Menu
What's New
By:
Filters

Analysis of how Shared Loots are Divided between team members

Serica

Serica

Moderator
Moderator
Joined
Nov 1, 2006
Posts
5,356
Location
Australia
Society
Antipodean Army
Avatar Name
Harena Serica Turbinis
MindArk has previously stated:
The Shared Loot feature shares the same loot mode setting as the team-option "Damage: Share Stack".
Click to expand...

Hanne|SDS futher stated:
Hanne|SDS said:
...
How it works:

Player A does 60 % Damage on the mob,
Player B does 30 % Damage on the mob,
Player C does 10 % Damage on the mob:

The probability is the same for all items in the loot. So for the next item, Player A still have 60% chance, Player B 30% and Player C 10%.
...
Click to expand...

Following the discussion here, and as noone else seems interested in taking it on as a project, Fishface and I will be doing some basic research into the mechanics on how teams loot are divided up.

In order to minimise the variables, we agreed on the following parameters:

Scope of the Project
  1. 2 person team (Fishface and me), set on Damage: stack Share (MA have said that the sharing is exactly the same as in Shared Loot teams)
  2. 10 looted mobs per hunt (any more is just tedious to spend time analysing at a time, and we're not interested in 'no looter' stats)
  3. We exchange chat log files via skype at the end of each session, so that we can both see the damage each of us inflicted.
  4. Mob chosen: Cornundacauda (Low HP, and neither of us have finished the iron mission on these)
  5. Weapon chosen: unamped Opalo (to minimise overdamaging)
  6. Mobs to be agreed and targetted one at a time, so no crossover of damage or loot in chat logs to confuse the issue.

We'll be posting the compiled data from each session here, and welcome any of the maths/programming guru's here to want to spend time analysing it too.

We'd appreciate if posts are be limited to reasoned discussion backed up by hard data, not wild speculation on numbers pulled out of thin air :)

Note to Mods: We'd appreciate you keeping an eye on the thread from time to time, and helping to keep it clean. :wise:
 
Current Hypothesis:
  1. When a mob is killed, MA calculate the damage inflicted by each player.
  2. Loot (if any) is then determined ( via "MA LOOT BALANCING BLACKBOX"), including total number of stackable type to be distributed.
    (Note: How MA chooses the items to be looted, or their total value, is NOT within the scope of this project, just how those items are distributed between team members)
  3. Non-Stackables:
    The non-stackable items are divided in the way Hanne has stated above:
    Player A does 60 % Damage on the mob,
    Player B does 30 % Damage on the mob,
    Player C does 10 % Damage on the mob:
    .. Items are shared where the probability are 60% that Player A gets an item, Probability is 30% for Player B and 10% for Player C.
    Click to expand...
  4. Stackables: This is the bit noone seems sure of, and that we're hoping to confirm...
    Stackables are shared according to the % damage inflicted.
    Click to expand...

    We think MA have used much the same algorithm as they do in working out mining loots
    (see Noodles post here)

    Stealing freely from that post (to save finding new words to express the same concept), our current hypothesis is:
    • For each type of stackable MA works out the total number of items to be distributed among the team members (see MA LOOT BALANCING BLACKBOX above)
    • They then calculate the nominal lootsize that each player gets (expressed down to 5 decimal places), as TT value of stack * % damage inflicted.
    • The nominal loot size is rounded to a whole number of resource units.
    • The rounding is done probabilistically, with more chance given to the number of stackables that is closest in value to the nominal lootsize.
    • Unlike mining however, team loots involve multiple players, and after point 3 there may be a remainder of 1 or more units of the stackable still to be distributed.
      Rather than coming up with a different algorithm for this, we suggest MA loop thru the above distribution process again as often as is needed to fully distribute the stack.
 
