Market Discussions

TLDR:
Statistical analysis showed that for most minerals, daily prices positively correlated with the no. of market orders (for that mineral) i.e. when one goes up the other also goes up, and when one goes down the other also goes down. The positive correlation was especially strong for tritanium and pyerite, moderate for zydrine and megacyte, and weak for mexallon.

The daily price for PLEX, on the other hand, had a weak negative correlation with the number of market orders for PLEX i.e. when one goes up the other may go down a bit. (stop reading now to avoid further brain trauma)

Background
VV's recent thread prompted me to look at some data. This is just a preliminary analysis to get used to handling Eve data.

Methods
Decided to initially look at a 92 day period, starting a month or so after the barge changes: Oct 15 to Nov. 14. As far as I know, there were no major events or patches/changes during this period.

Information from the Market Details Table was copied into Excel. Data was imported into Stata. Frequencies and distributions were studied . Scatter plots used to visualize relations. Pearson's r correlation coefficient was calculated for the continuous variables: no. of market orders; Quantity sold; and daily Price ("Avg."), for each mineral. Same for PLEX.

Did the same calculations for the the last year (to date), to compare with the 92 day data. The last year included all the ups and downs, patches, hulkgeddon, loot drop changes, barge changes etc. I note that this comparison period is not ideal, especially as the last year included the 92 day period being "compared".

Results
No. of Orders & Quantity Sold
For both the 3 month and yearly analysis, for all minerals and for PLEX, there was a strong positive correlation between the number of orders for that item and the quantity sold. For isogen, zydrine and megacyte, the positive correlation was weak to moderate.

Daily Price and No. of Market Orders
For tritanium and pyerite there was a strong positive correlation between daily price (avg.) and the number of market orders. The relation was moderately positive for megacyte and zydrine and weakly positive for mexallon. Weak negative correlations were noted between daily price and number of orders for nocxium and morphite.

Price and Quantity Sold
Moderately positive correlations were found between daily price and quantity sold, for tritanium and pyerite. A weak positive correlation existed for mexallon. These correlations were not seen in the year analysis- if anything very weak negative correlations were noted (see tables below).

I have restricted myself to some basic results. I would be delighted to provide further details.

Discussion
A positive correlation does not imply causality. With the above analysis it is not possibly to ascertain whether increase in the number of orders causes the price to increase, or vice versa. Further analysis may shed light on this.

For all the minerals as well as for PLEX, the very strong positive correlation between no. of market orders and quantity sold is consistent at the 3 month and yearly intervals. It appears that these variables are inter-dependent.

There is no information available on the number of buy and sell orders separately (a total figure for orders is provided in the market details table).

An increase in the number of buy orders is most likely to correlate with an increase in price, whereas an increase in the number of sell orders should have the opposite effect. One may expect these two effects to cancel each other out over time, but it is clear from the analysis that an overall increase in the number of orders for a mineral is related to an increase in price.

This trend is opposite for PLEX, where an increase in the total number of orders is related to a decrease in price (weak negative correlation).

Conclusion:
For minerals, the average price appeared to be positively correlated to the number of orders. The relation between price and quantity for minerals was not as clear, with some minerals showing a mild positive correlation for the selected 3 month period, and an overall negative correlation over the last year.

PLEX average price was neither correlated to the number of orders, nor to the quantity. In fact a weak negative correlation existed for both comparisons at both the 3 month and yearly interval.

Selected Tables
For masochists, with the r values (correlation coefficients), relating:
1. no. of orders
2. quantity sold
3. daily price- "avg"


Trit -92d | orders quant avg
-------------+---------------------------
orders | 1.0000
. quant | 0.5870 1.0000
.... avg | 0.4094 0.3750 1.0000

Trit - yr | orders quant avg
-------------+---------------------------
orders | 1.0000
. quant | 0.3380 1.0000
.... avg | 0.3561 -0.2519 1.0000


Pyerite 92d | orders quant avg
---------------+---------------------------
orders | 1.0000
.quant | 0.5196 1.0000
.... avg | 0.4065 0.3521 1.0000

Pyerite yr | orders quant avg
-------------+---------------------------
orders | 1.0000
.quant | 0.4031 1.0000
.... avg | 0.0747 -0.3769 1.0000


PLEX 92d | orders quant avg
-------------+---------------------------
orders| 1.0000
. quant | 0.9146 1.0000
.... avg | -0.1590 -0.1437 1.0000

PLEX yr | orders quant avg
-------------+---------------------------
orders | 1.0000
. quant | 0.9457 1.0000
.... avg | -0.1801 -0.1974 1.0000

Look forward to doing the analysis on items you recommend VV.

The tables for Mexallon are below.

Strong positive correlation between no. of market orders for Mexallon and quantity sold, at both 3 month and year analysis.

