Thursday 14 August 2014

Market Price History Accuracy

Market Price History is not an accurate indication of what things actually sell for.

For example here's a screenshot of the market history for PLEX in Domain.


Note the low prices on the dates 08-13 and 08-10 are 760 mil and 763 mil respectively. Yet when I look at my transaction log I can see I bought 5 on each those same dates for only 745 mil and 745.5 mil each.


Obviously price history is not a very accurate indication of what things actually sell for. So what is it then?

Market history does not show the prices things actually sell for. It seems to miss a lot of lower priced transactions (like my 10 PLEX above) which inflates the numbers meaning the prices things actually move at are often lower than market history indicates.

Maybe market history only uses sell orders without taking buy orders into account at all. That would certainly explain why my purchases (usually made with buy orders) are often lower than what market history says is the lowest price.

But that doesn't explain other inconsistencies in the history data, like how I sold 5 PLEX for 784 mil each on 08-13 when market history says the high price for that day was slightly less than that.

I know for a fact that on the 13th there were a whole lot of sell orders listed higher than 784 mil, many of them much higher so history can't be just taking periodic snapshots of listed sell orders as I first thought. That'd push the average way up.

Then again the average prices shown in my screenshots above are lower than I would have expected unless it IS taking buy orders into account too. But in that case why didn't my buy order purchases show up in the data?

Whatever the cause may be something funny is going on with market history. The problem is market history prices are often misleading, sometimes very misleading, especially for low volume items. In a worst case scenario, market history doesn't reflect anything near the real prices an item normally moves at but instead shows extremely overpriced listings that for the most part are NOT SELLING.

What I'd find a whole lot more useful than this is information on the prices things actually sell for. I'd much rather have accurate info on the prices (with volume numbers) something actually sold for than what someone listed it for.

On a related but slightly different note, in the case of PLEX I'd really like to see information on two specific things

1. How many PLEX were bought from CCP with real money.
2. How many PLEX were actually used (taken out of circulation and used for subs, transfers, etc).

Even yearly data on PLEX purchases and usage would be extremely useful though monthly data would be much better. No need to get fancy with in game statistics, regular reports or anything, a simple one time dev blog post containing the information would be just fine.


10 comments:

  1. I thought market history was for items sold not items sitting on market? If so... i've been wrong all these years.

    ReplyDelete
    Replies
    1. That's hard to say for sure. Sometimes it seems like it must be taking unsold listings into account too and other times not.

      It's like there's some weird algorithm messing up the data, tossing out data that should have been taken into account and using data that shouldn't have been used. Possibly intended to remove outliers and make manipulation more difficult but not doing a very good job of it.

      Delete
  2. The market history clips the min/max data. Try a simple experiment where you put up a sell order for a common item (such as ammo) at a ridiculously high price at a station without any other orders. Buy the item with an alt. Check the history in a day or two - you usually will not see that transaction listed as the max price for that day.

    Works sort of like how a teacher might bias the grading curve, by removing the lowest and highest scores on a test.

    Some players have said that Dr.EyjoG also used to edit the history, when the data did not look right to him. I guess that changing the data to match your theory is a standard practice in economic research.

    ReplyDelete
    Replies
    1. I can understand clipping extreme outliers (ridiculously high or low) but this is distorting the data.

      My PLEX purchases were within a few percentage points of both the average and median price. Not only that but since they were bough right on Amarr, that means all the higher priced station range buy orders had to get filled first... It wasn't just some guy not paying attention and thinking he was getting a higher price from out of range orders.

      In fact if Mr EyjoG would bother to fix the order filling code so it doesn't rip people off when they typo that'd go a long ways towards preventing most of the ridiculously low / high transactions from even happening in the first place.

      Delete
    2. This isn't just a one time thing either, I see it a lot

      Just off the top of my head I can say I used to buy a lot of Caldari Navy Adaptive Invulnerability Fields. There was a period of at least a month (probably more like 2-3 months) where CN AIFs were around 300 million (some even less) in *sell* orders. I was buying them on several of the main hubs for even less than that. As low as 220 million in multiples on the main hubs (not counting even cheaper singles on remote stations).

      Yet according to market history the price of CN AIFs on those hubs has never been under 300 million in the last year.

      That's complete bull. Market history distorted the data. It isn't an accurate reflection of what actually happened in the CN AIF market. Not even close.

      Delete
    3. Most items actually have a price range they tend sell in.

      Prices cycle through that range over time, with different cycle times for different items. Higher priced and lower volume items tend to have larger price ranges and slower cycle times between highs and lows. Lower priced and higher volume items tend to have smaller price ranges and cycle more quickly between highs and lows.

      When prices are at the high end of the range, the market starts to flood, eventually driving prices back down. When prices are at the low end of the range supply starts to dry up, eventually driving prices back up.

      The problem with market history is it isn't just clipping outliers, it's distorting reality by clipping the normal cycle.

      Delete
  3. This comment has been removed by the author.

    ReplyDelete
  4. The market history excludes some outliers from the price range. For example, if you sell a piece of ammo for 1b isk it will not display that as it's vastly outside the confidence intervals. The highest price accounts for the usual range that most buy orders actually sold at, and the lowest price is the lowest buy order (within reasonable interval) fullfiled.

    I'm pretty the 'highest' and 'lowest' prices are also just an average of the few highest values within that confidence interval.

    Perhaps the history manipulation you speak of has something to do with the way the system decides to ignore some of the highest priced offers or simply calculate averages between the few highest ones.

    ReplyDelete
  5. Hey mate
    what is the best way to contact you?
    we would like to talk to you about economy design for the upcoming Naval Action age of sail sandbox
    www.navalaction.com

    ReplyDelete
    Replies
    1. Send an in game mail to Moxnix Induli with more details about what you have in mind and we can arrange something from there.

      Delete