Last edited:
Code: 
Cornundacauda													
Maturity	Mature	Young	Old	Old	Old	Old	Mature	Prov	Old	Old		Totals
[B]Damage Done[/B]													
	Fish	85.3	57.5	97.8	99.6	91.7	98.6	110	129.6	92.5	113.2		975.8
	Serica	99.7	100.9	110.9	111.8	120.9	111.1	79.3	110.9	115.3	92.6		1053.4
		185	158.4	208.7	211.4	212.6	209.7	189.3	240.5	207.8	205.8		2029.2
[B]%Dam.done[/B]													
	Fish	46.11%	36.30%	46.86%	47.11%	43.13%	47.02%	58.11%	53.89%	44.51%	55.00%		48.09%
	Serica	53.89%	63.70%	53.14%	52.89%	56.87%	52.98%	41.89%	46.11%	55.49%	45.00%		51.91%
													
[B]Loot Received	[/B]												
Ammo (#)													
0.0001	Fish	582			1849		26213	882	2737	1055			33318
	Serica	587			2139		29815	600	2291	1317			36749
Coloniser Foot Guards	 (TTvalue)												
	Fish		1.2										1.2
	Serica		0										0
Bigwig Board (#)													
1.13	Fish						1						1
	Serica						1						1
Surface Hardener Component (#)													
0.2	Fish			1	1								2
	Serica			2									2
Wool (#)													
0.2	Fish			1	1						1		3
	Serica			2							1		3
Socket I Component (#)													
0.1	Fish								3				3
	Serica								3				3
Animal Muscle Oil (#)													
0.03	Fish	4			2		106	5		3	1		121
	Serica	4			1		122	4		3	1		135
Animal Hide (#)													
0.01	Fish			2									2
	Serica			2									2
Animal Oil Residue (#)													
0.01	Fish	1			1				1				3
	Serica	1			1			1			1		4
Nova Fragments (#)													
0.0001	Fish					11							13
	Serica					13							11
													
TT Value													
	Fish	0.1882	1.2	0.42	0.6549	0.0011	6.9313	0.2382	0.5837	0.1955	0.23		10.6429
	Serica	0.1887	0	0.82	0.2539	0.0013	7.7715	0.19	0.5291	0.2217	0.24		10.2162
	Total 	0.3769	1.2	1.24	0.9088	0.0024	14.7028	0.4282	1.1128	0.4172	0.47		20.8591
													
% as a total of TT													
	Fish	49.93%	100.00%	33.87%	72.06%	45.83%	47.14%	55.63%	52.45%	46.86%	48.94%		51.02%
	Serica	50.07%	0.00%	66.13%	27.94%	54.17%	52.86%	44.37%	47.55%	53.14%	51.06%		48.98%

Points worth noting:
  • There was only 1 non-stackable item looted. Fish did 36.3% damage on that mob, I did 63.7%... but he got the loot.
  • We were both using exactly the same weapon, but my share of damage ranged from 41.89% to 63.7%, Fish's ranged from 36.3% to 58.11%. Looking at the raw data, the differences are mostly in the number of 'MISS' shots, and the number of crit shots each of us had on a particular mob.
  • On Mob 4, I did more damage (52.89%), but Fish looted a higher # of muscle oil that me, he also got a couple of extra single stackables than I did.
  • On Mob 10, Fish did more damage, but I looted an extra Oil Residue, and he didnt.
 
Last edited:
Serica said:
MindArk has previously stated:


Hanne|SDS futher stated:


Following the discussion here, and as noone else seems interested in taking it on as a project, Fishface and I will be doing some basic research into the mechanics on how teams loot are divided up.

In order to minimise the variables, we agreed on the following parameters:

Scope of the Project
  1. 2 person team (Fishface and me), set on Damage: stack Share (MA have said that the sharing is exactly the same as in Shared Loot teams)
  2. 10 looted mobs per hunt (any more is just tedious to spend time analysing at a time, and we're not interested in 'no looter' stats)
  3. We exchange chat log files via skype at the end of each session, so that we can both see the damage each of us inflicted.
  4. Mob chosen: Cornundacauda (Low HP, and neither of us have finished the iron mission on these)
  5. Weapon chosen: unamped Opalo (to minimise overdamaging)
  6. Mobs to be agreed and targetted one at a time, so no crossover of damage or loot in chat logs to confuse the issue.

We'll be posting the compiled data from each session here, and welcome any of the maths/programming guru's here to want to spend time analysing it too.

We'd appreciate if posts are be limited to reasoned discussion backed up by hard data, not wild speculation on numbers pulled out of thin air :)

Note to Mods: We'd appreciate you keeping an eye on the thread from time to time, and helping to keep it clean. :wise:
Click to expand...

Seems pretty simple to me. Damage stack share is just that, stackables are sorted by damage. shrug easy peasy. Items are a different story. Are they split randomly? Are they included in stack share as to TT value?

In a team everyone see's how the items are allocated, and normally there has been an agreement made as to how they are split.

