Mining loot analysis

[falkao]

...
Edit: On second thought, with Alicenies at TT = 0.05, each click gives a 0.05 * 9 = 0.45 PED, while each click of Force Nexus only gives 0.01 * 20 = 0.20 PED. With Garcen Grease, at TT = 0.10, each click gives 0.10 * 8 = 0.80 PED. So.....hehehe.....maybe your right.

this is more or less on how I would have expected it. Not linear on units but more clicks on lower TT res. The relation itself is not known atm.

 

[falkao]

I analyzed again units with respective rounding using the model with loot = ,21*3^class and I came to the conclusion that the not so conservative approach is better. Hence the model is now ,22*3^class PED for class means and I've updated the main post. Expected return rate is now 96%instead of 92%. I will however do some verifications on class weights.

 

[falkao]

updated class weights slightly in main post and class limits for class 2.66, 3 and 4. Correction was needed due to overestimation when including higher TT res in low class weight estimation. Hence return rates are lower now by 1%.

 

[falkao]

A simple application for the loot model is the simulation of mining runs. Here an example.

10k avas do drop 996 probes. How do mining runs look like and where do they end up?

Fig. 1: Some mining runs
[br]Click to enlarge[/br]


Fig. 1 shows the mining runs for the first 10 avas. Some of them do start with a nice global and will loose thereafter, whereas others start with loss and will recover eventually, other not.

The final return is distributed as follows.

Fig. 2a: Histogram of final return
[br]Click to enlarge[/br]

better seen as

Fig. 2b: Kernel density estimate of final return
[br]Click to enlarge[/br]

Tab. 1: Statistics

Median final return is -26.2 PED which corresponds to a return rate of 95% (this is the model assumption). About 40% of the avas do profit, the rest is losing. Max profit was about 320 PED, whereas max loss is 208 PED. Overall 84k PED are retained by MA. The expected retained amount would have been 249k PED. Due to an excess of observations in the right tail (second peak) the observed amount is quite lower.

Doing the same simulation with 10k runs per Ava leads to the following.

Tab2: Stats with 10k runs per ava

Fig. 3: Histogram of final return with 10k runs per ava
[br]Click to enlarge[/br]


The distribution follows now a normal one and only about 20% of avas do profit. Max observed profit is 1163 PED and max loss is 1043 PED corresponding to 1.22 and .8 return rate respectively.


So what can we conclude from this? Having only 1k runs per ava will produce quite a lot of avas that will profit. If those go on with mining, then the most will lose as expected but there are still some (20%) that are able to profit.

Now I also know where all this loot stories and theories do come from.

There is one question that should be addressed. Is this system fair?

We don't know if MA implemented loot in that way. If we assume that everything is random then this will lead to what I have described. If there is some loot bias, the system will behave differently. With loot bias I mean a correction that does not allow that 20% do profit at the expense of others.

So without a loot bias the system is not fair, although everyone would have the same chance, there will always be a minority that will profit. From my point of view this should be explained in more detail by MA, so that everyone can be confident.

 

[Galadea]

Wonderfull analysis

Too bad i can't +rep you

 

[falkao]

If the system behaves purely random, then there are implications for MA as well. Income for MA would be delayed and I'm rather sure that they'll have some countermeasures, but I'll analyze this in more detail.

Here some further comments. If 20% will profit, then we have 20% of players that are going to tell how beautiful this game is and will believe that they have found a way to profit. About 1% will do exceptionally good, those will believe that they are the masters. About 1% will have a return rate less than or equal 85%, those are then unlucky ones and they will never know what they did wrong. Since 80% won't profit but 20% do, you'll have a lot of whining.

 

[sawachika]

If the system behaves purely random, then there are implications for MA as well. Income for MA would be delayed and I'm rather sure that they'll have some countermeasures, but I'll analyze this in more detail.

Here some further comments. If 20% will profit, then we have 20% of players that are going to tell how beautiful this game is and will believe that they have found a way to profit. About 1% will do exceptionally good, those will believe that they are the masters. About 1% will have a return rate less than or equal 85%, those are then unlucky ones and they will never know what they did wrong. Since 80% won't profit but 20% do, you'll have a lot of whining.

Interesting......hmmmm....now I wonder where am I? Am I that 20% or the 80%?

 

[falkao]

Interesting......hmmmm....now I wonder where am I? Am I that 20% or the 80%?

, we actually don't have any clue if the system is really purely random and hence does behave as described. If it is, then your question is justified.

 

[falkao]

here a further simulation.

