Make your own free website on Tripod.com

DKP vs Random

Results of a raid loot distribution using both zero sum DKP and /random models.
 
A guild of 150 members was run through 500 raids.  Each member had a different "hardcore" factor, meaning some members were more likely to attend raids then others.  Each player had 20 slots to upgrade, but a player would only upgrade a slot if the loot was 10% better then what they had.   Anywhere from 2 to 10 loots dropped per raid of various uberness, and each loot drop could be equipped by any player but each loot was specific to a single slot.

Zero Sum DKP
The following four charts display how the loots landed using the zero sum DKP distribution model.  Each loot that dropped cost 10 DKP.  Whichever player at the raid for whom this was considered an upgrade (i.e. 10% better then what they had in that slot) who had the highest current DKP balance would win the item.  No limit was placed on the number of items that could be won at a single raid.

This graph shows the average number of raids it took each member to win a loot drop in the DKP system.

dkp_image001.gif
DKP

This graph shows number of wins per raids attended in the DKP system.

dkp_image002.gif

This graph shows the uberness of the player (that is, how much his character has improved) vs. how many raids they attended.  DKP system.

dkp_image003.gif

The following graph shows how many raids players had to attend per win vs the number of raids attended.  For the DKP system, this graph shows that it does not matter how hardcore or causal you raid, it takes about the same effort to win items.

dkp_image004.gif

Pure Random
The following four charts show how loot was handed out in the pure random distribution model.  When loot dropped, whoever rolled highest among those for whom this was an upgrade in that slot would win the item.  No limit was placed on the number of items a player could win at a single raid, and there was no "two week" rule in the simulation.

This chart show the average number of raids it took for each player to win an item in the random system.

rnd_image001.gif

The following chart shows for the random system the total number of loot drops vs. number of raids attended that each player obtained.

rnd_image002.gif

For the random system the following chart shows how uber each character ended up given the number of raids they attended.

rnd_image003.gif

The following chart shows that, in the random model, the number of raids attended between wins did not depend much how often you raided.

rnd_image004.gif

Two Week Rule (Random)
The following charts show the effect that limiting how frequently a player can roll.  The constraint was added that if a player has won in the past 14 raids then then they may not roll for new loot. 
 
Not much changed over pure random other then the raids/win number seemed to go down and on average even out, which is what this rule was designed to accomplish.

twoweek_chart1.gif

twoweek_chart2.gif

twoweek_chart3.gif

twoweek_chart4.gif

Conclusions
Everyone needs to come to their own conclusion.  What is obvious to me from these simulation results is that loot distribution under a zero-sum DKP system results in a much more even distribution of loot.  Everyone gets loot in the DKP simulation, and everyone had to make about the same effort (measured by attendance) per loot. 
 
In the random simulations several casual (i.e. low attendance) raiders got nothing, and the effort required per loot had a much wider variation.  With random some got loot every 10 raids or less, others every 40 raids or more.  In both simulations the hardcore raiders wound up with more total loot, but it is possible in the random systems for an individual who attended 400 raids to wind up with less loot then someone who attended 100 raids. 
 
With the zero-sum DKP system I am reminded of LDoN - if you earn the points you eventually can buy the loot.  It is obvious that the hardcore raiders did not have an advantage over the casual raiders in the DKP system, as the raids/win and uberness graphs show a linear relationship between attendance and reward.