Menu
What's New
By:
Filters

Resource Area Concentration - A mathematical approach

R

R4tt3xx

I want to believe
Joined
Apr 10, 2005
Posts
2,174
Location
South Africa
Society
Freelancer
Avatar Name
Alexis Sky Greenstar
Greetings all

I have been mining for quite a while now, well since open beta and have noticed some significant changes regarding mining, one of the biggest ones is that actual areas have different concentrations of resources

Depending on the resource concentration the distance between claims will obviously change, this does not take into account density which will be discussed separately. So how do we find areas with high concentration. Simple...

Here is an easy to understand grid
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!

Each "!" represents a claim, the total size of this grid is 1km by 1km, the concentration is 100% and the density is 1 ped per claim, total resources = 100ped . We have 100 claims in this 1km area so the average distance between them will be 100m.1000 / 100^(1/2) If the concentration was say at 6%, ie 6 points per 1km, the distance between will be 1000 / 6^(1/2) = 408m

To obtain the correct concentration of resources in an area is not too hard depending on how accurate you wish to be. 7 probes should be good enough to determine the concentration.

Allow me to explain. First we test at 100% concentration, If we miss, we test half of that ie 50%, if we miss drop down to 25%, once you hit, add half the current percentage to your current percentage and test again, if you miss subtract half the next percentage, 7 probes is all it takes to get an accurate reading within 1.5%

Of course this post has dire consequences, first of all it only deals with concentration not density, which I am going to use to represent the average size of a claim. I am pretty confident that the 1km size of the each cell is correct.

Onto the consequences, miners will become smarter, checking their cells and only mining the ones with highest concentrations, I do not know how the recharging of these cells would work so I will check the cells that I typically mine for high concentrations first, it may be that only by seeding cells, ie adding to its concentration by literally spending ped on it without obtaining a claim, would increase both the concentration and density. If this is true we will have some truly dry spots.

I am taking a risk by actually posting this information as it will change the mining patterns of most miners for which it makes sense to, and who knows, perhaps miners will come to accept the spiral idea as plausable, but more on that a bit later.

Cheers for now, please comment below.

R4t
 
Have u ever testet that ores are not in the ground anymore and the return is avatarbased?
 
you still think that the ressources are allready in the ground and waiting to be found? that is not the case. that has been changed a looong time ago. go and carpet bomb the same area for a month and you will see that there is no pattern. the claims are random. just some server are hot and some are not, possibly the same like on hunting.
that is all the secret there is.
 
Count Joe said:
Have u ever testet that ores are not in the ground anymore and the return is avatarbased?
Click to expand...

This method would also apply to the "hotness" of the servers, surely hot servers would have their claims closer together, making them easier to find ... At least avatars would perceive them that way.
 
Me Really Never said:
you still think that the ressources are allready in the ground and waiting to be found? that is not the case. that has been changed a looong time ago. go and carpet bomb the same area for a month and you will see that there is no pattern. the claims are random. just some server are hot and some are not, possibly the same like on hunting.
that is all the secret there is.
Click to expand...

No they are not in the ground, but their concentrations are set, how you interact with the areas "pool" depends on the miner.

I am seeing that walking into a new cell, triggers the "generation" of resources for that cell starting at where you entered it. This is consistent with the way it was done in beta.

Another observation that I made was the missing "shadow mining" glitch / flaw, this suggests to me and I am going to confirm that each area has it's own set pool.

The devil is in the details, most carpet bombers will not see something like this. Oh and each probe that you drop, obviously alters the concentration and density and re-centres the generated resource claims around the avatar.
 
Last edited:
R4tt3xx said:
(...)
Here is an easy to understand grid
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!
!!!!!!!!!!

Each "!" represents a claim, (...)
Click to expand...

More like each "!" represents a NRF.
 
antonioliveira said:
More like each "!" represents a NRF.
Click to expand...

