I bet you thought we were done.but no, we are just getting started! So far we have determined the current market value of an item based on one scan. It also gets rid of more subtle outliers to determine the average. This method ensures that no poisoning of our market value can take place by those who post high volume items at astronomical prices. In this example, our final market value is 14.5. In this example, it throws out any data points that are not between 7.502 and 18.785 which means the 5 will get thrown out.Īfter step 2, the data set looks like this:įinally, we calculate our current market value by simply taking the average of the remaining data points. In this step, AuctionDB throws out any data points that are more than 1.5 times the standard deviation away from the average. Our data has a standard deviation of 3.761. Now, we find the standard deviation for our data. We now simply take the average of the data that survived step 1. It would not ignore the 13 even though it's more than 20% greater than 5 because it is not yet 15% of the way through the data.Īfter step 1, the data set looks like this: In the example data above, there are no large increases in price between 15 and 30 percent of the way through, but it would totally ignore everything after (but not including) the 16 because that's the last number in the bottom 30% of the data. It will consider at most the lowest 30% of the auctions. After it is through 15% of the auctions, any increase of 20% or more in price from one auction to the next will trigger the algorithm to throw out that auction and any above it. The same can be said for cloth, leather, ore, volatiles, and other items where the quantities required by trade skills are larger.ĪuctionDB uses this to detect more subtle outliers. For example, there is more hypnotic dust on the auction house than greater celestial essences as hypnotic dust is used in much greater quantities than greater celestial essences. This assumption turns out to be pretty accurate. It is assumed that the number of an item that is required on average is proportional to the number currently on the auction house. AuctionDB attempts to factor this into its market value. This is an inherent weakness with most market value estimates.
If you bought the 15 cheapest, you'd pay 16.3 gold per item. If you bought the 5 cheapest auctions in this example, you'd pay 12.2 gold per item.
It is easy to see that the value of the item depends on how many is typically bought at a time. AuctionDB calculates the market value in multiple steps which attempt to correct for outliers, give a moving value over time, and give a much more accurate market value in general than a simple average. Performing a simple average of this data set would give you a market value of 25.79 which is obviously too high. Let's use the following data set as an example (assume these are gold values and each number represents the buyout of a single item): AuctionDB uses an algorithm with a bunch of steps to make it as accurate as possible while using as little memory/disk space to store the data as possible. This page attempts to explain exactly how AuctionDB calculates the market value of items.
0 kommentar(er)
