These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

Market Discussions

 
  • Topic is locked indefinitely.
 

Predictive Analytics - Let's Tell the Future

Author
Plleasure Hub
Municorn
#1 - 2016-07-23 07:05:23 UTC
Hiya, MD. I hope all you space bears are doing well. Sorry that I haven't offered many exciting posts of late. It looks like you guys have had some real class acts passing through. I thought I'd try to spark an interesting discussion about prediction in EVE markets.

One reason I've been so quiet is due to some awesome stuff I've gotten into IRL. I'm interning as a business intelligence developer for an IRL industrial conglomerate. I'm also in a business analytics course with my favorite professor (entering my last year at uni - woop woop!)

I've been learning some fascinating stuff about data analytics. Apparently, it is entirely possible to predict trends in markets with the right historical data. Prediction is not perfect, but it is damned interesting. Successful companies are using predictive analytics to boost sales, growth, and provide guidance to their business strategies [link].

My question to you lot is this: how would you use the power of prediction to profit or achieve your goals in EVE? Would you consider using a web app offering market predictions to make investment decisions?

If you have any ideas on ways to use this power, please post or message me (if you want a more private convo). I might be taking on an EVE-related project for my spring semester capstone course. Your suggestions could help guide the features of the next EVE 3rd party tool!

Truly yours,
Plleasure

"There's no meaningful difference between a real and a virtual world. It's pointless to ask anyone who they really are. All you can do is accept and believe in them, because whoever they are in your mind, is their true identity." — Kazuto Kirigaya

Vaerah Vahrokha
Vahrokh Consulting
#2 - 2016-07-23 19:05:38 UTC  |  Edited by: Vaerah Vahrokha
You can't predict the markets. But you can do a lot none the less.

Here's how I have made the hundreds (billions) with ease:

1) Learn the game Attention or at least learn a niche like nobody else. For years I tracked and lived the ices markets.
Lived as in, I literally created an ice mining fleet with all sorts of logistics and learned what happens at all the levels, from being a miner to the fleet leader, to organizing hi sec mining / anomaly mining / missions mining (!) (for minerals) / low sec mining, null sec mining. Only place I did not try it are WHs.

Watch the markets, watch the seasonality of demand and supply, look at how price changes throughout the summer, across the years. Look at what happens to supply during 0.0 wars, during events (Hulkageddon days come to mind) and so on.

This lets you learn behavioral analysis first hand.

Same applies to PLEX, really. And to many, many other markets.

2) Create a market model. You can spend your days headbutting 0.01 ISKers or you could pour in an huge position and - using the knowledge learned above - let it sit in the markets and grow it over time.

3) You may leverage on your Excel skills to do 2, you may use real life charting platforms like I did and apply RL markets trading systems adapting them to EvE.

4) Or, if you are the Quant Guy, you may happily analyze the markets, find out that EvE markets have so many juicy inefficiencies and build an algorythm out of them. Backtest it, have some out of sample data to get additional confirmation and then you are set. There are software that can do that, find out what they are Pirate


The 4th but also the 3rd point can sort of help with your research.
InterStellar Architect
InterStellar Architects Corporation
#3 - 2016-07-23 23:47:59 UTC
Plleasure Hub wrote:
Hiya, MD. I hope all you space bears are doing well. Sorry that I haven't offered many exciting posts of late. It looks like you guys have had some real class acts passing through. I thought I'd try to spark an interesting discussion about prediction in EVE markets.

One reason I've been so quiet is due to some awesome stuff I've gotten into IRL. I'm interning as a business intelligence developer for an IRL industrial conglomerate. I'm also in a business analytics course with my favorite professor (entering my last year at uni - woop woop!)

I've been learning some fascinating stuff about data analytics. Apparently, it is entirely possible to predict trends in markets with the right historical data. Prediction is not perfect, but it is damned interesting. Successful companies are using predictive analytics to boost sales, growth, and provide guidance to their business strategies [link].

My question to you lot is this: how would you use the power of prediction to profit or achieve your goals in EVE? Would you consider using a web app offering market predictions to make investment decisions?

If you have any ideas on ways to use this power, please post or message me (if you want a more private convo). I might be taking on an EVE-related project for my spring semester capstone course. Your suggestions could help guide the features of the next EVE 3rd party tool!