Again depends on concentration, average is what 27% so .... 1000/27^(1/2) 192m gap between claims, so about 5 claims per 1km.
 
I don't understand the power of (1/2). Where does that come from?
 
Kerham said:
I don't understand the power of (1/2). Where does that come from?
Click to expand...

Square root

Distance = Size of the square / sqrt(amount of points in square)
 
I see mining as a game, when you find a claim there as always more nearby, you must make the right choices to find them, there are often clues.

tests I have done previously showed me for sure resources are no longer in ground but are still concentrated in areas.
grab a load of rookie finders, blow 20 ped, run in circles, auto-use planetside, you will get 50% return but see how the claims come and how the are concentrated.
 
Until someone posts a lbml screenshot with a hit rate over 50% on 1000+ drops... i'm not going to believe a word of anyone's mining theories.

If people demand proof.. that's how you get it.

I carpet bomb, drop as fast as a can w/o overlapping.. and I do just fine TT wise.
 
Rocket192 said:
Until someone posts a lbml screenshot with a hit rate over 50% on 1000+ drops... i'm not going to believe a word of anyone's mining theories.

If people demand proof.. that's how you get it.

I carpet bomb, drop as fast as a can w/o overlapping.. and I do just fine TT wise.
Click to expand...

It would require a but-ton of prep work, you would need to know the concentration of each cell you go into and adjust your mining pattern on the fly as it goes down before moving onto the next cell and starting the entire process all over again...

The problem is that the prep work will eat away at your profits if you are not going after high mu resources.

The concept works, but is a pain to implement. You would have to keep track of each probe and basically draw up a balance sheet with the total either subtracting or adding to the areas concentration.

At this moment in time, I have not idea where high ped value claims fit into this.

But we trundle on regardless ..
 
ImUp2Mischief said:
I see mining as a game, when you find a claim there as always more nearby, you must make the right choices to find them, there are often clues.

tests I have done previously showed me for sure resources are no longer in ground but are still concentrated in areas.
grab a load of rookie finders, blow 20 ped, run in circles, auto-use planetside, you will get 50% return but see how the claims come and how the are concentrated.
Click to expand...

Exactly ... I can do it in 7 double bomb probes.
 
Coordinates of drop and coordinates of possible claim are not mutually independent in a statistical way as claim cannot be discovered outside your finder search radius at any given time.
So claim pattern will in some way reflect your exact drop pattern so you need to somehow "subtract" drop pattern to see "pure" claim data.
Only after that we can even start thinking about discovering patterns in claim data.
 
stc44 said:
Coordinates of drop and coordinates of possible claim are not mutually independent in a statistical way as claim cannot be discovered outside your finder search radius at any given time.
So claim pattern will in some way reflect your exact drop pattern so you need to somehow "subtract" drop pattern to see "pure" claim data.
Only after that we can even start thinking about discovering patterns in claim data.
Click to expand...

I do not quite understand this, but let me put my own spin on the probe vs claim offset.

I have noticed that over time the distance between my claims and probes drift, the effect is more noticable when I am not using a rookie.

I can assume that it is because I am not adjusting the concentration value in real time. I wonder if it is not possible to use that distance to correct for drift....

Again I apologise for perhaps going off topic but your posts intent is not very clear, at least not to me.

Thanks for the comment :)
 
I have decided to give you guys a quick example of how this works.

In this example, you will see the concentration drop from 100 % (100 claims) to 0 % (0 claims) in the span of a minute.

There are two variants, namely an archimedean and a fermat spiral styles. I would like to remind viewers that this is merely an example to illustrate the mechanics.

The origin is always the point where a probe is used.

L8r
 
Last edited:
R4tt3xx said:
I have decided to give you guys a quick example of how this works.

In [URL="https://drive.google.com/file/d/0Bz393x-q88QWZG9aX2tSU3M3RnM/view?usp=sharing]this[/URL] example, you will see the concentration drop from 100 % (100 claims) to 0 % (0 claims) in the span of a minute.

