Speaking as a programmer, this is actually very hard. The problem is being able to classify a synthesizer into a definite category based on just a text description. As an example, I recently got a minimoog on ebay, and as many of you probably know, searching for a minimoog isn't that simple. Do a search for "minimoog", all categories. At this very moment, there are five of them for sale, including one in Australia and one keyboardless version, out of 39 results displayed. You also have voyagers, sample cds, tshirts, mugs, vsts, as well as the odd synth where someone throws "minimoog" in the title so you have to look at it when you are really looking for a damn minimoog.Jabberwalky wrote:So when is someone going to create a database system that tracks all sold items on Ebay and can show a graph? If I knew how...I would. One of you programmers could make a lot of money with this
While it's easy for a brain to sort the garbage out from the goods, for a computer this is no easy task. Narrow the search like this:
1. Narrow the search by category "Musical Instruments". 25 results, all 5 minis still there. Narrowing the category more right now doesn't work, as it loses some results. The program will have to search in multiple categories (synthesizers and keyboard instruments), but ebay won't let you do this so I'll ignore it for now.
2. Add "-voyager" to the search. 13 results, all minis remain.
3. Similarly, add in "-sample -shirt -pc -soviet -arturia" and it's down to 7. There's the 5 minis and then a couple of parts. At this point, there's not much more as humans we can do except add more special cases to weed out each individual part. You're already risking cutting out good matches too. If someone writes "this sounds way better than the arturia", then it's extremely difficult for a computer to know it's really a minimoog.
OK, so this is just a longwinded way of saying it's not easy to classify these. But the technology does exist. It's basically just a bayesian spam filter that needs to be specifically trained for each instrument. But you really need a lot of data to properly train them, and then you need somebody to look through every results and be the "oracle" to say which ones are good.
So now you have a synthesizer-auction-classifying algorithm and it's doing a semi-decent job. But how does it use this information? The price of a synth is based on a lot of things. Serial numbers, condition, service record, current popularity, seller's feedback, shipping, and tons more little pieces of information that humans take into account. You also need to look at auction length, buy it now price, number of bids (and possibly how many different bidders there were), if its being relisted, etc. Then there's the ripoff/bargain anomaly factor. There's also lots of cross-correlation between each synthesizer. You could stop before all of this though, and just make some graphs of prices, but there's many correlations in play that would give a more accurate view of price trends.
:blah5: :blah5: :blah5:
Anyways, it's possible, but difficult. This sort of undertaking might better be taken by some stock market prediction firm.