Return rate distribution for 3 different scenarios:
10k drops by 10k avas
1k drops by 100k avas
100k drops by 1k avas

Break even is marked by the vertical black line.

[br]Click to enlarge[/br]

Please note how the variance of the retrun rate diminishes when increasing run length (drops) per ava. Avas doing short runs (below 1k drops per run) will have the impression as they are doing sometimes very good and sometimes not. Avas doing long runs will have the impression as they wont have that much success but also not that much loss. Off course, if we sum up all results from all avas with short runs and compare it to the sum of all avas with long runs, the result will be the same. However, perception is completely different.

Does this explain why experienced users will quit sooner or later? It is for sure not a motivation to see that the max own overall return rate is less compared to those that started.




Since in the above graph run length and number of avas did vary here for completeness the same from above but with a constant number of avas (1k). It leads to the same distributions.

[br]Click to enlarge[/br]

 

[falkao]

I've added loot simulation to the main post and also a section about why some sizes are missing when using amps.

I guess we've covered now nearly every aspect.

 

[sawachika]

From the looks of the graphs you gave, I seem to get the impression that it would be bad to drop too little and to drop too much.

The black line being the baseline where you get no profit/no loss.

The blue line being, if you drop 10k probes/bombs, you'll get a 20% chance to profit and an 80% chance to not profit.

The red line being, if you drop 100k probes/bombs, the chance of you profitting is almost close to 0.

The green line being, if you drop 1k probes/bombs, you'll get a 40% chance of profitting.

Hence, does that mean its a good suggestion to drop like 1k probes/bombs (To follow the green line) and get a 40% chance to profit. Rather than following the other 2 lines. This will ultimately provide a good guideline to miners, imho, that a sensible way of mining is to do runs of 1k probes/bombs. No more, no less. And once you profitted. Its time to pull out of the run and wait for a cool-down period. Cause continuing on the run would ultimately bring you back to the 1.0 return rate or even lower.

Nice find!

(Now we know that it ain't right to do VERY LONG runs of many drops. As you will end up losing. And that we should stop as soon as we profit. "I think.")

 

[falkao]

...

Hence, does that mean its a good suggestion to drop like 1k probes/bombs (To follow the green line) and get a 40% chance to profit. Rather than following the other 2 lines.

since the loot distribution is heavy right tailed and the probability to get a global is about 1% (unamped enmatter) you can expect in mean about 2.7 globals out of 1k drops. There will be however runs where you'll get quite more, say 4 or quite less, say 0 to 1. Those runs with an excess in globals will be those 40% of runs where one profits.

However, doing 10k runs with 1k drops is the same with respect to overall return rate as when doing 1k runs with 10k drops. In The difference is the observed return rate distribution over runs. In the 10k runs with 1k drops there will be more single runs where you profit, whereas with 1k runs doing 10k drops there will be less. Overall spend Peds and return is the same however. Hence if you look at the single runs with 1k runs you're in the green line, however summing them up, you're in the blue one.

The matter changes if runs do come from different avas. Take 1000 avas that are doing 1k drops. 40% of them will end up with a profit. Hence the perception of loot is biased with respect to runs with a higher number of drops. This explains the myths we have about loot. I'm sure that the 40% lucky ones do think that they did something different as those that did loose.

...
This will ultimately provide a good guideline to miners, imho, that a sensible way of mining is to do runs of 1k probes/bombs. No more, no less. And once you profitted. Its time to pull out of the run and wait for a cool-down period. Cause continuing on the run would ultimately bring you back to the 1.0 return rate or even lower.

What I described does not help to higher returns. Sure, if you're doing very long runs, you might get disappointed as your return rate will spread less and hence you won't get that much runs where you can say that you profited.

If one is only testing out mining, then one suggestion can be to start dropping bombs/probes till a maximum of 1k drops but stop when you see that you're profiting. If however the returns are less than 80% you have 2 choices, or to stop or to continue to recover from it.

(Now we know that it ain't right to do VERY LONG runs of many drops. As you will end up losing. And that we should stop as soon as we profit. "I think.")

correct, but then you have to stop mining completely and I guess that's not the intention.

 

[Jimmy B]

Is this is based on tt loot? With markup on loot and those return figures, this probably means most miners profit?

I always thought miners had it easy

 

[falkao]

Is this is based on tt loot? With markup on loot and those return figures, this probably means most miners profit?

I always thought miners had it easy

yes, all is based on TT, hence adding markup returns do look quite good, better than with hunting.

 

[Anna Ufo]

Yes, i think TT is the only good way on experiments, because imho the system works based on tt values. This includes loot and skill system at least.

 

[falkao]

one further observation.