Truly yours,
Plleasure


I have 3 words for you: ask Sabriz Adoudel
Robby Rova
Doomheim
#4 - 2016-07-24 04:24:41 UTC
The problem with predictive analysis in Eve online is that the items are NOT static and neither is the demand. At some point you'll have 20k players average and another 50k. This plays a huge role in demand and is impossible to predict without numbers from CCP. Alongside of this is the fact that the items are subject to change constantly. Ships will change bonuses / hard points and items will change stats. Everything is subject to change at ANY time throughout the course of new Eden. And they change frequently enough to where any prediction would be considered null due too not enough data to make an accurate prediction.
Vaerah Vahrokha
Vahrokh Consulting
#5 - 2016-07-24 07:33:44 UTC
Robby Rova wrote:
The problem with predictive analysis in Eve online is that the items are NOT static and neither is the demand. At some point you'll have 20k players average and another 50k. This plays a huge role in demand and is impossible to predict without numbers from CCP. Alongside of this is the fact that the items are subject to change constantly. Ships will change bonuses / hard points and items will change stats. Everything is subject to change at ANY time throughout the course of new Eden. And they change frequently enough to where any prediction would be considered null due too not enough data to make an accurate prediction.


EvE markets are a simplified version of the RL ones. RL is NOT static either, and it's a mess with as many wars, deaths, births, technological changes as EvE has got, and more.

That's why a good idea is to trade EvE's most liquid markets, where "accidents" have somewhat less of an impact and are somewhat smoothed out.
Robby Rova
Doomheim
#6 - 2016-07-25 01:16:40 UTC
Vaerah Vahrokha wrote:
Robby Rova wrote:
The problem with predictive analysis in Eve online is that the items are NOT static and neither is the demand. At some point you'll have 20k players average and another 50k. This plays a huge role in demand and is impossible to predict without numbers from CCP. Alongside of this is the fact that the items are subject to change constantly. Ships will change bonuses / hard points and items will change stats. Everything is subject to change at ANY time throughout the course of new Eden. And they change frequently enough to where any prediction would be considered null due too not enough data to make an accurate prediction.


EvE markets are a simplified version of the RL ones. RL is NOT static either, and it's a mess with as many wars, deaths, births, technological changes as EvE has got, and more.

That's why a good idea is to trade EvE's most liquid markets, where "accidents" have somewhat less of an impact and are somewhat smoothed out.


You're right, but RL is a lot more static than the eve marketplace which is why predictive analysis works. In the current setting we don't see any sort of stability with anything other than minerals. But even those need about a years worth of market data to make any sort of prediction with at least a 50/50 chance of being right. You're also screwed when a change occurs where the mineral drop tables get screwed with significantly, we saw this with moon minerals when they first made the t2 construction material changes. The good old days when tech was the bottleneck. You see entire markets shift drastically without any chance of guessing a number. You can say it'll be high, but won't ever be able to predict how high. There's not a stable long term demand like there is in RL.

The only possible item you could try to predict are is plex. It's the only thing that will have somewhat stable demand and the market data to give an accurate prediction.
William Legrand-Marx
Nemesis Ad Astra
#7 - 2016-07-26 11:54:29 UTC
Alright, I have seen what other said and you can predict markets. In fact those who can make money on it, they don't make the market to profit from it, they trade on it. Another fact, corporate forecasting is very much about it. You get macroeconomists into the deal and let them do their job.
EVE economy is much softer with less shocks than real economy, thus less variables to consider. Can be a good teaching groun for economists as I said many times already. As others said, RL is static - again wrong. From production capacities of SME to stock fluctuations.

To consider forecasting you have to understand what is in the eve economy. We got main things used - ships and construction sites. Fro ships you got parts, minerals, blueprints and so on. For constuction be it PI or Citadel you got pretty much the same. See who produces and buys or volumes, price changes, reasons behind price changes and volume changes, for how long, where and so on. This all will help follow the news about somehing and know what impact it will have.

Saying that forecasting EVE is impossible or hard is words of lazy men. Forecasting real economy can be done and it is done daily from banks to grandma buying new internet connection.

Saying that manipulating EVE is hard is words of person that never did it. manipulating real world is harder mainly due to increased number of factors, that is why geopolitics take place. Instead of shifting market from within you make if different from outside. That is why you lobby governments. Only technological leaps make change from within. Like internet development and network creationism from social to military.