There are two variants, namely an archimedean and a fermat spiral styles. I would like to remind viewers that this is merely an example to illustrate the mechanics.

The origin is always the point where a probe is used.

L8r
Click to expand...

This site can’t be reached

%22https’s server DNS address could not be found.


the link doesnt work to me
 
just wanted to add a quick interesting observation i just made.
i was mining on cyrene in the area around the white spot on the map (theres an asteroid above you in the air). i was mining on the ground and got 2 claims i couldnt reach. after checking coords i saw that the usual heigth of the claim is 110ish meters but the the 2 missing claims are at 1000ish meters. so the claim is on the asteroid above me around 900m higher than my position. i cant reach em though because theres some kind of shield around it. also interesting is that the ressource is in 176m depth. but that would actually be in the air shortly below the asteroid lol...
 
R4tt3xx said:
I do not understand the point.
Click to expand...

People (and pigeons) identify patterns even in random systems where there is no relation between their actions and the outcomes.
 
Antrace said:
This site can’t be reached

%22https’s server DNS address could not be found.


the link doesnt work to me
Click to expand...

Fixed, thanks
 
Kosh said:
People (and pigeons) identify patterns even in random systems where there is no relation between their actions and the outcomes.
Click to expand...

This is not a random system ..... I am waiting for my new credit card (+-2 days) to arrive so that I have more funds to attempt a non rookie run with a running balance sheet of sorts which will adjust the concentration depending on how much I alter the area.

I tested an area last night using the data collected from the night before and it was pretty much the same. I can safely assume that no one mined that area :)

I will reveal more data as I obtain it, but low peds is an issue that I will resolve shortly.

The spreadsheet predicts where claims will be, when configured correctly it can even determine type, but I have not been dynamically adjusting it as I mine. I am the type of guy that looks for ways around a problem in order to get the proper result with the least amount of manual effort so drawing up a real time balance sheet is not really mu cup of tea..
 
Last edited:
R4tt3xx said:
This is not a random system ..... I am waiting for my new credit card (+-2 days) to arrive so that I have more funds to attempt a non rookie run with a running balance sheet of sorts which will adjust the concentration depending on how much I alter the area.

I tested an area last night using the data collected from the night before and it was pretty much the same. I can safely assume that no one mined that area :)

I will reveal more data as I obtain it, but low peds is an issue that I will resolve shortly.

The spreadsheet predicts where claims will be, when configured correctly it can even determine type, but I have not been dynamically adjusting it as I mine. I am the type of guy that looks for ways around a problem in order to get the proper result with the least amount of manual effort so drawing up a real time balance sheet is not really mu cup of tea..
Click to expand...

I ask again as you seems to like testing and calculating - have u ever tested if everything is based on your avatar?
 
R4tt3xx said:
I do not quite understand this
Click to expand...
Then prepare for the wall of text :)

Imagine u're flying in the air and made a photo of the flowers on the ground trying to find out which area have more red flowers. In this case all observations are immediately available.
This is however not the case in mining
You may drop a single probe at a time on a single (x,y) coordinates and observe an area around these coordinates at a finite radius. In this case everything gets worse and this is where statistical dependency come into play.

Assume that I programmed loot algorithm which gives you a claim every 10 minutes. You decided to make a run in a way like "I'll run straight to the north, drop 10 probes on the way, then move west 110m turn south, drop 10 more probes and so on so I'll cover a rectangle area". You manage to move between consequent drop coordinates in around 1 minute because that's your speed. In this case every 10th of your drop will be successful. But because one side of the rectangle is exactly 10 drops away from another you will observe claims only on the "edges" of that rectangle. Then u will assume that these edges are good area to mine because they form easily recognizable geometric pattern (straight line) and everything in between is a dead area. Which is false conclusion. Assume side of the rectangle is 9 drops. Then u will observe a single diagonal line pattern and conclude there's a vein of resource. Depending on how exactly u drop there will be different patterns observed none of which reflects true distribution of the claims.

