Median Price Not One

Maybe I just don't understand how its obtained, but it has always bothered me that the median prices listed for just about every item are wrong.

A median, as defined by wikipedia is: "...the number separating the higher half of a sample, a population, or a probability distribution, from the lower half."

This would lead me to conclude that the number equal distance between the highest recorded sale and the lowest recorded sale in the 25 listed sale histories of each item would be the 'median' price.

This is often never the case.

Take the neckchopper, on Odin. Highest sell in history: 48,000. The lowest sell: 39,000. A derived median would be 43,500, not the 40,300 price that's listed.

I would think the algorithm for finding a median would be simple to write, maybe I'm wrong (I'm no programmer). Anyway, I have just always found this annoying and very 'buggy' so I thought I'd bring it up.

I think you are misunderstanding the meaning of median. It's the number in the middle of a distribution of numbers.

For example, if the numbers were:
13, 15, 22, 25, 27, 31, 99

The median is 25, because it's the number in the middle of the group. 25 gives a more accurate representation of normal numbers, since the average is 33.

It is probably not perfect but it protects incorrect numbers when a person accidentally buys a Yagudo Drink for 30k or RMTs skew prices.

The median price is not taken from the last 25 transactions. It is taken from the last X number of days.

In addition, to avoid completely crushing the database every time someone loads an item page, the median is only computed once per day ahead of time, and therefore does not reflect any sales made between the last calculation pulse and now.

Bumping this because I think something is definitely broke with it. Median seems to always be the second most recent sale in my tests. I made a thread here....

http://www.ffxiah.com/forum/?topic_id=917&lp=1#5427

Yup, appeared to be bugged on the individual item pages but not on the rest of the site computations. Scragg says he's fixed it. Thanks for reporting -- subtle but nasty bug. =)