All in all, we can see now that PLEX, for example, is growing in price again. This is a chance to see if Citadel will fix high PLEX price. The cycle started and it will peak around time the previous cycle happened. We will see it being done, if CCP of course secretly won't flood market with PLEX. If PLEX shoots last year top or more. It will mean Citadels are not working, which means Quant will have to do extra hours. So as I said, you can predict. Macroeconomics is about analysis and forecasting, add sme econometrics to it and some math and you got yourself a good night show to watch.

That is why I decided to make a trade guild. I want to teach people how to do that. Hopefully increasing market awareness of the players and making market a more competitive place. EVE University doesnt propose that, they have basic explanation of market mechanics. I tried to recruit through their recruit system but my recruitment post was deleted. I suppose they dont like the idea.

So, tools to evaluate could be great but withou knowledge no matter what data you show, if person sees only numbers and charts instead of evolution then there is no point in developing that tool because people won't see great use in it as they don't have knowledge to use it in first place.

Places like the rading guild would actually change that but the more blind sheeple trades on market the better for those who profit from them, which means I am the bad guy, for them.

There is nothing worthy in this world even if others think it is worth something...

Trogdor Montague
Jaque Pot Holdings
#8 - 2016-07-27 02:24:44 UTC
I want to play! :)
Vaerah Vahrokha
Vahrokh Consulting
#9 - 2016-07-27 08:13:55 UTC
Robby Rova wrote:

You're right, but RL is a lot more static than the eve marketplace which is why predictive analysis works. In the current setting we don't see any sort of stability with anything other than minerals. But even those need about a years worth of market data to make any sort of prediction with at least a 50/50 chance of being right. You're also screwed when a change occurs where the mineral drop tables get screwed with significantly, we saw this with moon minerals when they first made the t2 construction material changes. The good old days when tech was the bottleneck. You see entire markets shift drastically without any chance of guessing a number. You can say it'll be high, but won't ever be able to predict how high. There's not a stable long term demand like there is in RL.

The only possible item you could try to predict are is plex. It's the only thing that will have somewhat stable demand and the market data to give an accurate prediction.


I fail to see how RL is more static.
Current RL markets are less volatile than they used to be because of a series of temporary factors like QE, but we still see sudden crashes (Swiss National Bank removing "peg"), Brexit and bumps (Nintendo with Pokemon Go). And less, but still real markets swinging in commodities, including oil and seasonal futures.

Also, "about a years worth of market data" is well below the minimum amount needed to provide statistical confidence. 8 years is the basic standard for algorythmic trading. In EvE I did not exactly measure it because I have market data from 2009 anyway, so it's going to be more than enough.
Markets change, sometimes change a lot, due to patches.... and that's just how markets are meant to be.

In fact an efficient market rolls a tight range around it's underlying value and is not really profitable as is. The real gains come when something happens and markets need to reposition their value to another price. Savy investors and speculators are able to ride these changes, which form "trends" that some times last months. A liquid market featuring a trend bringing something from 1000 ISK to 2000 ISK is going to double whatever amount of ISK.

Sometimes the changes in price are sudden and cannot be exploited. That's the only case where your reasonment is right.

Many times changes are announced (i.e. patch notes) and many jump on the bandwagon and effectively "predict" the price.

Other times changes are seasonal or come from another affected product that got changed (or its gameplay make it rarer / more common) and again, it's possible to "sniff" the change while it happens and make a profit off it.
Robby Rova
Doomheim
#10 - 2016-07-27 18:22:16 UTC
Vaerah Vahrokha wrote:
Robby Rova wrote:

You're right, but RL is a lot more static than the eve marketplace which is why predictive analysis works. In the current setting we don't see any sort of stability with anything other than minerals. But even those need about a years worth of market data to make any sort of prediction with at least a 50/50 chance of being right. You're also screwed when a change occurs where the mineral drop tables get screwed with significantly, we saw this with moon minerals when they first made the t2 construction material changes. The good old days when tech was the bottleneck. You see entire markets shift drastically without any chance of guessing a number. You can say it'll be high, but won't ever be able to predict how high. There's not a stable long term demand like there is in RL.

The only possible item you could try to predict are is plex. It's the only thing that will have somewhat stable demand and the market data to give an accurate prediction.


