# Profitability Ranking (basic markets)

1. Jul 26, 2010

### oddjobmj

Hello!

I've come here today seeking information about how I can build a ranking system that will sort through some data I have gathered about a particular market and then rank items in order of profitability using this data. For those interested in the data-set it's from a game called EVE Online. My intent is to learn, but to do this I've challenged myself in this game with a project.

To understand the problem I'll give a little information about the game. In this game there is a market where various items are bought and sold. The items might be ammunition, weapons for your ship, armor, commodities, space station parts, brain implants, etc. These items are build-able in the game. What I've done is collected market data using numerous automated tools. The data I collect is the # of units bought each day, the average price these are bought at, and the cost to manufacture this item.

So, it is simple for me to write up a script that tells me the % profit because I have the cost and market value to work with. However, items with a higher profit % are not always profitable to actually manufacture and sell simply because the market volume for a particular item might be very low. It wouldn't help me to build an item at a projected 500% profit if it is a stagnant asset that no one wants to buy.

At this point I have an equation set up something like:
(% profit)(market value)(market volume)/10,000,000

Of course, dividing by 10 million simply makes the 'rank' a more manageable number. I also know this is completely insufficient... It results in a moderately fair ranking system, but it is all too common for an item to show up high on the list simply because it costs 500 times more than another item even though only two units sell in a day.

I've looked up ranking systems and most of what I read about were pages describing Bayesian ranking where the most common examples were of simple ranking of user opinions about a certain item like a movie where the modification to this data was that items with lower votes remained closer to the average rank of all the other items. On the other hand, items with a higher # of votes progressively move towards the average rank of the votes that particular item actually received. I hoped to treat my market volume like the # of votes, but realized this is unfair because not all of the items have the same market value/cost.

Before I finish I want to say that I can imagine that the basic idea here could be discussed in extreme depth. Instead of forming a perfect equation to predict profitability I simply want two things. 1) To better understanding ranking systems with regard to markets AND 2) Come up with a 'fair' ranking system for this particular project so that with limited human intervention we can expect that the items on the top of the list should be considered for production first.

Please feel free to post any questions.

I greatly appreciate your time and effort.

Thank you,
Mark W.

2. Jul 28, 2010

### oddjobmj

I think I've made some progress.

I'm trying to modify an equation I found on the IMDb Top 250 movies list.

http://www.imdb.com/chart/top

Taking a small portion of this I'll modify it to suite my needs.
So, I'll replace v with s and apply the units sold to that variable.
The issue here is that our 'm' our minimum units sold for consideration isn't going to be a flat rate throughout my list like it is on the movie list. Mine has to change dynamically based on the original cost of building a unit of each object. i.e. Objects with a high value will require a much smaller quantity sold and objects with a low value will require many more units sold to be considered. I figure the best way to do this is to pick a large number and divide it by the value of an object. Here are examples of two extremes that will show up on our viewer:

High-value Example:
Low-value Example:
This is closer, but still the 'm' value for my high example is a little too low and the 'm' value for my low example is too high. How can I bring these values closer together? Moving the initial 10 Billion value lower or higher pushes one of the variables in the wrong direction while bringing the second closer to the desired...

Also, any suggestions on modifying the second part of the original imdb equations?

Thanks,
Mark W.

3. Aug 1, 2010

### oddjobmj

Can anyone help me or even explain why I haven't yet had a response? Maybe let me know if I worded it poorly or didn't provide essential information.

Thanks,
Mark W.

4. Aug 1, 2010

### CRGreathouse

I think your underlying assumptions are wrong: the naive profit model you reject is, in fact, more appropriate. Of course you don't want to overproduce on the low-demand items, but it's not hard to figure out a good number to make. Basically, you should strive to have inventory in stock for the k most profitable items, where k is taken as large as possible.