Finders do have 8 uses per min and this corresponds to 480 drops per hour. Let's assume atm that we don't extract finds.

Cost of those 480 drops is about 253 PED and you should get a return of at least 243 PED (95% return rate). Hence one hour will cost you maximally 12.7 PED. If we consider that extraction needs time, we should consider only half of the costs leading to 6.34 PED per hour. If you mine amped, then costs will be proportionally higher according to the amp you use.

All this fits quite good to MA's given figures of 5 to 15 PED per hour and hence a further indication that rate of return is about 95% or higher.

 

[Jimmy B]

Finders do have 8 uses per min and this corresponds to 480 drops per hour. Let's assume atm that we don't extract finds.

When I mine 40 bombs and 40 probes takes me over an hour normally. But I'm quite slow I quite like amps too, so I probably still get through that amount of PED.

 

[falkao]

When I mine 40 bombs and 40 probes takes me over an hour normally. But I'm quite slow I quite like amps too, so I probably still get through that amount of PED.

well, my example was more a cross check to see if everything fits. When I go out mining, I'm however able to to burn 100 probes in quite a short time . I practice mainly random dropping.

 

[Noodles]

Catching up on this thread. I kept track of my hitrate for a while.
[br]Click to enlarge[/br]

Nearly perfect agreement with Steffel's 27.1% number. Some of the drops were ore (mostly unamped) and some were enmatter (mostly 101 amped).

I have some comments on your updates to the first post (great work!). Will try to get to them later.

 

[LividRayne]

Good work guys, I have been trying to analyze mining for years but never to this extreme, just little case studies on my own. If only there were a way to tell when your "hot" and when your "not"

This might be weird but I think hunting and mining loots go hand in hand. If you up in one, your up in the other. I have only had 2 HOFs (990 ped narc pvp4, and 872 ped allophyl) but what was weird is that they were within the same week and that i had globaled on either mining or hunting once a day for the week span. Hope this might help a little bit if anything a consideration of the tie between "hot" and "cold" for both mining and hunting

 

[falkao]

..

Nearly perfect agreement with Steffel's 27.1% number. Some of the drops were ore (mostly unamped) and some were enmatter (mostly 101 amped).

can you provide the exact numbers? finds/drops

 

[falkao]

This might be weird but I think hunting and mining loots go hand in hand. If you up in one, your up in the other. ..

very hard to verify. I have more the impression that everything works on its own. For example, when changing mob I had the impression that a new series starts independently from the one before.

With respect to mining it might be interesting to know if ore and enmatter finds do have it's own series or if they a part of the same pot.

 

[Galadea]

With respect to mining it might be interesting to know if ore and enmatter finds do have it's own series or if they a part of the same pot.

I would say that they don't share the pot. I and a lot of other miners have experienced that.

 

[falkao]

I would say that they don't share the pot. I and a lot of other miners have experienced that.

that's very probable. My personal opinion is that every activity within EU has its own pot, like enmatter is independent from ore, or hunting one mob is independent from another mob, but I can't prove it atm. Hence, all this is only a hypothesis and might get falsified.

To be more precise here the possibilities:

1) Everything is random. In this case ore and enmatter loot are independent from each other as well as loot within each ress. type.

A loot bias exists with the following possibilities:

2) Loot is per activity within profession. In this case every ress. type has it's own pot in which it behaves mostly random with the possibility of a loot bias within ress. type. Loot between ress. types is independent from each other.

3) Loot is per professions. In this case loot is mostly random but a loot bias exist. In this case loot within and between ress is not independent.

4) Loot is over all professions. In this case a loot bias might get applied to every single activity you do within EU. Hence mining loot might get corrected with hunting or vice versa.


To falsify one or more of the above mentioned hypothesis we might need datasets from different avas doing first one activity and then another one. Like 50 miners doing first 1000 enmatter drops and then 1000 ore. You see, this would be very hard to organize.

 

[Legion]

that's very probable. My personal opinion is that every activity within EU has its own pot, like enmatter is independent from ore, or hunting one mob is independent from another mob, but I can't prove it atm. Hence, all this is only a hypothesis and might get falsified.

I think this is correct that everything is sort of independent of each other, but it goes deeper.

Ore is separated from enmatter and hunting from mining but also each ore and enmatter are separated, and so is each area.

What i have come to believe lately is that everything in eu including your avatar is based on "biorythms" http://en.wikipedia.org/wiki/Biorhythm

How i think it works:

Each ore, mob, avatar, spawn has it's own "biorytmh" and this goes up and down just as explained on the website above.