Shared loot.... no one knows anything, no idea how items are allocated, nor stackables so far as I know. How about it MA how is this done?
 
Shimmer said:
Seems pretty simple to me. Damage stack share is just that, stackables are sorted by damage. shrug easy peasy. Items are a different story. Are they split randomly? Are they included in stack share as to TT value?

In a team everyone see's how the items are allocated, and normally there has been an agreement made as to how they are split.

Shared loot.... no one knows anything, no idea how items are allocated, nor stackables so far as I know. How about it MA how is this done?
Click to expand...

Noone knew how armor worked either, til players got together the data and worked it out.

This thread has been created in that same spirit of dispassionate research.
 
Serica said:
Noone knew how armor worked either, til players got together the data and worked it out.

This thread has been created in that same spirit of dispassionate research.
Click to expand...

Dang Serica, you must be kidding. How certain are you of all those old armor tests today? It's become silly in my opinion. The "secrets" of the loot system should not extend to basic questions of % of loots given in a number of circumstances.

People playing this game shouldn't have to rely on things like Entropedia or Tracker. These should be published figures by MA at this point. I can see the need in the past though.

I shall bow out of this thread now Lady Serica, only for help you have given me in the past. The past is the past though, and MA needs to bring itself up to now, much less the future.

Bow
 
Shimmer said:
Seems pretty simple to me. Damage stack share is just that, stackables are sorted by damage. shrug easy peasy. Items are a different story. Are they split randomly? Are they included in stack share as to TT value?

In a team everyone see's how the items are allocated, and normally there has been an agreement made as to how they are split.

Shared loot.... no one knows anything, no idea how items are allocated, nor stackables so far as I know. How about it MA how is this done?
Click to expand...

For the most part, I agree with what Serica has said in the OP.

as far as brain oil is concerned, I can confirm, from the Sand King experience, it is not distributed as per damage... even higher damage person can get lower number of brain oil even though it's stackable.
Someone has suggested that brain oil is treated like an item!
 
Infinity01 said:
For the most part, I agree with what Serica has said in the OP.

as far as brain oil is concerned, I can confirm, from the Sand King experience, it is not distributed as per damage... even higher damage person can get lower number of brain oil even though it's stackable.
Someone has suggested that brain oil is treated like an item!
Click to expand...

That's because you live in Shaping's world, Serter's world is different as is Crystal Man's.

yea yea I said I'd go out :(
 
Infinity01 said:
..

as far as brain oil is concerned, I can confirm, from the Sand King experience, it is not distributed as per damage... even higher damage person can get lower number of brain oil even though it's stackable.
Someone has suggested that brain oil is treated like an item!
Click to expand...

No, it's perfectly possible to see how that split can happen.

You said in the earlier thread that someone using an apis+beast got 3 brain oil, and you got 1 while using apis+dante.

Firstly .. while you have compared weapons used, you have no idea of the actual damage done to that particular mob by each of you. Even using exactly the same weapon, Fish and I did not each do the same actual damage. Misses and crits .. and the individual effect of lag .. can make quite some difference to 'expected damage' and 'actual damage'. As I said in that earlier thread, the only way you can truly make any comparison is to have both parties count the actual damage recorded in chat.log files (and cross-check to verify).

Secondly, I'll show how that 3 brain oil vs 1 brain oil can work, even if you had each averaged the damage of your respective weapons, using our current hypothesis laid out above:

PERSON A: Effective dps of Apis/Dante = 59.5
PERSON B: Effective dps of Apis/Beast = 52.5
Total effective dps = 112
so Apis/Dante does (59.5/112) 53.13% of damage, Apis/Beast does 46.87% of damage.

MA decides the total brain oil for the mob.
You're only 2 out of a large number of people, and you have no idea how many brain oil were looted in total.
So let say, after totalling each player's damage done to the mob, and working out each players nominal tt share of the brain oil, A's share based on damage comes out at 1.6468 units, and B's nominal share comes out to 1.4531.
This is then rounded probabilstially with more chance given to the number of stackables that is closest in value to the nominal lootsize.
However, on this particular loot, A's nominal share is rounded down to 1, and B's is rounded up to 2.
There's still a few brain oil left to allocate after that first pass tho, so it goes back thru the loop to apportion the remainder.
This time, A's nominal share of the remainder is .5313 and B's is .4687.
Unfortunately for A, once again his nominal share is rounded down (this time to 0), and B's is rounded up to 1.