Same thing could happen in case u're observing time patterns. Assume I programmed and algorithm which gives you a claim on a fixed y coordinate. You move in a 10drop rectangle pattern and observe claims to appear on the same coordinate. But you conclude they appear after 10 minutes and there's loot wave with 10 min period.
What happens if you come tomorrow to use that dependencies you've found today? U will fail because none of these reflect real distribution. Tomorrow you will conclude that previously discovered vein has been mined out by somebody else earlier on. Then you'll try to wait 10 minutes between the drops as per previously discovered pattern but that seem not to work either. So u will keep "discovering" patterns whch looks exremely good today but don't work tomorrow.

Thats how statistical dependency leads to false conclusions.

Statistics says that if mutual dependencies between observed factors in data samples are possible (and it's quite possible to have both claim clusters and timed waves in loot generation algoithm) then you just cannot make any viable theories on how exactly every single factor influences another as you will almost surely have false conclusions now and then. Your only hope is to somehow eliminate mutual influence of all factors before examining patterns of an each single one separately. There are methods in case of linear dependency (see principal component analysis as an example of theoretically well based one) but thats damn hard in case of nonlinear without any knowledge of the process.

In our case it's almost surely non-linear dependencies on both of the factors since dependency over time means either up/down trending or in our case most likely cycles. And there's no linear deependency between coordinates in a way "higher x coordinate means higher claim density or resource size". So each of our 2 factors produces nonlinear patterns by itself and also there's mutual nonlinear dependency. Currently there are no theoretically well based methods for dealing with such dependencies.
So pretty much all the researcher may do at this moment is to collect immense amount of data, and continuously perform "voodoo dances" on these data like grouping by conditional probability, local linearization etc. But these are still shamanistic voodoo dances regardless of being perormed on a higher level of understanding what's going on. Or wait until better statistic methods are being developed.

Or reformulate an original task.
 
Count Joe said:
I ask again as you seems to like testing and calculating - have u ever tested if everything is based on your avatar?
Click to expand...

That is possible yes, if it works for players too, the chance of it being avatar based would drop...
 
stc44 said:
Then prepare for the wall of text :)

Imagine u're flying in the air and made a photo of the flowers on the ground trying to find out which area have more red flowers. In this case all observations are immediately available.
This is however not the case in mining
You may drop a single probe at a time on a single (x,y) coordinates and observe an area around these coordinates at a finite radius. In this case everything gets worse and this is where statistical dependency come into play.

Assume that I programmed loot algorithm which gives you a claim every 10 minutes. You decided to make a run in a way like "I'll run straight to the north, drop 10 probes on the way, then move west 110m turn south, drop 10 more probes and so on so I'll cover a rectangle area". You manage to move between consequent drop coordinates in around 1 minute because that's your speed. In this case every 10th of your drop will be successful. But because one side of the rectangle is exactly 10 drops away from another you will observe claims only on the "edges" of that rectangle. Then u will assume that these edges are good area to mine because they form easily recognizable geometric pattern (straight line) and everything in between is a dead area. Which is false conclusion. Assume side of the rectangle is 9 drops. Then u will observe a single diagonal line pattern and conclude there's a vein of resource. Depending on how exactly u drop there will be different patterns observed none of which reflects true distribution of the claims.

Same thing could happen in case u're observing time patterns. Assume I programmed and algorithm which gives you a claim on a fixed y coordinate. You move in a 10drop rectangle pattern and observe claims to appear on the same coordinate. But you conclude they appear after 10 minutes and there's loot wave with 10 min period.
What happens if you come tomorrow to use that dependencies you've found today? U will fail because none of these reflect real distribution. Tomorrow you will conclude that previously discovered vein has been mined out by somebody else earlier on. Then you'll try to wait 10 minutes between the drops as per previously discovered pattern but that seem not to work either. So u will keep "discovering" patterns whch looks exremely good today but don't work tomorrow.

