How to analyze loot

[falkao]

Originally Posted by Dawis

Well this Is precisely what seems a problem for me. Correlation between health and loot in fact needs to be linear - at least that is my basis assumption.
Not only linear but proportional. thus 10x the health 10x the loot.

what makes you assume that. Data shows something different.

Originally Posted by Dawis

My worries are that taking into account only loots above 50 PED you will get overestimated results for small monsters, while bigger monsters will be closer to actual loot size.

thats only a problme of sampling, but not with estimation if you have an exp dist. For a normal dist. the situation would be as you have described it.

Originally Posted by Dawis

Well of course it all depends on what you want to achieve - how you define "good loot" - I define it as better than average loot per killed hitpoints.
Thus i mob A gives you X loot per 1000 HP killed
and mob B gives you 1.1X loot per 1000 HP killed
then mob B is considered mob that loots better.

[/quote]

now I know what you mean. However, I didn't see any differences between mobs with same HP when modelling data. So I think there is none, otherwise we would already have noticed it.

To be sure here a comparison of the hp 300 mobs I have.


Click to enlarge

Statistically there is no significant difference between those 4 groups. As you see from Umbra my data is poor. However, still no significance.

[Tussi]

Originally Posted by jdegre

i'm sorry but this reasoning is completely flawed. "it must be fast, therefore it must be simple", right?. there are thousands of examples of computing systems which are extremely complex and very very fast.

Just because a system is simple (and fast to compute) dosn't mean that the output of it is predictable - especially when the output is spread over a large number of avatars.

Tussi

[falkao]

Originally Posted by jdegre

np

btw, now that we talk about systems that could be related to loot, i'm sure you're aware that some people believes that the skill system and loot are related somehow. and i'm not talking about noob players saying "damn, the ubers take all the loot!!"... i mean very very experienced players stating that loot has some relationship with your skill/profession levels and how it changes (for better or for worse) when you cross certain "special" points.

so, without the intention of shifting the focus of this thread, of course, maybe i'd not take the skill system out of the equation just yet.

to verify this we would need samples per skills, that we don't have atm. Therfore it can't be excluded atm but should be verified. This is not a problem for tghe global data and would imply only some further variance. However, if we would like to introduce a test day where only nobbs or only ubers do hunt, then we should see something, if it exists.

edit: oops I have overseen. Global data includes ava name. So maybe we already have some data. Let me check.

[Dawis]

now I know what you mean. However, I didn't see any differences between mobs with same HP when modeling data. So I think there is none, otherwise we would already have noticed it.

Yes.. from now on I think I will call it minimal observable loot problem As the problem rises from fact that you can not observe loot less than 50 ped, which could create some problems to my mind.

Originally Posted by falkao

what makes you assume that. Data shows something different.

If I am right - data should show something different. Of course size of globals does not grow lineary with HP.

thats only a problem of sampling, but not with estimation if you have an exp dist. For a normal dist. the situation would be as you have described it.

May be the problem is that I don`t know enough of the terms that you use.
Can you explain me better what you mean by exp dist ? If it is exponential distance then I do not get quite what it means. If by exponential distance you mean that you only measure the best of the loots of each monster, then I get it.
I have some background in math, but I am obviously not so good in statistics.
So can you explain how this exp dist helps you to get rid of minimal observable loot problem = 50 PED?

EDIT: - Reviewed the thread. Found out that with expt dist you most probably meant exp distribution. I know this type of distribution from my statistics course.
Well seems I misuderstood your main goal a bit. Now it seems to me that you are only trying to find distribution of loot values. Well it is quite obvious that it has exponential nature or close - well of course the actual data does not have to fit perfectly with any model.

[Legion]

falkao i guess ur not checking ur pm's?

[Kosh]

I asked this on page 3, but got no response.

This is great analysis, a couple of questions come to mind.

The loot from hunting and mining seems to be discrete and not continuous. You do not really get all values, have you tried fitting discrete distribution models?

You assume each mob you loot is not independent to the previous mobs, what if this is not the case? could be some kind of stochastic process. The box-plots imply the variance does not change every hunt, but I am not sure that means the loots are not effected by the previous loots.

[Doer]

Originally Posted by Dawis

Well - take simple average of Argo globals and simple average of Aurli globals.
Aurli global size will of course will be higher, but not much higher. Although Aurli has many times the health of argo the average global will be reduced by lots of small globals.
Let me elaborte:
Lets assume loot function
Loot = Health * C(x)

Lets assume C(c) generates constants from unknown parameter vector x from 0 (no loot)to very high constants (ATH)

Argo case Health = 200 --> loot = 200*C(x)
Aurli case Health = 2000 --> loot = 2000*C(x)

Remember that actualy we want to observe properties of c(x) function not whole loot function.
Of course in second case loot will exceed 50 ped much more frequently.
To compare loot is size with respect to monster you must divide it by HP (which strongly corelates to cost to kill that mob)
so 50 PED argo = 50 / 200 = 0.4 = C(x)
50 PED Aurli = 50 / 2000 = 0.04 = C(x)
So in aurli case we would take average from C(x) results in range 0.04; inf
but in Argo case we would take average from C(x) results in ragne 0.4; inf
That is big difference.

What i was trying to say is what Falkao explained much more technically with

Originally Posted by Falkao

fortunately a truncated exp will have the same characteritics as one that is not truncated. The only difference is the different mean you'll get. If you know the truncation point the original mean is estimated mean - truncation point. However, it is also possible that globals start at 50.

So substitute "truncated exp" for "tail" in my post.

[falkao]

Originally Posted by Kosh

I asked this on page 3, but got no response.

This is great analysis, a couple of questions come to mind.

The loot from hunting and mining seems to be discrete and not continuous. You do not really get all values, have you tried fitting discrete distribution models?

You assume each mob you loot is not independent to the previous mobs, what if this is not the case? could be some kind of stochastic process. The box-plots imply the variance does not change every hunt, but I am not sure that means the loots are not effected by the previous loots.

sry Kosh for the late reply. The data I used is Starfinders global data. The values are integers and are rounded. It seems floor is used, so rounded downwards.
The purist would say you have to use a discrete distribution. What I'm doing is a continuous approximation and thats sufficiently precise for our data.

If X has an exp dist with mean 1/a then Y = floor(X) has the following density

f(y) = a*exp(-a*n)*(1-exp(-a)) for n= 0, 1, ...

f(x) = a* exp(-a*x)

so the difference between those is a*exp(-a*n)*exp(-a) for x = n = 0,1.

for a = .1 and n = 0 the diff is 0.09 PED.
for a = .1 and n = 1000 the diff is 0 PED.

if a gets smaller (higher mean) say a = .01 and n = 0 (the worst case) the diff is .009.

So that is sufficiently precise from my point of view.

[TommyDaGreat]

Originally Posted by Snabble

heres how i analyze loots

Animal hide-- tt food - check
Animal oil -- tt food - check

etc etc

Animal hides are not TT food :P

You can earn 5 ped for every 100 ped of them you have in storage. And you know what they say about small water drops and oceans

[falkao]

Originally Posted by jdegre

np

..maybe i'd not take the skill system out of the equation just yet.

I did a first check and found 2 avas hunting the same mob (aurlis) with sufficient high number of globals (59/69).


Click to enlarge

Comparing loot leads to p = .419 (Mann-Whitney U-test), so not statistically different. This says nothing atm since we don't know if they have similar skills or not. Maybe one of them can tell us more, therfore the names.

If this i not allowed, then I'll remove them.