Outcome: A receives 1+0 = 1 brain oil in loot, B receives 2+1 = 3 brain oil in loot.
 
Hey,

just like i wrote in the last thread you are treating this unnecessarily complicated. And because you are too fond of the old time experiments you missed all the points the others and I made about how this can be treated. (don't get me wrong, I too am thankful of those finds especially the armor and ming stuff but they don't have anything to do with this what so ever!)

I can see two possible ways stackables, e.i. ammo and low tt oils, hides etc. could be distributed, given that the system works at least close to what MA says.

1.) The total damage is calculated and then distributed accordingly, so 100 HP mob and 60/40 total dmg will net one player 60% off one kind of stackalbe and the other the remaining 40%. In this case rounding only makes a minimal difference. So if only 2 of one stackable are in the loot, both players should each get one!

2.) The relative damage of every participant is calculated and that is then the probability of each player to get 1 part of the stackable. That would mean for a 100HP mob and a 60/40 dmg ration and 100 eye oils to be given out: the first eye oil goes to player A(60%dmg) with a probability of 60% and to player B with a probability of 40%, then the next eye oil, until all is gone.


Option 2 can net you big differences if only small amounts of some stackables are in the mob. Just as we all have seen in team hunts before. If the stack of items is big enough though (like ammo for example) then the difference is statisticly very small. Option 1 can't explain your data, nor can it explain the brain oil issue, nor other stuff everyone sees on team hunts. Also option one could lead to a lot of of cases where special rules would have to be coded. Like the stack of 2 items, 2 payers and 90/10 dmg ratio. Why not run stackables through the same system as "items". "" because i actually think, with the support of your data, all parts of a stack are treated as an item.

My conclusion: I can not proof that option 2 is correct, because i don't want to conduct an experiment that can but i can, with my observations and now your data, show that option 1 is incorrect. That is enough for me. If you want to go further, try to run an experiment that tries to find out if option 2 is right. (why would you choose small mobs?)



Best Regards
Thark

PS: if i made a mistake somewhere, please point me to it. I tend to sometimes interpret stuff the way i want it to be, i don't see me doing this right now, but ofc if i did i would have writen something else xD
 
So basically, you are saying: everything is divided as described, any remainders are decided according to the rules of item split on those rules? Which distorts it a lot if the stackable to be divided is a relatively high TT.
 
Thark said:
Hey,

just like i wrote in the last thread you are treating this unnecessarily complicated. And because you are too fond of the old time experiments you missed all the points the others and I made about how this can be treated. (don't get me wrong, I too am thankful of those finds especially the armor and ming stuff but they don't have anything to do with this what so ever!)

I can see two possible ways stackables, e.i. ammo and low tt oils, hides etc. could be distributed, given that the system works at least close to what MA says.

1.) The total damage is calculated and then distributed accordingly, so 100 HP mob and 60/40 total dmg will net one player 60% off one kind of stackalbe and the other the remaining 40%. In this case rounding only makes a minimal difference. So if only 2 of one stackable are in the loot, both players should each get one!

2.) The relative damage of every participant is calculated and that is then the probability of each player to get 1 part of the stackable. That would mean for a 100HP mob and a 60/40 dmg ration and 100 eye oils to be given out: the first eye oil goes to player A(60%dmg) with a probability of 60% and to player B with a probability of 40%, then the next eye oil, until all is gone.


Option 2 can net you big differences if only small amounts of some stackables are in the mob. Just as we all have seen in team hunts before. If the stack of items is big enough though (like ammo for example) then the difference is statisticly very small. Option 1 can't explain your data, nor can it explain the brain oil issue, nor other stuff everyone sees on team hunts. Also option one could lead to a lot of of cases where special rules would have to be coded. Like the stack of 2 items, 2 payers and 90/10 dmg ratio. Why not run stackables through the same system as "items". "" because i actually think, with the support of your data, all parts of a stack are treated as an item.

My conclusion: I can not proof that option 2 is correct, because i don't want to conduct an experiment that can but i can, with my observations and now your data, show that option 1 is incorrect. That is enough for me. If you want to go further, try to run an experiment that tries to find out if option 2 is right. (why would you choose small mobs?)



Best Regards
Thark

PS: if i made a mistake somewhere, please point me to it. I tend to sometimes interpret stuff the way i want it to be, i don't see me doing this right now, but ofc if i did i would have writen something else xD
Click to expand...