There is no relation between daily price of mexallon and no. of orders at the yearly level,. Weakly positive correlation for the 3 months.

Daily price of mexallon is moderately positively related to quantity sold in the 3 months studied, but no relation exists at the year level (using this method). (see tables at the end)

The table immediately below relates the daily price of mexallon (mxavg) to the daily price of all other minerals during the 3 month period. (this is the nice stuff)

Strong positive correlation can be seen between Mexallon price (mxavg) and pyerite price (pavg) during this 3 month period (oct 15 to nov. 14, 2012) Weak negative correlations of Mexallon price are seen with isogen (iavg), morphite (moavg) and zydrine (zavg) prices.

................. | mxavg
-------------+---------------
mxavg | 1.0000
iavg | -0.2682
mgavg | -0.1624
moavg | -0.2842
navg | -0.1448
pavg | 0.5456
tavg | 0.1196
zavg | -0.2507

The tables for Mexallon relating no of orders to quantity sold and to daily price (numbers are Pearson's r correlation coefficients)
Key: mx = mexallon 92 day study period, mxy = mexallon year period, avg=daily price.

..............| mxorders mxquant mxavg
-------------+---------------------------
mxorders | 1.0000
. mxquant | 0.5829 1.0000
....mxavg | 0.1873 0.2886 1.0000

. pwcorr mxyorders mxyquant mxyavg

...............| mxyord~s mxyquant mxyavg
-------------+---------------------------
mxyorders | 1.0000
.mxyquant | 0.4754 1.0000
....mxyavg | 0.0411 -0.0616 1.0000

I say above that predictive intervals may be generated easily, but that may be a mis-statement. The issue is knowing the variance (or standard deviation).

I looked at the high and low vales that we are given in the market details table (where I got the info from), and these values do not look like standard deviation. They do not vary equally around the mean. So much so that at times the mean (which they call "avg.", so maybe it is not the mean) lies outside the high-low interval!

So what are these high and low values? How do we find the variance?

Edit: Maybe the "avg" is the median (as the median is a type of average). They're possibly using the median, as outliers tend to effect the mean a lot.

Still not sure what high and low represent.

Have you tried this intellectually savory operation: check for the relation between PLEX and Technetium and when. And why.

Hmm, I'm seeing a downtrend starting in the mineral prices. The price of PLEX is also getting nearer 550M over the past few weeks.

Noo, why people don't read my thread. Then they lose money

"Orders" in the Market History table is the number of transactions executed. Since each execution is necessarily the matching of a buy order to a sell order, the figures are identical. The Market History table does not record the number of "standing orders" (buy or sell) that were posted on the market at that date, but only completed transactions.



I understand the high and low to be the actual high and low transactional prices after the removal of outliers using some unknown (to me) trimming method. In this forum (and elsewhere), I have seen speculation about the trimming method, but have not myself seen (from CCP or player testing) anything to confirm the actual method. I would be interested to learn the trimming method.

I have never myself seen any daily "Avg" outside the Low" and "High" (or any supposed "Median" on the chart outside the Donchian channel), and would be interested in seeing an example of a "mean" outside the low and high.

At some point historically (maybe 4 years), the Market History charts in-game were labelled "Median Day Average" but displayed the mean instead of the median. The Market History table said "Avg" and showed the mean also (which is fair there, since mean is indeed a type of average). I tested the error with the chart and bug-reported it, but eventually—after a BH successfully reproduced it and added detail beyond my screenshots—it was flagged "not reproduceable" and was "attached to a defect"... I'm not really sure what that means as far as fixing the error goes, though I never noticed it as fixed in any patch notes.

I believe, though I did not myself test, that even longer ago (e.g., 8 years) the values were medians, but that at some point, they were changed to means because it is computationally simpler (less demanding on the market servers) to keep updating mean throughout each trading day. It's possible that intraday "avg" numbers are a mean, but that historical figures (those for the last full day prior to the latest downtime) are calculated differently from intraday figures and are actually medians, but I've never seen any evidence of that.

An old thread on this sort of topic: http://community.eveonline.com/ingameboard.asp?a=topic&threadID=1226826.

Also: statistics are fun! If you go beyond the basics, keep in mind that commodity prices are not typically normally distributed (not many negative-priced transactions!), so there are some caveats in some methods that assume sphericity.

i'm a miner, it's not like i'm "losing" money on this. i'm just not making as much.
minerals are down, plex is down, the time it takes me to buy a plex in hours is largely unchanged.

*shrug*
not like i can influence mineral prices, if prices are down then they're down. just going to have to accept that and only sell the minimum i need to be able to afford what i need to purchase and wait until prices are more favorable in order to dump the rest of my stock.

as long as i'm liquid enough to purchase a plex for my account i'm fine. other than that i can hold minerals until the end of time. it would seem we're at the "too late" stage anyway, looking at trit's moving average.