Thats how statistical dependency leads to false conclusions.

Statistics says that if mutual dependencies between observed factors in data samples are possible (and it's quite possible to have both claim clusters and timed waves in loot generation algoithm) then you just cannot make any viable theories on how exactly every single factor influences another as you will almost surely have false conclusions now and then. Your only hope is to somehow eliminate mutual influence of all factors before examining patterns of an each single one separately. There are methods in case of linear dependency (see principal component analysis as an example of theoretically well based one) but thats damn hard in case of nonlinear without any knowledge of the process.

In our case it's almost surely non-linear dependencies on both of the factors since dependency over time means either up/down trending or in our case most likely cycles. And there's no linear deependency between coordinates in a way "higher x coordinate means higher claim density or resource size". So each of our 2 factors produces nonlinear patterns by itself and also there's mutual nonlinear dependency. Currently there are no theoretically well based methods for dealing with such dependencies.
So pretty much all the researcher may do at this moment is to collect immense amount of data, and continuously perform "voodoo dances" on these data like grouping by conditional probability, local linearization etc. But these are still shamanistic voodoo dances regardless of being perormed on a higher level of understanding what's going on. Or wait until better statistic methods are being developed.

Or reformulate an original task.
Click to expand...

Agreed, this concept has to work consistantly over time and result in correct predictions. So it needs to be tested on a daily basis and show consistant results.

Once that phase is over, limited release, then full if it ever gets that far...
 
So the sheet has changed again, this time focusing on literally balancing probes spent and resources found WITH MARKUP !! I tested the setup without it and was not very happy with the results, adding mu to the value yielded a profit.
 
And the hitrate tanked soon there after .....
 
I think I am going to go thru the math quickly just to make sure that it is correct.

####
####
####
####

I have a 2d array consisting of 4 units in each line, so 16 in total. These units are 1 distance apart from each other. The concentration sits at 100% as I cannot fit any more units into the set area without overlapping.

Distance = Length of field (4) / The Square root of the total amount because we are dealing with a 2d array (sqrt(16)) which gives us 1 distance, which is correct

Dist = Length / sqrt(total units)

Based on the above let's get really weird..

I take 1 unit away and re-arrange the remaining units in the square, what is the distance now ?

Distance = 4 / sqrt(15) = 1.03 distance

The concentration = 15/16*100 = 93.75 %

Another array

##############

This is a one dimensional array so ...

Distance = Length / units^(1/dimensions)

Dist = 16 / 16^1 = 1 distance

Remove 1

Dist = 16 / 15^1 = 1.0667 distance


In order to find the correct concentration % we would adjust it based on the results of previous attempts.

Attempt 1 assume 50%
Attempt 2 would assume a change of half that ie 25%
Attempt 3 would assume a change of half that ie 12.5%

If I had hit on all 3, my concentration would be the sum of all 3 changes so 50+25+12.5 = 87.5%

If I had missed on all 3 well 50-25-12.5=12.5 %

The basic idea here is to add to the previous value on a hit and subtract on a miss. Please note that the base concentration will change depending on the resources that you obtain from the scanned area, so using a rookie tool is advised.

1 Hit 50 + 25
2 Miss 50 + 25 - 12.5
3 Hit 50 + 25 - 12.5 + 6.25
4 Miss 50 + 25 - 12.5 + 6.25 - 3.125

Conc = 96.875 - Pretty Good

Assuming each claim is worth 1 ped (and I have 96.875 claims in my area) , there should be 96.875 peds worth of resources in the scanned area.. Each ore probe I drop would add 1 to the concentration and subtract the resulting claim's value from it, resulting in a decrease in the amount of claims in the area. Please note that it is possible that Mindark uses the found resources mu value as a base.

There is soo much more to this, treat this as an introduction :)

Cheers
 
You must log in or register to reply here.
Back
Top