I am pretty much your option 2) beliver: that each item (even single weapon cell) has a chance of being looted proportional to dmg done by team participants.

Although i think it is NOT implemented this way (but simulates this). Why i think so: imagine that when team loots a lot of small items (like ath with +200k oils inside) and for each item there is a lottery tick to decide who shoudl get what stackable item. With items with lower value than 1pec (like amo for example) it is even worse. This would lead to situation that server (client is not alowed to do this) side would have to count countless random ticks for each item looted. Multiply this by all teams curenlty hunting (and recently shared loot event waves) and you will get that server has to count few biliards things per sec (which is not possible even with fastests computers) and it has a lot of other work to do too ;).

Imho it is simulated for all stackables (with proper formula with random factor) to act like the option 2). And this can be counted with limited operations or proportional linearly to number of members in team. But still from our point of view it would look like random distribution proportional to damage dealt ;).

Falagor
:bandit:
 
Nice test.

I was about to suggest stackables might also be distibuted randomly with a weighting based on percentage of damage done. But Thark beat me to that.

It is tricky to calculate the damage done though. Should the killshot be counted for the total reported in chat, or just the HP remaining in the mob prior to the killshot? Also, how is regeneration counted? I did a test with Oleg once, where we teamed, he put a few shots into a mob and then we let it regenerate and I killed it. Each time we did that I got all the loot. When I killed it before it fully regenerated, Oleg got some loot.

I think there may be value in doing a further test similar to what you've done here, but on something low regen and big like Leviathon, with one team member using a gun about half the dps of the other. Or maybe another on corns but where one person only fires one shot. Just to see what happens in more extreme situations.

Shimmer said:
How certain are you of all those old armor tests today?
Click to expand...

If you mean the armor tests documented in the thread in my sig, it is still all valid today. I can say that with near certainty following some tests I did when Rextelum recently were implemented.
 
Last edited:
Interesting research, subscribing...

Don't forget to test the Shared Loot distribution in both solo and team, as it may differ.
 
Thanks for the responses so far.

Just for now, we're only at the data collection stage. We've limited the scope as described to try to make it both manageable and consistent as it will be collected across a period of time.

One premise in our hypothesis is that programmers will tend to use similar algorithms to do much the same things - which is why we think there is likely to be a similarity between shared hunting loots, and what has already been put together to account to the calculation of mining loots after tax.

Another is that programmers are more likely to use a loop than a whole bunch of fresh code to account for the remainders after the first allocation.

In both cases, it's an Occam's Razor premise - the simplest option that makes the fewest new assumptions.

Thark said:
... 2.) The relative damage of every participant is calculated and that is then the probability of each player to get 1 part of the stackable. That would mean for a 100HP mob and a 60/40 dmg ration and 100 eye oils to be given out: the first eye oil goes to player A(60%dmg) with a probability of 60% and to player B with a probability of 40%, then the next eye oil, until all is gone. ...
Click to expand...

As Falagor said, the sheer number of loops that would need to be done to implement your hypothesis, for all stackable items looted by teams, would put an untenable load on the servers and also, imho, simply take too long to run for proper gameplay.

JimmyB said:
It is tricky to calculate the damage done though. Should the killshot be counted for the total reported in chat, or just the HP remaining in the mob prior to the killshot? Also, how is regeneration counted?
Click to expand...

There is no way to accurately measure the exact HP remaining in the mob prior to the killshot. Whatever we try would only be an estimate. So, we've tried to keep it as simple as possible by both of us using unamped Opalos (average damage of 6).
We could drop it down further to use say a KiwioII (average damage of 1.5), and might do so if the Opalo shows a wide variance in damage done, but it would add to the tedium of the data collection and recording. :girl:

I did consider looking at regeneration, even if only to update Entropedia for those mobs ... and that's when I noticed that the timestamp is quite a few seconds different between our chatlogs :) I could sit there and try to normalise them all, but I felt it was (at this stage) beyond the scope of our project.

JimmyB said:
I think there may be value in doing a further test similar to what you've done here, but on something low regen and big like Leviathon, with one team member using a gun about half the dps of the other. Or maybe another on corns but where one person only fires one shot. Just to see what happens in more extreme situation
Click to expand...

Certainly there would be value in other tests being done too.

