Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
IMustashU AQuestion
State War Academy Caldari State
1
|
Posted - 2015.04.14 15:12:28 -
[1] - Quote
I have written a python script to pull crest market data in attempt to discover good items to trade in a more scientific way than just going through them one by one and eyeballing it.
My aim is to "score" all the items based off of profit, margin, volume and competition.
The current formula i am using is:
if sevenDayAvgVolume >= 2: score = ((pureProfit*profitMargin)*sevenDayAvgVolume) else: score = 0
I am ignoring anything with a avg daily volume of less than 2 mostly to filter out offcer stuff and other items that never get traded but have massive margins as i have no interest in attempting to trade them as they would lock up too much isk.
I am currently trying to find a good way to measure how competitive a market is and score lower competitive items higher as it dose not matter how much profit in a item their is if you have 2+ people constantly 0.01 isking your never going to be able to buy or sell them.
I would appreciate any input, i do plan on releasing the tool eventually but currently its just a cli python script. |
Koniforous
Tauren Transit
223
|
Posted - 2015.04.14 17:10:59 -
[2] - Quote
To determine competitiveness develop a formula that compares daily order completions to overall orders close to the 7day average. Do this for only buy or sell, however, labeling the item as competitive for one or the other individually.
TAUTX: Private Bank and Lending
TAUTX: Collateral Liquidation / Discount Items
|
Shakuul
Infinitus Sapientia
1
|
Posted - 2015.04.14 19:24:16 -
[3] - Quote
score = ((pureProfit*profitMargin)*sevenDayAvgVolume)*%ofTimeAsLowestOrder
%ofTimeAsLowestOrder is going to be a function of how often your competitors' orders are being updated and how often you want to update your own.
One quick way to figure this out would be the following: - count orders updated in the past day, call this N - define a number of times you're willing to update per day, call this U
your time as the lowest order is going to be roughly U / (U + N), so if there were 8 order updates in the past 24 hours and you plan to update twice per day, you'll have about 20% of the volume. This of course assumes order updates are uncorrelated, have a uniform distribution over time, and that there arent a low number of competitors updating many many times per day.
another approach might be to look at time between orders. So suppose there is one order updated 6h ago and another updated 5h30mins ago. Under this approach, you'd assume your order stays the lowest for 30mins before it is undercut, so updating twice a day you would capture 1/24= 4% of daily volume.
my guess is the second method is going to be more accurate, but try them for yourself! |
Cista2
Phoibe Enterprises
77
|
Posted - 2015.04.14 20:26:18 -
[4] - Quote
Nice idea...... and in fact I used to use the same formula, on my many paper notes.
Problem 1 is that such a formula is only descriptive, not predictive. If price drops with 15%, your profit is gone.
Other problems are that competitiveness and weekly turnover are inversely related, so you have more competition and less profit on those with high turnover. Since you cannot describe competitiveness as objectively as you can the other factors, it follows that whatever value you assign to that variable is what decides the outcome of your formula.
And both these two factors (competition and turnover) are again related to the price fluctuations, as are the profit margins. 0.01 iskers and true traders float from item to item on these waves.
You can say that the actions of traders work towards making every item equally profitable over time, while the price fluctuations that are independent of traders are making some items much more profitable than others. This all is just again to say that a formula that paints a picture of a moment in time, cannot really predict if this is the item you should buy today to sell later today or next week. But it can give you some hints. |
IMustashU AQuestion
State War Academy Caldari State
1
|
Posted - 2015.04.14 21:35:15 -
[5] - Quote
Shakuulanother wrote: approach might be to look at time between orders. So suppose there is one order updated 6h ago and another updated 5h30mins ago. Under this approach, you'd assume your order stays the lowest for 30mins before it is undercut, so updating twice a day you would capture 1/24= 4% of daily volume.
my guess is the second method is going to be more accurate, but try them for yourself!
Ive also been thing of a variation of this idea take the most recent X% of orders and then find the time average time between issued and now (note issued date time is reset when an order is updated) and then use that as a multiplier in the score.
I will try both and see as their not complex to program.
Cista2 wrote:Problem 1 is that such a formula is only descriptive, not predictive. If price drops with 15%, your profit is gone.
This is something im acutely aware of my aim was to create a computationally light weight indicator that basically says look at these items first, so that i don't have to go through 100's of items manually.
I did think about trying to build some kind of predictor but im not a statistician or economist and the only ones i really know much about are candlestick graphs and bollinger bands if anyone is more educated in these areas and wants to work with a programmer id be happy to team up on a project.
Cista2 wrote:Other problems are that competitiveness and weekly turnover are inversely related, so you have more competition and less profit on those with high turnover.
I think and could be wrong that although higher profit and margin attracts competitiveness they are inextricably linked and so some items will have more competition and some will have less and the trick is to find the ones that have less.
Cista2 wrote:Since you cannot describe competitiveness as objectively as you can the other factors, it follows that whatever value you assign to that variable is what decides the outcome of your formula. .
This is a very good point originally i had thought i would just experiment with weighting it but now thinking about it their is no reason why it has to be part of the formula itself but instead have a separate score for competitiveness again used as an indicator.
I'm not trying to build something that will be the work for me just to give me more information to make better decision on what to invest in. |
Bellamy Flynn
Hole Truth
0
|
Posted - 2015.04.17 15:01:28 -
[6] - Quote
I love this type of project but last time I tried to work out a formula was before crest and Eve Central didn't have the volume information I was interested in. I would be very interested to see the source for the data pull portion because I find the crest documentation kind of lacking.
Anybody have a good example of a crest import or could point me to a good resource? It's such a shame to have so much data available to process when I can't figure out how to import it.
|
IMustashU AQuestion
State War Academy Caldari State
2
|
Posted - 2015.04.17 15:09:58 -
[7] - Quote
Bellamy Flynn wrote:I love this type of project but last time I tried to work out a formula was before crest and Eve Central didn't have the volume information I was interested in. I would be very interested to see the source for the data pull portion because I find the crest documentation kind of lacking. Anybody have a good example of a crest import or could point me to a good resource? It's such a shame to have so much data available to process when I can't figure out how to import it.
I use the python library pycrest, i found the docs to be verry lacking as well your best best is to go read up in the techlab https://forums.eveonline.com/default.aspx?g=topics&f=263 |
Cixi
Hedion University Amarr Empire
4
|
Posted - 2015.04.17 17:35:49 -
[8] - Quote
Just an idea for you, if you want to score an item for trading, you need to know how the competition is going, and this is something hard to do with the information we have at our disposition (eg: there is 10 B/O, but is it 10 traders or 1 trader with 10 B/O?)
I will be watching this thread anyway |
IMustashU AQuestion
State War Academy Caldari State
3
|
Posted - 2015.04.17 17:58:01 -
[9] - Quote
Cixi wrote:Just an idea for you, if you want to score an item for trading, you need to know how the competition is going, and this is something hard to do with the information we have at our disposition (eg: there is 10 B/O, but is it 10 traders or 1 trader with 10 B/O?) I will be watching this thread anyway
Its not possible to tell who orders belong to without filling them then checking your journal, I believe this is by design. |
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |