Being the clan’s auction bish on Alleria1, my primary role is to pretty much not move anywhere but the AH, bank, and mailbox, with occasional forays out of the area to gain experience. But, mostly the AH and bank thing.

So far I’ve been following a fairly basic system that has kept food on the table and beer in the fridge.  Right now, everyone left on Alleria has 10K gold, after Grimm cleaned us out on he way to Azuremyst.

A lovely new tool

Now, in the past few months, The Undermine Journal has returned to life, and, being a dutiful auction bish, I decided to check it out. Raw data is a beautiful thing, so when he added an XML feed of his regular data, I jumped at the chance to do some research.  Using a little Python code here and there, Illume was able to automate the collection and extraction of Glyph data, which she and I pored over.

Let me back up just a little bit.  A big problem with the AH prices has always been the so-called "goblins" posting glyphs for ridiculous amounts. Lovely tools like TUJ can’t really extract much in the way of at-a-glance data as to what glyphs are currently the best to post. Furthermore, they tend to do silly things like undercutting to stay at the ‘top’ of the listings when one is browsing for glyphs to buy.  Behavior like this tends to skew market prices to significantly meaningless levels unless you know how to filter.

I don’t know how to filter.  Well, I do, but I can’t.

The simplest way to filter is to watch what prices that items are actually selling for. Unfortunately, TUJ does not have that data in the XML feed, and you can’t get that from Blizz, either.

The next best way is to use statistics.  I like to take the standard deviation of the average of a price over a given time, and then add 2 times the standard deviation above and below that2. Any prices outside of the band thus established should be considered an error and not trusted.

With over 400 glyph recipes out there, this screams for automation. Unfortunately, I haven’t got access to the data I need to make those tools yet, so as of right now, statistical analysis of post prices is out of the question.

Back up and punt

So, if the main problem is that there are a number of unreliably high prices for glyphs cluttering up my lovely data feed, what can I do about it? I considered filtering based on an exception filter (as erroneous prices showed, add the glyph to a list of exceptions that will never get listed) but the permanence of that solution precluded some random glyph suddenly coming into favor.

How about turning the game on its ear?

I have a new experiment going on now, and after a couple of weeks the results are encouraging, at least with the local glyph goblins on Alleria.

Basically, I am accepting the TUJ feed at face value – with moderate sanity checking – but letting the practice of undercutting work for me.

I always undercut, as I have said in the past. So if a glyph is posted for 250g, I will generally post for 5 to ten percent lower.  I’ve enhanced that a bit. Now, if they undercut me, I’ll cancel and repost at undercut prices.

There are two possible effects:

  1. Statistically, prices will get dragged down until they either sell, or bottom out.  Regardless, this reduces the false positives.
  2. I sell some glyphs for some really good prices.

So far, it’s been mixed – that is to say, I am seeing both taking place.  Some glyphs are dropping, leaving the good prices at the top, and some are actually selling for these ridiculous prices.

AuctionLite, you served me well

One thing I had to do was let go of AuctionLite.  While it has served me well for quite some time, it lacks an easy way to (a) see which of your auctions are being undercut, (b) and let you cancel them right there.  I had to flip back and forth between various tabs, very inefficient.

Fortunately, Auctionator had my back. This addon recently came to my attention as a potential replacement for AuctionLite, but since I didn’t need to replace it, I didn’t pay much attention.  Once I started looking around for something to streamline the cancel cycle, though, it came back to mind.

And, honestly, I don’t even know if Auctioneer (AKA AUC-suite) does this as well as Auctionator does, but one thing is for sure – if it did, it would do so at the expense of a lot more memory.

The dawn of a new age

Thus tasked and thus armed, I set out a couple of weeks ago to try out my new strategy. 

So far, it’s been very rewarding.  False positives are decreasing, throughput is up, and profits are up.   Glyphs that rarely sold before, sell quite frequently now, and for consistently higher prices.  On the flip side, some glyphs that traditionally brought in more, are taking a hit because I am no longer posting and letting them sit – and thus, the final price is usually below what I would have otherwise gotten.

Still, the tradeoffs are well worth it. I would say I have doubled our income in this short period.  It’s possible this is a fluke, or that the goblins will decide on some retribution, but I really doubt I figure on their radars.

The next part of this experiment will be when Grimm applies this body of work to Azuremyst. He has recently dropped Mining and picked up Inscription, and is in the process of skilling up. Once he tops out, we’ll see how well this process carries over.


  1. It still sounds so odd to add that server disclaimer! []
  2. This is often called a ‘two-sigma band’ in engineering circles. []
Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Please leave these two fields as-is:

Protected by Invisible Defender. Showed 403 to 35,798 bad guys.

Notify me of followup comments via e-mail. You can also subscribe without commenting.