And I hope that someone will do them, record the data from all team members and post the results here.

Terminator said:
Don't forget to test the Shared Loot distribution in both solo and team, as it may differ.
Click to expand...

If you mean by testing in a wave spawn, I can't see any realistic way of obtaining chat logs from every participant in order to collect data on individual damage done to compare to loots received.
Feel free to try tho, and post the results here when you do :)
 
I dont know if this has been considered:

If player A) Went hunting before the test and lost 50 ped. (eg shooting Hogglo's and got nothing much back)

If player B) Went hunting before the test and gained 50 ped. (eg shooting Drones and globalled)


Then player A & B team together and record how the loot is allocated in an experiment.


This is (IMO) an added and important variable which will make it difficult to measure how team loots work.


In summary, players come with history which will in part determine what they will loot.
 
Mega said:
I dont know if this has been considered:

If player A) Went hunting before the test and lost 50 ped. (eg shooting Hogglo's and got nothing much back)

If player B) Went hunting before the test and gained 50 ped. (eg shooting Drones and globalled)


Then player A & B team together and record how the loot is allocated in an experiment.


This is (IMO) an added and important variable which will make it difficult to measure how team loots work.


In summary, players come with history which will in part determine what they will loot.
Click to expand...

We've all seen days where a particular player is 'hot'. I particularly recall the day Fishface got a HOF in all categories then available :)

In my experience, team hunting with a player who has the Midas Touch will increase the overall loot of the team (specifically how that happens is part of that MA BALANCING BLACKBOX that we're not looking into), but I do not believe that it would mean that the 'hot' member gets a greater proportion of the stackables than the damage they inflict.

For now tho, we'll just see what the recorded data reveals over time.
 
Fabulous thread!

Thark said:
2.) The relative damage of every participant is calculated and that is then the probability of each player to get 1 part of the stackable. That would mean for a 100HP mob and a 60/40 dmg ration and 100 eye oils to be given out: the first eye oil goes to player A(60%dmg) with a probability of 60% and to player B with a probability of 40%, then the next eye oil, until all is gone.
Click to expand...

It is not hard to show that this option 2, while a reasonable hypothesis, is incorrect. Several years ago I took a very small (unpublished) data set, in a less rigorous manner than Serica is doing. If you simulate option 2 (Monte Carlo-ish), it clearly shows that the variance that would be present is greater than the actual variance observed. So I ruled out option 2, and I'm sure it could be ruled out using these data also. Or I'll eat Jimmy's hat :dunce:

My thought is that, while Serica may seem to be making this more complicated than it is, it is probably about as complicated as she makes it to be. Why, I have no idea, but it's probably as she suggests, that the code is already there to use.

Serica, I do have a thought about your hypothesis...when rounding probabilistically, what happens if everyone rounds up, and you have to distribute more stackables than are given to the loot? Is the loot changed? That's what seems to happen in mining. Also in mining, it is quite possible that the final loot value can be decreased due to probabilistic rounding. Could that happen here? How would it change your proposed algorithm? Much easier to tease out the answers to these questions in mining compared to hunting...

Keep chugging with these experiments, the results will be quite interesting!
 
Serica said:
No, it's perfectly possible to see how that split can happen.

PERSON A: Effective dps of Apis/Dante = 59.5
PERSON B: Effective dps of Apis/Beast = 52.5
Total effective dps = 112
so Apis/Dante does (59.5/112) 53.13% of damage, Apis/Beast does 46.87% of damage.

MA decides the total brain oil for the mob.
You're only 2 out of a large number of people, and you have no idea how many brain oil were looted in total.
So let say, after totalling each player's damage done to the mob, and working out each players nominal tt share of the brain oil, A's share based on damage comes out at 1.6468 units, and B's nominal share comes out to 1.4531.
This is then rounded probabilstially with more chance given to the number of stackables that is closest in value to the nominal lootsize.
However, on this particular loot, A's nominal share is rounded down to 1, and B's is rounded up to 2.
There's still a few brain oil left to allocate after that first pass tho, so it goes back thru the loop to apportion the remainder.
This time, A's nominal share of the remainder is .5313 and B's is .4687.
Unfortunately for A, once again his nominal share is rounded down (this time to 0), and B's is rounded up to 1.

Outcome: A receives 1+0 = 1 brain oil in loot, B receives 2+1 = 3 brain oil in loot.
Click to expand...

Sorry Serica, I'm not sure what kinda rounding algorithm you are talking about..
When it comes to the person with higher damage, each time you are rounding down. even though the number is >0.5. Common sense tells it is usually rounded up when >0.5, and if <0.5 rounded down.

In both case you did the opposite...what mathematical algorithm is that.. no offence here... I'm just curious...

Also, I've not used only Apis + dante ... my Apis was tierd to level 2... and I did give few crit hit to the sand king...

So. I'm not that very convinced with yoru explanation here...

And not only this.. I've seen several times in team hunt... my hunting partner globaled/hofed immediately after i got crit hit and send to revival. or got disconnection...
Recently, after Sand King I've teamed up again and hunted the same mob and same areal... but decided to shoot at sepeare mobs... I've used the same tierd 2 apis+ dante and my partner used apis + beast..

My partner got several globals and 1 hof (profited) ... and I got 2 globals in the 50s.(lost few 100s peds).

And this kinda exteremely bad loot to for months...
So, pending better explanation, I'd think this is selective avatar loot manipulation.. or something I don't know.
 
Last edited:
Infinity01 said:
...

In both case you did the opposite...what mathematical algorithm is that.. no offence here... I'm just curious...

....
Click to expand...

I guess you haven't read the post by Noodles that I've given links to more than once before. I suggest you take a look both at that, and the much longer Mining Loot Analysis thread that it is based on.

It's exactly the same algorithm (as posted rather succintly by Noodles) that has been proposed for calculating the number of items of a mineral actually looted when mining in a taxable LA.
 
Just a note to say that there may be days that Fish and I can't get together to do a run.

We both have RL commitments at times, especially at this time of year :)

Also, we're not always on the same planet... so if we happen to both end up on a different planet at the same time, we'll try to pick a mission mob of about the same HP as Cornundacauda, and report on that instead.
 
Thanks heaps for this serica. It's always been evident for me that stackables are distributed according to some sort of probability, whether thats applied in the rounding or per stackable is what needs to be found it. I'll try get in game sometime to help out with this. :)
 
Does no one think it could be :

1 mob in shared loot

Person A 10% damage
Person B 15% damage
Person C 30% damage
Person D 40% damage
Person E 5% damage

And Ma

MA "person A you loose 100ped last hunt hmm you have 2brain oil
Person B dude your avatar is too ugly you don t have anything
Person C hmm i m generous 2 oil too
Person D ahahaha try again 0 oil next time don t shoot so much !
Person E ohhh a cute noob with opalo baaahhhh you take the full TT esi Gj dude have fun
Person F it s seem you didn t shoot but MA like you take an oil too "



No really you never think about it ? ;)



well joke over , gl Serica i m interesting to see how shared loot can work.
 
Serica said:
There is no way to accurately measure the exact HP remaining in the mob prior to the killshot. Whatever we try would only be an estimate. So, we've tried to keep it as simple as possible by both of us using unamped Opalos (average damage of 6).

We could drop it down further to use say a KiwioII (average damage of 1.5), and might do so if the Opalo shows a wide variance in damage done, but it would add to the tedium of the data collection and recording.
Click to expand...

Yeah, I mentioned it more to be borne in mind whilst analysing the results. It's tricky to measure for sure. You could just use the Kiwios when it's nearly dead, but I wouldn't bother initially either.

Noodles said:
It is not hard to show that this option 2, while a reasonable hypothesis, is incorrect. Several years ago I took a very small (unpublished) data set, in a less rigorous manner than Serica is doing. If you simulate option 2 (Monte Carlo-ish), it clearly shows that the variance that would be present is greater than the actual variance observed. So I ruled out option 2, and I'm sure it could be ruled out using these data also. Or I'll eat Jimmy's hat :dunce:
Click to expand...

Haha, hat ready but I doubt it'll come to that!
 
Noodles said:
...
Serica, I do have a thought about your hypothesis...when rounding probabilistically, what happens if everyone rounds up, and you have to distribute more stackables than are given to the loot? Is the loot changed? That's what seems to happen in mining. Also in mining, it is quite possible that the final loot value can be decreased due to probabilistic rounding. Could that happen here? How would it change your proposed algorithm? ...
Click to expand...

I've been thinking this over before responding.

I've been noticing that the loot seems to show in decreasing order of TT value in the chat logs so far, except for ammo, which is shown first. (I'd not really noticed that pattern in the past.)

