Menu
What's New
By:
Filters

FYI: How Loot Returns Work as Function of Looter Level and Efficiency, Based on Actual Data, as of March 3rd, 2022

Just wanted to say thanks very much for putting in so much time and effort to publish great clear-cut statistics. Inspired some great comments here too. From personal experience it seems very accurate and also confirms my personal experience between looter and efficiency and explains why my TT returns have actually improved moving from a 88% eff weapon down to 70% weapon as my looter levels have grown greatly during that time between the two different weapons, data points.
 
atomicstorm said:
...

This game at its heart is a PvP game. The phrase: don't teach the fishes - is very important here. Teach the fishes - and you become the suckerfish.
Click to expand...
One cannot argue against that. It is, however, the most complicated and elaborate pvp game outside of RL I'm aware of. You have to marvel at its engineering.
 
Our human mind i snot made for "extremely small" and "Extremely large" numbers.
a stupid example i made to colleagues in Italy:
if we put shoulder to shoulder the whole italian population can we expect to cover the fuil perimeter of the sea shores?
and in positive cases how many times?
The answer is surprising... many think we can not many say that we can one and a half times.
we got 59.300.000 (2021) ppl, 8300 Km coasts.... using just 50 cm average shoulder width...
the distance is made of (8.300 *1000 * 2) = 16.600.000 Peole... so 59300.000 / 16.600.000 = 3.50 times.

if we speak about virus, and i say "if we line up 100 Corona virus, can we see them ? answer 90% is "sure those little bastards"
biggest viruses are 150/1b Meter so in millimeters = 0.00015 ... 100 of them are 0.015 millimeter and we can not see it with bare eyes

The same works for large sample randomizers distribution.
Entropia has a randomizer cycle that is really long. ppl name the number 7... it can come in our mind that look for a pattern (we are programmed to find a pattern, to adapt our past experience to the future) as a "most common result of a 2 dice roll)... it can be a 7 state randomizer (there is a study on that on a post on a RT instence roll).

i wrote already "do not use Student T in entropia). the world of entropia is (statisticallyu speaking) Leptokurtic and etheroskedastic.
we have high peak and fat tails, and the model is Skewed to the left side (higher losses)-
It is like the distribution (canonical) of equity markets and its brownian motion is in some extent GARCH imho.

i was a professional option trader for 19 years and i experienced similar paterns.

Entropia is fractal, we ahve small, medium and large randomizer cycle. i am gaming jsut since 2 year and cycled less than 4m PED so my sample is relatively small but killing small creatures (under 3 ped kill cost) made me collect almost 1m loot instances and the Deception of "coin flipping" is a reminescence of the past.
the is a "dogwalk effect" and a "mean reversion effect" in the returns.

Baseline is that we need to aim to collect 3% average markup on kills and 6% in crafting.... OR sell lourt skills to balance the cost. each skill tick is about 1 pec value and game reward skill as part of the "return". if we account skill manufactured loot is about equal to decay (codex gifts us 0.5% to returns, iron mission "gifted" less for lower ones. Rocktropia, older iron we can finalize in other planets, traeskero, rex and leviathan and all storyline mission are "double dip" and lead to profit.

OP is solid numbercrunching but is overvalued because caps at 100... need to take off MA and PP slice... so we can expect game to be capped at 98.5...99% return (PLUS SKILLS).
i personally think that interaction between Looter and Eff and not just a sum (or a new player with a grindhouse should expect 85% returns....
but in this case it is "gut feeling" and falls into "deception" too.

Very interesting post. i take it as "expect return between 93 and 99" and keep shooting looking for some sellable MU :cool:
 
Without reading fully, surely you know the fractal question: How long is a coastline?
You're still in the free edit time window... ;) (nice to see a more in-depth use of knowledge being shared here though)!

The pounding heart of EU remains its loot mechanics. There have been, and are, glaring problems with the wave dynamics, but overall it has stood the test of time well, whether mobs, mining, crafting, harvesting...
What is then done with the loot - and thus its mu - is also a battleground ;)
 
This is good info. The difference between ~60% eff weapons and 75+ eff weapons was not as high as I had assumed. I can reasonably aspire to 100 looter....someday....but i cannot ever see myself in possession of a weapon equal to half my annual salary ;)
 
A couple pointers:
  • TT return is capped below 100%; no matter what gun you use. You can use Marco's dev gun and still get 98% TT return.
  • There is such thing as a personal lootpool, don't care what MA says. It safeguards against losses to much to quick and moreso it makes sure you don't get multiple multipliers in a short term. This is what they did to fix the mining amp exploit and they basically expanded it to all professions.
So while the scale might be true. If you fall below you are going to be compensated in the long run to get to 90% return and if you outperform it you will be capped.
 
Entropia Dao said:
There is such thing as a personal lootpool, don't care what MA says.
Click to expand...
MA chooses its words carefully. A bucket is not the same as a pool of water. What MA tracks IMHO is our tt. MA gives us smaller and larger buckets to fish out water with. We don't get our own pools, we get buckets, sized at whatever the algo tracking our returns determines. If we get fish or crabs when they are added to the pool (mu), MA doesn't care. They just care about the volumes we collect and add back in...
 
Entropia Dao said:
  • TT return is capped below 100%; no matter what gun you use. You can use Marco's dev gun and still get 98% TT return.
Click to expand...
I don't believe that is true. 99%+ is defo possible. Looter is still too new to draw any conclusion about what happens over level 100 with only handful of players above that level.
 
Last edited:
I'm even starting to believe theres no such thing as fixed tt return like a fixed number that your returns should converge to.
It may even work like: all people spent 1000 ped last minute. The next minute there are 990 ped to claim between all people based on their efficiency+looter.
 
This is how I see it: TT return might be capped to the current status of our collective loot pool.
If MA set the average predicted return to be collectively at 96% and it drop below this those above will probably be
able to get higher per centage individualy than what the collectively target is.
So if those below the set value improve their return, those at the top will drop some.
 
jetsina said:
MA chooses its words carefully. A bucket is not the same as a pool of water. What MA tracks IMHO is our tt. MA gives us smaller and larger buckets to fish out water with. We don't get our own pools, we get buckets, sized at whatever the algo tracking our returns determines. If we get fish or crabs when they are added to the pool (mu), MA doesn't care. They just care about the volumes we collect and add back in...
Click to expand...

It is interesting what you are saying.
It depends on the definition of "pool". If they determine a pool as a preset amount you can loot, then yes there is no pool in the literal sense.
But they do track our TT return and there is an adjustment of loot afterwards. This adjustment might just happen to multiplied loot, which would make regular loot follow the eff/loot profession curve. Only point is that regular loot without multipliers won't give you a good hunting run.

Maybe the biggest question for MA would be to ask the following:
"Are TT return and other expenses tracked and calculated when a loot event happens"

The answer will be yes. I've seen multiple indications of this.
It wasn't that way in 2008 for sure, but they started doing it by 2010.
 
Evey said:
I don't believe that is true. 99%+ is defo possible. Looter is still to draw any conclusion about what happens over level 100 with only handful of players above that level.
Click to expand...

Maybe, I am not high enough level, nor have a high eff. gun to check the upper limit.
I do know that on the other side of the spectrum we get compensated.
I can only assume that MA capped it at some point and maybe tweaked it after the lvl13 amp exploit in mining.

I follow the logic that MA must make sure nobody can exploit (either intentionally or by accident) the loot system, since that would cost them money.
Maybe return is capped at 99%. maybe at 101%. And maybe it is not a hard cap, but just an extra fractional multiplier whenever you get a multiplied loot.

In example. If the system would give you a 100x multiplied loot, it multiplies it with a variable based on tracked tt return. If you lost a lot, that multiplier would be higher then 1; if you had good loot it would be smaller then 1.
It would not affect most loot events which are between 0.25x-2x return on costs, but it would effect the few loot events that make or break a hunt.

From what I've observed and following MA's changes to the system over the years, it seems that they indeed would program loot this way.

Entropy would mean all information is kept in one form or another. Maybe there is an algorithm that keeps track of all expenses and returns.
This would explain certain things.

Ahh well, I hope that the source code (or at least the calculations) get released in the event EU ever stops being. I'm not a math geek, but I would surely look through this. It is like Entropia is one big riddle. :tiphat:
 
I'm not being facetious when I say this. I really do think a probability and statistics course for the serious Entropia player is a must, particularly if you are going to hypothesize how the loot system might work.
 
Jhereg said:
I'm not being facetious when I say this. I really do think a probability and statistics course for the serious Entropia player is a must, particularly if you are going to hypothesize how the loot system might work.
Click to expand...

I finished a college statistics class 6 months ago and I must say it won't tell that much about Entropia.

Also, the only player who I know has a mathematics degree is Star, and he's not playing that often anymore. I don't know if he sold most of his uber stuff or when he stopped playing, but it might telling something about loot 2.0
:twocents:
 
Entropia Dao said:
I finished a college statistics class 6 months ago and I must say it won't tell that much about Entropia.

Also, the only player who I know has a mathematics degree is Star, and he's not playing that often anymore. I don't know if he sold most of his uber stuff or when he stopped playing, but it might telling something about loot 2.0
:twocents:
Click to expand...
I bought some of his stuff a couple years ago. Are you a new player?
 
Oh you did a math course a while back? Cool, think about it this way then:

If you were to setup a system, lets just call it a gambling system, would you rather:
  • Not how you would code this, I know, but its easier to read if I say it like this, lets save the programming / optimization for stackoverflow.
  • Im just gonna make up nice round numbers here.
  • Lets just say there are only 3 buckets/kinds of multipliers cuz I dont wanna spend too much time on this.
  • Also, within those buckets we actually have a little variation, its not always 0.5*TT_IN = TT_OUT but maybe 0.6 but with that statistics course, you can prolly come up with tons of ways to add a little bit of fuzziness to those, maybe 0.5+RA_NUMBER/BUCKET_LIMIT, what ever.
A) Have a system that rolls a number, say 1 to 200.000.
Rolled NumberMultiplierTT_RETURNChanceExpected payout
1-190.0000.50.5 * TT_IN95%47.5%
190.001 - 199.9991010 * TT_IN4.9995%49.995%
200.00020002000 * TT_IN0.0005%1%
Total Return:98.495%
  • Explanation: On a 1500HP mob, that gives you a "global" ~ every 20 mobs, and a nice juicy, 10k hof every 200k mobs, which people tell me is a way too low sample size right now :devilish:
  • You get 98.495% Player Return in this system. MA can't lose, there is no way for you to exploit this. This is why online casinos, Roulette etc. use a system like this, very simple, basic but robust system.
  • You can tweak numbers, add buckets, what ever
  • You can add stuff too, math gets more complicated but add simple variables to add "Efficiency Impact" or you add a constant, root, mod or what ever to TT_IN. Tons of "easy" ways and the only thing MA needs to check is if their math still works out (rightmost column can't add up to more than 1), tho lets be fair, this loot formular adjustment is the only time they will actually run a simulation to beta test their shit. hrhr

B) OR would you rather have a system that also rolls a number, say 1 to 200.000, but keeps track of where your Player Return is at and adjusts, either your chance to get a multiplier OR the size of the multi. Both of these have sub-options but end up the same. Its just a sliding scale that dynamically changes one or the other variable.
It would make sure someone is less likely to hit big quickly in a row or loose too much. Keep in mind the above example tho. I tried making a nice and simple table for this like above but it got confusing and I dont wanna spend forever formatting something for a forum. A piece of paper / white board would be so much easier. But one can convince one self fairly easily that this is the "loot pool"-theory, right?
Like I said I didnt manage to write this down cleanly for a forum, but if you didnt HAVE to take that statistics course, but wanted to or maybe had fun, then this is a nice little exercise, at least it was for me. Trying to write this down and figure out the expected payout with some nice round numbers? Actually a little tricky, at least for me.

Now, the first one is so basic, that there simply is no exploit. Raising your bet at any point? No problem, since payout out assumes infinite rolls, what ever you raise just gets eaten by that infinity, mathematically speaking. E.g. Betting on all the allowed roulette fields, yields what 97% player return? Same with betting Black/Red - doubling your bet thingy, if I remember right.

Not the second one tho, that is more akin to black jack + basic strategy + counting cards. There, you cleverly raise the bets when your deck is "hot", thus beating the house. That is also why that can't be illegal but they will ask you to leave and never come back. The reason that works is because they didnt used to have infinite decks so you could count cards and raise the bets when you were likely to win. For a computer program tho, infinite decks are no problem, just pick another random number. Another can of worms I know, lets not open that random can right now tho. Practically, you would figure out a way to know when you are "due a correction" as people have called it here, then go to leviathans and get yourself some TT profit. And since we are on a computer, there are plenty of very unsophisticated ways to figure out if you are due a correction, and dang, our crafter friends clicking the EP4s woulda figured that out, at least I believe that.


