Auctioneer has come to the point where its profit calculations could benefit from a statistics/economics major. So we're currently looking for one so that s/he can improve those functions for the 4.0 release of Auctioneer. Such a person needs to have a basic understanding of programming concepts, programming experience is a plus. However the most important qualification is that the person needs to be well versed in stats/econ models and formulas.
If you would be interested in helping out as a volunteer coder, please visit us at our IRC channel http://norganna.org/irc/ or reply to this post. If you reply to this post you do not need to give out your e-mail since we can see your registered e-mail address.


Thanks in Advance,

MentalPower
Lead Developer of the Auctioneer AddOns package.

Great to see Auctioneer moving along so well... I've missed being around! Work is oppressing me from my habits

Keep up the great work guys! Drop me a line if I can help with coding stats from a REAL statistician!

.k.

jtravers88 visit us at our IRC channel http://norganna.org/irc/ so we can talk ideas.

Karavirs, your assistance is always welcome :).

Ndlovu is definitely on the right track. If you haven't found anyone yet, email the department heads of a couple of major University Economics departments. This could be an awesome project for a student. And the bait: this is for an add-on to a game that is played by over 7 million people worldwide. Of course, you could also get such an overwhelming response that your mailbox might get flooded with hopeful students. BTW, I emailed my Economics professor to see if he was interested or knew of any students that might be. Good luck.

I can help. I am business and economics double major.
email me at hand@rpi.edu let me know how i can help

Hop into IRC so we can talk. http://norganna.org/irc/

This is a virtual world, not a real world. Thus the economy is overly simplified. As a simple example: you cant find a cheaper way to make something by substituting cheaper components. There is no "Slott's Mana Potion with extra lime". Thus I believe you are aiming for overkill. I use a handful of algo's, and it works nicely.

I have the education requirements(BA Economics with concentration in computer apps minors in CS and MATH), but I am not too familiar with Lua. Perhaps I can just describe what I believe would make the best use of economic/statistical theory in my opinion. With the addon in the present state each item that pops up has averages for the price and a recommended price to post bid and buyout. Data for the averages is calculated from the times each item each been viewed previously on the AH. However, currently the addon is not using significance figures from a basic regression. It is a fairly simple calculation actually. Using a basic formula you calculate the items presently listed in the AH that have a price significantly below previous prices. Based on the inputs after many scans you can calculate the expected price for an item. As the number of times you have seen the listed item in the past increases the reliability of the expected value rises. This is a basic result of using a linear regression model. Simply use the data saved from previous scans including the number of times the items have been "seen at AH." You just want to run a basic regression. I don't think you should have to get too fancy. And here is why /////(lengthy explanation warning. feel free to skip) //////The AH is not a real AH, hence, the players have no real life loss when items are inadvertently or purposefully listed below the markets equilibrium price. I might consider doing an ARCH (auto regressive conditional heteroskedasticity model) to examine if market volatility is suspect to be a function of time. (A common practice in the real market) However, I suspect that you do not keep track of the time and date that each price is recorded, so there is not enough data for this analysis and furthermore it may be unnecessary due to the fact that there are easier and lower risk options available with more simple analysis. Also, since you do not record the time that prices are listed this throws out pretty much every method of analysis other than the simple cross sectional analysis. (This is why I recommend an OLS regression) The fact is that many players make mistakes when listing items because they list them at a price below the markets demand for that item. (Again due to the nature of the 'virtual' market) This occurs far more often in this virtual market than in the RL market. (I speak from a year of game play and Auctioner use)///end of explanation/// When I use auctioneer I like to make the lowest risk purchases that will produce a profit and spend almost all of my gold. This means that I often look for buyouts that are uncommonly low and buy them up and relist them. This strategy is very effective. However, a simple regression would make this strategy far more effective. When I look at the items sorted by profit I look for items that have been listed many times at a price higher than the current one. If a price is lower than the previous prices and there are only 1 or two times that the item has been listed previously, then it might be that the item was just listed unreasonably high before. In fact, odds are the you will see items listed at unreasonably high prices because they tend to stay on the AH longer. Usually I spend only a few minutes looking for items to buy after scanning because the ones that are truly profitable dissappear quickly. I typically use 30 times listed as a good indicator of the significance of the previous price. This number comes from statistics. You could increase the performance of the addon by actually calculating the significance of a currently listed price below its historical price OR you could easily add a function that sorts though and makes use of the number of times listed without even programming a basic regression. (It doesn't have to be too complicated, but you can make it as complicated as you want it to be.) I might also suggest that you use some sort of grouping based on the characteristics of the items for sale. For example, two different head armors may offer exactly the same stat increases, but have different names and thus be in different groups, but in actuality they should be counted as the same. So thats my input for now. If you program a basic regression on the addon you should be careful to run only the items that have been selected by some criteria such as high profit or more than a certain number of the item listed. There are so many items that running a regression for every single one is not practical computationally. I suggest that you run regressions on items that have a large volume of trade (Seen greater than 30 times) and that have a fairly large percentage of profit. Say 50%. This should limit the number of regressions to about 100 or so. Still no easy computation, but with a computer processor it is beyond reasonable. Please let me know if my suggestions have helped you or if you would like to ask me anything else. I would be glad to help in any way. Please use my registered e-mail address. -Thomas Shields Oh yeah, the value that you should post along side the other numbers listed in the Auctioneer table is the number of standard deviations that the price of the item is below the mean price. You should only list items that have a P value of <.05 In other words, you will only list items that have a normal distribution of prices.