One thing that occurred to me was that the total 'ammo' component of the loot might be adjusted to allow for those sorts of variances, as it has the lowest TT/unit.

Another possibility is a 'cascade' effect down thru the loots in order of decreasing TT value, but somehow I find that less likely as it would need the total loot components for each item to be rolled one at a time.
ie the total loot value is calculated, the BLACKBOX is then rolled for the highest tt loot component, that's distributed, then the BLACKBOX is rerolled for the remainder to pick the next loot component etc etc.

Or, as you suggest, the total loot might be adjusted. Perhaps if we get a nice hof of .. oh 1999 ped, and our combined loots come to 2000.03 ped, we might know the answer. :laugh:

As we gather more data, perhaps something will become evident - at this stage tho, it would be only speculation.
 
A thought about the overkill problem. In loots with ammo, could you use ammo as a proxy for actual damage done? There are so many ammo units in loot now, the rounding error should be minimal.

Or, as a first step, you could compare ammo ratios to chat log damage ratios and who fired the kill shot, and see if they match or are a bit different.

Anyway...keep up the good work!
 
jojo07 said:
And Ma

MA "person A you loose 100ped last hunt hmm you have 2brain oil
Click to expand...

My conspiracy theory long time ago, was when it took a long time until you got a loot message (could be "didn't carry any loot", or a loot window popping up), was that that kill did trigger something, people at MA HQ rushed into a board meeting, were briefed about the situation (player, mob killed and so on), and then they made a decision. Once the desision was made they entered the result into a terminal connected to the computer deciding the loots; it could be "No loot", "Normal loot" (say 50 pec), "Global" or "Special" (like ATH or Mod Fap), and when they pressed ENTER the loot window (or no loot) popped up on the screen.
 
Noodles said:
It is not hard to show that this option 2, while a reasonable hypothesis, is incorrect. Several years ago I took a very small (unpublished) data set, in a less rigorous manner than Serica is doing. If you simulate option 2 (Monte Carlo-ish), it clearly shows that the variance that would be present is greater than the actual variance observed. So I ruled out option 2, and I'm sure it could be ruled out using these data also. Or I'll eat Jimmy's hat :dunce:
Click to expand...

hey,

i'm very curious as to how you calculated the expected variance, or if you just chose one? One could easily code different ones, it's also plausible to put in a limit of how far off one data point can be. (dang it's hard do write about statistics and stuff in english, and i don't want to look up every dang word - i hope you know what i mean)


A few other things came to mind after i wrote my last post.

1.) Could it be that dmg dealt has more decimalplaces? e.i. you deal 54.3456 dmg? that could account for the small differences in big stacks of items. But still doesn't explain the Brain oil stuff.

2.) How many calculations are plausible for one team loot? I mean if let's say there are 3 team loots that need to be given out and calculated by one server at the same time. How many calculations can a server do in let's say one tenth of a second? And how much time would it take to run 10k items (like ammo) through my idea? I'm not into programming that kind of stuff and right now i dunno how i could find that out.
Plus when the sandking(again just an exaple cuz there are many people there and "big" stacks) is killed we always have a little delay before the window pops up, not like the lootlag that sometimes happens, but more like it takes time to determine the loot or it's distribution.


But you guys had some good arguments, maybe i made my life to easy with my idea. I'll go kill two levis and see what i can come up with. (besides my overdue ATH ;))

Regards
Thark
 
Thark said:
hey,

i'm very curious as to how you calculated the expected variance, or if you just chose one? One could easily code different ones, it's also plausible to put in a limit of how far off one data point can be. (dang it's hard do write about statistics and stuff in english, and i don't want to look up every dang word - i hope you know what i mean)
Click to expand...

A statistician could tell you how to calculate the variance. As I'm not...what you can do write your program/spreadsheet to do the weighted coin flips on each item in a stack. You'll quickly see that when you recalculate a bunch of times (i.e. Monte Carlo), the variation in the amount that each hunter gets for doing the same % of damage is quite large for larger stacks, while in actual loot, the amount each hunter gets is always quite close to the % of damage done.

Another way of saying this is that, for small stacks of stackables (e.g. brain oil from a sand king), this coin flip method might seem to work, but for larger stacks (e.g. 100 muscle oil split between two hunters), it's not consistent with what is observed.
 
You must log in or register to reply here.
Back
Top