatively, report the mean of the 1/4-3/4 range.

    I am not a programmer, and I don't know how hard it is to code. But if it is possible, it will make a great product (which AA surely is) even better.
    •  
      CommentAuthordinesh
    • CommentTimeJul 7th 2008
     
    as mentioned, we have an outlier filter which tries to deal with the problem of large, outlandish prices (at least on the high side). The biggest problem with it is that it doesn't work on day 1 - the first time you scan, we don't know what the current mean is, so we have no way to know that 20 is bogus. on the other hand, if you already knew that 2.3 was your mean, then you got those 5 data points, the outlier filter would throw away that 20 listing.

    also, as I said about, we are not prevented from writing a median, IQ, or any other kind of statistic and plugging it in. No one has had the interest or availablility to date. When someone does, it will get written I am sure.
World of Warcraft™ and Blizzard Entertainment™ are trademarks or registered trademarks of Blizzard Entertainment, Inc. in the U.S. and/or other countries.