Menu
What's New
By:
Filters

Resource Area Concentration - A mathematical approach

What u're trying to do is called weighted moving average smoothing.
However that should be done on 2-dimensions not 1 (like smoothing over all adjacent drops by coordinates not just by consequent drops) since your initial data are at least 2D.
Problem is you cannot apply that type of calculations at all to data which have finite number of factors.
There's measuring theory that differentiates types of scales. Claim/nrf is a categorial scale with only 2 attributes possible. Measuring theory forbids any arithmetic operation on categorial scaled data allowing only 2 operations on them (equal to, not equal to)
Your method won't work because it have obvious theoretic flaws in its core. Also prone to curve-fitting issues because u have finite amount of data but infinite amount of options of choosing smoothing weights each giving different overall result.
 
stc44 said:
What u're trying to do is called weighted moving average smoothing.
However that should be done on 2-dimensions not 1 (like smoothing over all adjacent drops by coordinates not just by consequent drops) since your initial data are at least 2D.
Problem is you cannot apply that type of calculations at all to data which have finite number of factors.
There's measuring theory that differentiates types of scales. Claim/nrf is a categorial scale with only 2 attributes possible. Measuring theory forbids any arithmetic operation on categorial scaled data allowing only 2 operations on them (equal to, not equal to)
Your method won't work because it have obvious theoretic flaws in its core. Also prone to curve-fitting issues because u have finite amount of data but infinite amount of options of choosing smoothing weights each giving different overall result.
Click to expand...

Ok so you understand what I am trying to do and say that it will not work as it requires more than 2 states.. I accept that, but there are more than 2 states. I am attempting it at this level right now as I do not know how to represent it any other way.

Also please tell me what the theoretic flaws are and I will look them up and allow my brain to process the info ..

Thanks for the comment, appreciate it :)
 
R4tt3xx said:
Also please tell me what the theoretic flaws are and I will look them up
Click to expand...
see https://en.wikipedia.org/wiki/Level_of_measurement nominal level scale.
Existence of claims is a nominal attribute having only 2 possible values (claim, nrf)
In this case see mathematical operations allowed on them in the same section and that type of data allows only equality / non-equality comparisons.
However weighted smoothing u're using involves arithmetic operations such as summation.
This is not allowed on that type of data.
What u're up to do is like having 2 white T-shirts and 2 black and by performing some voodoo calculations and conclude that u have 4 grey T-shirts.
Or having 3 billion men + 3 billion women on Earth and conclude that entire Earth population are hermaphrodytes.
 
I am going to change topics a bit and ask another question... This will involve maths and is in the form of excel formulae

4 formulae that are interconnected. If 3 variables are provided, variable 4 can be determined.

Formula 1 Size Of Grid = (Max Amount of Points)^(1/root)*Radius*2
Formula 2 Max Amount of Points = (Grid/2/Radius)^(1/root)
Formula 3 Radius = (Grid/2/(Max Amount of Points)^(1/root))
Formula 4 Root=Log((Max Amount of Points),(Grid/2/Radius)

Here is the backbone of my sheet..

X=(cos(2*pi()*0.618*(Current Point)))*((Current Point)^(1/root))*Radius
Y=(-sin(2*pi()*0.618*(Current Point)))*((Current Point)^(1/root))*Radius

It's based on a fermat spiral and I have been meaning for someone besides myself just to go thru the math in order to see if there are no apparent flaws with it. The spiral is intentionally arranged the way it is via cos and -sin.

I also suspected that the spiral rotates so I did the following to counter that.

X=(cos(2*pi()*(0.618*(Current Point)+((Current Point)/(Max Amount Of Points)))*((Current Point)^(1/root))*Radius
Y=(-sin(2*pi()*(0.618*(Current Point)+((Current Point)/(Max Amount Of Points)))*((Current Point)^(1/root))*Radius

This adjustment places each point where it would be if the spiral would be rotated to that point.

So that is the math part done.

Where I used this was to say, when I dropped a probe and missed, I set the current point to the minute of the drop and the centre of the spiral to my coordinates. I would then hit all the claims on the spiral till it vanished, then start again.. Obviously the spiral's rotating over time was countered as above.

This is how I roll, why adapt to a scenario when you can just bypass it and get similar results.

This system still works depending on how concentrated the area is..

Please comment on the math, do I have something obviously wrong ?
 
Last edited:
stc44 said:
math itself looks ok
Click to expand...

Cool, next piece is this.. I got stretching and elliptical done, but have no clue on how to do Schwarz-Christoffel Mapping. This is a pretty cool one as it would allow you to map a pretty uniformly distributed circular matrix (ie fermat spiral) onto any polygon (or land area).

I mean it should be a simple matter for a graphics card to perform the calculations needed for the mapping.
 
this looks like the definition of overthinking something. anyway wish you gl on finding something out. if there is anything to find out.
 
stc44 said:
No idea how it could be done in general case.
I'd possibly try to do that in two steps.
1. article about mapping circle to a regular n-polygon.
2. article about mapping regular n-polygon to an arbitrary n-polygon
Also since land area size is not that big compared to finder radius, I'm quite sure it's possible to use other more computationally friendly transformations without significant errors.
Click to expand...

I have absolutely no idea on how to do this transformation in Excel, the others were a piece of cake.
 
While mining over the weekend, (and 2 hofs later), I had a brain wave...

In order to test what I think I have discovered, I would need a finder with the least amount of variance, which would be the little rookie ...

The rookie normally hits on a class 1 which is a max of 0.24 ped, so I would have a max of 4 headers in a 108m grid ie 0.24/0.05 = 4.8 +- 27m between them on the spiral. Best location to look would be at header 1 of which the offset is classified. If I am correct it should yield near 50% hitrate.

I also have reason to believe that the number sequence Mindark is using for locations that produce hits are prime numbers for the most part, then alternate to non primes during high loot waves resulting in higher loots and concentration.

Using rookies also has another plus, if areas deplete over time, rookies will only deplete the area at 10% of the rate of normal finders, so one should expect to find and I will quote an old term, veins !!!

Will post the result with pics if successful
 
I am seeing the appearance of what can only be described as a waveform, the decreases the amount of claims in an area as time goes on, this waveform is very obvious when looking at the frequency and size of hall of fames.

The wave-length also changes over time, with the lowest detected length of 15 minutes.

My mining sheet's concentration now varies uniformly over the number of maximum claims. This is truly a player driven dynamical system.
 
You must log in or register to reply here.
Back
Top