I fail to see how RL is more static.
Current RL markets are less volatile than they used to be because of a series of temporary factors like QE, but we still see sudden crashes (Swiss National Bank removing "peg"), Brexit and bumps (Nintendo with Pokemon Go). And less, but still real markets swinging in commodities, including oil and seasonal futures.

Also, "about a years worth of market data" is well below the minimum amount needed to provide statistical confidence. 8 years is the basic standard for algorythmic trading. In EvE I did not exactly measure it because I have market data from 2009 anyway, so it's going to be more than enough.
Markets change, sometimes change a lot, due to patches.... and that's just how markets are meant to be.

In fact an efficient market rolls a tight range around it's underlying value and is not really profitable as is. The real gains come when something happens and markets need to reposition their value to another price. Savy investors and speculators are able to ride these changes, which form "trends" that some times last months. A liquid market featuring a trend bringing something from 1000 ISK to 2000 ISK is going to double whatever amount of ISK.

Sometimes the changes in price are sudden and cannot be exploited. That's the only case where your reasonment is right.

Many times changes are announced (i.e. patch notes) and many jump on the bandwagon and effectively "predict" the price.

Other times changes are seasonal or come from another affected product that got changed (or its gameplay make it rarer / more common) and again, it's possible to "sniff" the change while it happens and make a profit off it.


I shouldn't say static, i'll replace it with stable.

What i'm saying is, the fact that items in RL markets suddenly cannot be changed to something else is what makes RL more stable than eve. For example with the conversion of trade items to planetary items - any data prior to the change is irrelevant. Same with battleships & tiericide. Any item that changes will have a different set of demands than the item prior. And you CANNOT make an accurate assessment of that item with market data prior to the change. you'd be operating on data that is no longer valid.

Maybe a year's worth of data is enough, i say maybe because you're operating on something that you cannot accurately predict will happen again. 2-3 years is comfortable amount of data to accurately judge what will happen if all else stays equal. So only items that have been unchanged for that period of time can be accurately predicted.

changes are put into devblogs and yes, you're right we have a reasonable amount of time to react. But that's not prediction based upon that particular item's previous history. Its a culmination of what makes up that item - and still - you cannot predict when and where the price will hike and adjust due to stockpiles -etc. But at the end of the day, you don't need predictive analytics to determine what will rise and fall in a devblog. It's telling you right then and there what to buy, and the only analytics you have to do is figure out what is directly relate-able to the item being changed and how will it effect those items.

The only time i have ever seen predictive analytics work in eve was when they shuffled the moon minerals around for t2 materials. We had the data with the necessary amount of time to accurately estimate the amount of t2 goods being produced to use that as a reference to realize that Technetium was going to be a HUGE bottleneck. (https://oldforums.eveonline.com/?a=topic&threadID=1207775)

my point being is that the RL market has a lot more variables that ultimately remain relatively stable and have had YEARS upon YEARS to determine what will go up and down. As a population, we simply do not have an accurate assessment of the majority of the variables needed to make an accurate prediction. They are either hidden by CCP or we simply do not have enough time with that item in it's current state. When we do have those variables i think we can make it work, EI: High volume items, minerals in particular. But lower level items and ships we'd have no chance and getting and accurate assessment. Which to my understand was what the OP Was talking about - All items in eve.
TheSmokingHertog
Julia's Interstellar Trade Emperium
#11 - 2016-07-28 04:47:17 UTC
This is the game I play, loving it! EVE is real.

"Dogma is kind of like quantum physics, observing the dogma state will change it." ~ CCP Prism X

"Schrödinger's Missile. I dig it." ~ Makari Aeron

-= "Brain in a Box on Singularity" - April 2015 =-

Jdestars
Stars Research systems Incorporation
#12 - 2016-10-10 12:25:04 UTC  |  Edited by: Jdestars
Basicly i use export and just some 'cooking equation' used in signal analyse ... average harmonic and other but useless with the the time ;)

So you can use this kind of tool but ... in integrate chain if there are a lot (over than player ) of Automated Predictive tool which take control of decision you finish by predict what ? the mechanical of the tools programmation ?

If one player take control of a part of Evemarket ( Possessing the majority of stock in the order book ) do you think the tool can help you ?

Eve its good place for test Deep learning Network with large data if you include some rules like industrial chain supply item requiremnet
im sure dev or install this tool is a good technical challenge so do it for the challenge , not for a game result ( Isk) ;)