Advertisement

Making/Money: Auctioneer Wishlist

WARNING - MATH AHEAD!

A few weeks ago, a reader e-mailed me with questions about some Auctioneer data that they have been collecting. Though the question was regarding trying to determine inflation patterns on their server, it got me to thinking about what Auctioneer actually captures. And what other data points would be most helpful in making evaluations of item pricing or looking into economic trending.

While I'm not about to go through all of Auctioneer's algorithm rules, I would recommend that anyone who uses this mod check them out on Norganna's Wiki. They explain where the recommended prices come from. Basically, it's a comparison of the historical data accumulated from all of your auction house scans and the most recent snapshot data. The market value ends up being either the median of historical or recent data, flavored with the user-set preferences for markup or markdown.

While using the median values grants the data some of the benefits of ignoring outlier values, it relies on there being enough data points to represent the full range of possibilities. This causes a paradox. On the one hand, we would want to have lots of data points so that the median really does indicate a central value (stats-y wish, clearly). On the other, older data may no longer reflect newer market conditions. And it doesn't take into consideration the greed factor.

Let's face it - when using Auctioneer it's generally to ensure that we're getting a fair price for our listings. Of course, we try to price to sell... but still want as much as we can get (Yay Capitalism). So will following recommendations from Auctioneer get you all that the market is willing to pay? Probably not.

The median will give us the midpoint of whatever data we have collected. This can be skewed if there are not many scans of the AH if, for instance, one seller posts twenty stacks of Peacebloom at the same time for the same price and we capture data on all of the listings. No distinction is made between prices set by a single seller or a group of sellers.

A potential means of getting around this is to only pull one price per item per seller per scan. In this way, you would still have data on when the price was what it was and would be able to weight sellers equally regardless of how many items they had posted.

To counter getting outdated recommendations, it would be helpful if Auctioneer held data only for a few weeks. By automatically dumping old data and using new scans, you would be fairly certain to come at least close to recent auctions. But this would make logging in routinely just to pull AH scans would be necessary so that you didn't lose all of your data.

Another way to get recency without sacrificing all of the historical data would be to create a more complex algorithm that weights current data more heavily that older data. While it could certainly be done, it may take slightly longer to calculate and store it into the Auctioneer data for it to be available via tooltip. And it may not be what the developers of Auctioneer want to do (not affiliated with them in any way = wouldn't know).

Here comes the truly wishful thinking. It would be rather helpful if Auctioneer could capture indicators of whether an auction was the only one for that item when it was posted. Sellers who enter a market that has no competition do not have to price lower than anyone else to get a sale. If they remain the only auction listed then they have a temporary monopoly. However, if another seller enters the market, they would want to undercut the first auction so as to get the next sale, even if it is at a slightly lower price. But the price they request is usually based on what the current prices are on the AH.

If it were possible to know whether an auction was created in the absence of other auctions, then these prices could be treated differently than the usual algorithmic calculation. For instance, if a monopoly price was well above the median of non-monopoly prices, a penalty might be assigned to it forcing the value towards the median for inclusion in the calculation. If the price is below (maybe this seller doesn't use Auctioneer and therefore had no clue as to what price to set) then it might be assigned an inflator or ignored when calculating the final price.

There are so many other things we could ask for from a mod already as useful and prevalent as Auctioneer. Getting them may be another story. They rely on Blizzard to provide access to data and some of the things that would be really nice to have simply may not be available. Still - it's always nice to imagine a world with more numbers.

What else would you want to see in Auctioneer? Would a more advanced algorithm be helpful? Do you take any other information into account when you set prices on the AH?


Alexis Kassan is a numbers nerd. She spends her days with statistical programs and her nights with spreadsheets and textbooks. She's also a MMORPG addict, having gotten sucked into Ultima Online at a formative age. In her time away from work, books and games, she can usually be found drowning in pools of sprinkles. If you have a question about in-game economics or how crafting fits in with them, hit her up at alexis DOT kassan at weblogsinc DOT com.