Now if you go hunt and your hunting rythm is high and you find a mob and an area with low biorythms no matter what you do you will loot bad, but if you find a good area and the right mob voila. This is also then tied into the calculations falkao and the other math gurus has made here.

 

[Sircus]

So what does all that imply? Is it possible to profit when doing short runs? Yes and no. Due to the loot distribution that MA is using, those that play only for limited time and hence doing a lower number of drops, do have a 40% chance to profit from loot only.
(Hence we will have 40% noobs that are telling how able miners they are.) If those go on, their return rate will become lower and converge to the expected mean return. The contrary is also true. If those that were unlucky continue mining, then they will higher their return rate.

Taken from your first post, highlighted the text that I wanted to ask a question on.

So if you can see that you had bad runs, should one then keep increasing the lvl of amp untill you find something big to compensate? (Assuming there was no MU on amps)

 

[Jimmy B]

Taken from your first post, highlighted the text that I wanted to ask a question on.

So if you can see that you had bad runs, should one then keep increasing the lvl of amp untill you find something big to compensate? (Assuming there was no MU on amps)

This isn't really the correct interpretation of falkao's comment.

The expected tt return for mining activities, as estimated by this thread's data and analysis, is 95% of tt costs. If you mine a lot, you will likely end up near this result. If you drop just 100 bombs, you will probably not (you might profit quite a bit, you might lost quite a bit).

If someone starts out unlucky (say 40% return after 200 bombs) his overall return is very likely to increase as he continues. That doesn't mean he is expected to profit as he continues, he's expected to get 95% as he continues. And if he does get 95% from the next 1000 bombs, his overall return from the 1200 bombs will be much closer to 95% than 40% now.

So falkao is not saying if you lose 200 PED today, tomorrow you will probably get it back. On average tomorrow you are expected to get 95% back just like every other day. And so probably tomorrow your overall return will be better than 40% (even if you lose tomorrow).

 

[Sircus]

This isn't really the correct interpretation of falkao's comment.

The expected tt return for mining activities, as estimated by this thread's data and analysis, is 95% of tt costs. If you mine a lot, you will likely end up near this result. If you drop just 100 bombs, you will probably not (you might profit quite a bit, you might lost quite a bit).

If someone starts out unlucky (say 40% return after 200 bombs) his overall return is very likely to increase as he continues. That doesn't mean he is expected to profit as he continues, he's expected to get 95% as he continues. And if he does get 95% from the next 1000 bombs, his overall return from the 1200 bombs will be much closer to 95% than 40% now.

So falkao is not saying if you lose 200 PED today, tomorrow you will probably get it back. On average tomorrow you are expected to get 95% back just like every other day. And so probably tomorrow your overall return will be better than 40% (even if you lose tomorrow).

Ah I phrased my question bad, my keyword was not profit but the expected 95%.

Am I still asking the same question if I say: Amp does not add tt it multiplies? so by finding a big find later on, (to compensate for my current 40% return) then the multiplication would make a higher then 95% return temporarly?

 

[Jimmy B]

Am I still asking the same question if I say: Amp does not add tt it multiplies? so by finding a big find later on, (to compensate for my current 40% return) then the multiplication would make a higher then 95% return temporarly?

There is no compensating for your 40% return.

Whether you have 40% or 150% return, in future runs you are expected to have 95% return in the long run.

Let's take Person A who got 40% from 100 bombs today, and Person B who got 150% from 100 bombs today. Over the next 1000 bombs Person A is expected to get 95% back. Over the next 1000 bombs Person B is also expected to get 95% back. One of course may be luckier than the other, but that is just as likely to be Person B as Person A. But in the long-run both their returns will get closer and closer to 95%.

falkao is not saying that what happens today affects what happens tomorrow.

If you toss a coin 10 times and get ten heads, you have exactly the same chance of getting a head on the next toss as a tail. However in the next 100 tosses you will very likely have closer to 50/50 tails/heads. Let's say you do get 50 heads and 50 tails. From the 110 tosses you have 60 heads and 50 tails. So as expected you are much closer to 50% heads 50% tails than you were at the start when you had 100% heads 0% tails. The first 10 tosses didn't have any impact on the outcome of the following tosses though.

[Note: I'm not saying for that EU definitely has no 'memory'. But so far as I understand it, the data falkao has from his hunting study and the data from this thread both show no evidence of previous results affecting future results. Thus so far falkao's studies are in favour of Witte's point of view rather than Lavawalker's (starting from the top of this page). Could be there just hasn't been enough data to see memory yet, but this is what the evidence so far points to I believe.]

 

[FreezyPops]

removed.....