Now, I don't know Swedish law or if PE even falls under that or maybe EU law but either way, it could be that MA tries to follow gambling laws in some country just to be safe. Eventho they say they aren't a casino. But take German/EU law for example. There are laws for online casinos which say (I forgot the number and, was that in malta? cuz thats where all those casinos are at, for us in the EU?) 92% must be returned to the player. That is a theoretical number tho. That assumes infinite "rolls" to figure out the players loss. Lotto for example has 50% (no math needed, they just give back 50% in the "loot pool", rest is for the govmt/ads or what ever), for fun, you can calculate the return to player for blackjack with basic strategy. It is actually fun, cuz not super trivial, also actually look up how card counting works, surprisingly, the complicated thing isnt the counting but the math behind raising your bet without it being obvious to the casino. Anyway, you CAN have a "minimum" without a loot pool or tracking your current player return. You can also follow the law this way and most importantly, you/the player cant tell the difference UNLESS they try to beat the house and fail. Which leads to my point, if you were MA, why would you do a system with only down sides for you in which you could actually be beat? After all, counting cards doesnt work with an infinite amount of decks, which MA has, cuz its all digital.

I kinda wanted to go for the extra credit and give you a quick python script to simulate system 1 that gives you a nice chart after X rolls. But its time to go to work. I just think we humans are too prone to seeing patterns where there potentially is another explanation, so I posted this. Tho I think its like the 24th of this kinda post on this forum but w/e.

cheers
 
Last edited:
Great thread, i have linked it to many peoples already when we were talking about loot.
Also my numbers after 750k cycle is about matches to the expected return from the last table.
 
Thark said:
I kinda wanted to go for the extra credit and give you a quick python script to simulate system 1 that gives you a nice chart after X rolls. But its time to go to work
Click to expand...
I already made that script :)
You can find it here: https://github.com/opentropia/pytropia/blob/master/loot-sim.py
The actual multipliers and probabilities in the script are based on actual data, around 200K of individually tracked kills.
The script also has a model for "bonus shrap" and to scale the loot based on looter/eff.
I still wouldn't call it accurate since 2000x multis are very rare (like 1 in 200000) so to get accurate models millions of (individual) kills needs to be tracked.
It's still pretty fun to run it and experiment. You can easily see the "kickbacks" you get from pure randomness :laugh:
 
Great work and great conversation.. I might have to test whether my loot return changes when I am not hunting and drinking :). My guess is that I am not taking the game nearly as serious when drinking as I do when not drinking. All joking aside, I enjoyed the read
 
Great info OP and love the Data you collected but you forgot some very important Looter stats.

1. If wearing my favorite pajama's (shown to increase multi's by 13.33%)
2 The secret hidden Stamina multi (once over 120 stamina you really get the swirls) :woot:
 
Jhereg said:
A quick preface.

There's a lot of bad information out there. As with everything, take what you read with a grain of salt. I've decided to share this, mostly because I think it would be fun to see what reactions I'm going to get out of this. But I've started to lose motivation for playing this game, because it has become an endless grind and because it was no longer about knowledge and science, but beliefs. "In God we trust, all others must bring data."
There are those who think knowledge is power and should be consolidated to lift oneself above the others. I think knowledge is power, but can be used to lift everyone up instead. The data below was the culmination of about two weeks of work and subsequent analysis, and a couple particularly helpful lightbulb moments. It was hard work, but without some critical thinking, we might still be at the "We need a million data points in order to get any real idea of your returns" phase.
Thanks to @miathan for providing the python script for parsing and tabulating the data, and @minim for their input in data as well.

Motivation: To determine the impact of efficiency and looter level on tt returns, and tease out any obvious trends.

Procedures:

Weapons with three different efficiencies were selected (92.5%, 62.8%, and 29.5%). Each weapon was run on three different mobs (bristlehog, drones, thorifoid berserkers) where I was able to hunt without armor and healing (using lifesteal). The distribution in multipliers were extracted and analyzed. The main region of interest of the distribution fell beneath 1.0x, so see the figure below for an example of such a distribution:
j21TbWV.png

As you can see, the multipliers distributions have groupings. By analyzing the mean and standard deviation of the most populated grouping (Between ~.33x and .63x), we can try to find some trends in looter/efficiency behavior...

Number of kills (samples) performed in each case is shown in the table below:
6yBTiUK.png


While I was able to get some really decent results by analyzing only certain multiplier groupings, or certain fractions of the multiplier groupings (e.g. the bottom 2% of multipliers in a group), it did become apparent that I would need maybe quadruple the number of data points to get data that was going to be clean enough. While this is not on the order of 1 million kills, this was still very time consuming and I wasn't about to throw another 5 weeks into this project if I didn't have to.
At this point, I realized that instead of trying to statistically analyze the mean/sigma of the distribution, I could take the very bottom multiplier I got out of the entire sample. and this generally behaved extremely nicely and within my expectations for pretty much all the samples. While not as statistically robust, I found this method to give me the best results, so this is what I went with. The resulting data is in the following figures and explanations.


fM4Delm.png


As you can see, at all the looter levels considered here, efficiency scales linearly. The slope here represents the impact of efficiency on the multiplier. If we take the slopes and divide by the y intercept, we get 7.25%. As MA has stated before, the expected efficiency effect is about 7% full range, so this value seems to align well with what MA has suggested before, as well as my previous turret testing when extrapolated out to 100%.

Teasing out the impact of looter is a bit more complicated but as we can see here, there's an obvious effect of looter levels on multipliers and it is very apparent. We can take the intercept and plot them on another plot. This would effectively represent looter impact at 0 efficiency:
DMWrF4v.png

I was quite happy to see that the points all fell on a line. I had some doubts about whether all the professions for looter fell on the same scale, but would seem that they do, which simplifies testing. To consider the effect of looter up to lvl 80 animal looter which is what I have, I take the slope here (.0002) (.000167), multiply by 80, and divide by the expected multiplier for animal looter at eff=0 (.2335) based on the previous plot. When you do that, you get 6.9% 5.73%. So the expected improvement from 0 to lvl 80 looter is ~7% 5.73% TT return, assuming this linear behavior occurs over the entire range. If we take this value and scale it for the entire 0-100 looter level range, it comes out to about 7.2%. We have no expectations from MA of what looter does so I have nothing to compare this against. But I can say that my animal looter represents a ~1.65% better tt return than robot looter...

It is quite interesting that it appears the impact of looter level and efficiency are approximately equal in scaling between 0 and 100...

Discussion (Take these with a grain of thought, as some of this is based on my opinion of what the looter system might look like extrapolated out to higher levels.)

Based on this data, it appears that both efficiency effect and looter effect are linear. In addition, robot, mutant, and animal looter all appear to lay along on the same scale.

Assuming MA had foresight based on what problems loot1.0 caused, I'd expect some capping eventually of looter level effect. (No proof of this, until I can get to higher looter level). But based on other aspects of the game having lvl100 as a max (eff, weapon reqs), lvl 100 doesn't seem to be unreasonable. Again, I'd have to test this as my looter level moves up.

Just for shits and giggles, I assumed at lvl 100 + eff 100 is 100% TT return. I built a table with either the impacts either being multiplicative or additive ( so looter effect times efficiency effect versus looter effect plus efficiency effect). Here's what it might look like...totally just for fun. While this represents the trend, I have no idea what the tt return should be at 100 looter and 100% efficiency so this is just a proposed possibility

<ERRATA'ed: NEW TABLE, taking into account 7% effect over looter range 0-100 instead of 8.5%...)
eOlJj3H.png



Hopefully, this will help elucidate some stuff about eff and looter.

Many others who have been willing to take the time to collect the data, and then to share it with the player base have come before me, and I hope there will be others who share this path with me. I hope to work with others to understand better the mechanics in EU in the future.

Yours truly,
Zho

20220303: Errata: Adjusted looter vs multiplier line to show larger number of decimal places, which reduced the amount of looter prof effect on tt return. I have updated the post to reflect the updates.
Click to expand...
Amazing post! Thanks!!!

Not saying it's true but would be smart of MA to give close to 100 returns for the first 10-15 levels. Hard to test but that would help get people hooked on the game at the start... 🙈
 
jak said:
Amazing post! Thanks!!!

Not saying it's true but would be smart of MA to give close to 100 returns for the first 10-15 levels. Hard to test but that would help get people hooked on the game at the start... 🙈
Click to expand...
"Great post! I know you actually tested and recorded data, but heres a random made up idea which goes against that data, which I didnt test whatsoever, that I still feel the need to post in response."
 
The Abomb said:
Great info OP and love the Data you collected but you forgot some very important Looter stats.

1. If wearing my favorite pajama's (shown to increase multi's by 13.33%)
2 The secret hidden Stamina multi (once over 120 stamina you really get the swirls) :woot:
Click to expand...
posts like this are why you are the people's champ
 
jak said:
Amazing post! Thanks!!!

Not saying it's true but would be smart of MA to give close to 100 returns for the first 10-15 levels. Hard to test but that would help get people hooked on the game at the start... 🙈
Click to expand...
People would just cycle new avatars to exploit this.
 
Xen said:
People would just cycle new avatars to exploit this.
Click to expand...
Would they though? A noob even at 100% TT isn’t going to generate anything substantial. Muleing is pretty easy to detect unless you want to put a whole lot of effort into disguising it, and still wouldn’t be worth the risk in breaking TOS and giving MA a reason to suspend a withdrawal.
 
  • Like
Reactions: jak
You must log in or register to reply here.
Back
Top