due to some discussion generated by MA's developer notes nr.3 i have offered to track and make public my raw mining data in a format that allows analysis.
...
In the mentioned thread there was a discussion about loot and if it was triggered from a system that has memory or not and if it is related to the existence of a “personal loot pool”.
I have posted the following graph over there.
It shows the inverse of the cumulative distribution function (survivor or survival function) for the number of drops you do need till you hit a global (defined as loot >= 50 PED for ore and >=25 PED for enmatter). The sample consisted of 7360 consecutive amp 4 enmatter probe drops containing 85 globals and was delivered by Steffel back in 2008/2009.
What does the graph depict?
Let’s take the given value at 10 drops. It is roughly 90% which tells us that it is very probable that a globals comes after 10 drops and not before. The given probability of 90% also implies that there is a 10% chance to get one within 10 drops.
The median (prob of 50%) is at about 60 drops. This tells us that half of the globals will come after and half before 60 drops. At 237 drops the survival function is below 5% indicating that only 5% of the globals do come after this number of drops.
The shape of the distribution is a well known one and it is called exponential distribution. It can be used to model waiting times till events that are independent from each other do occur, quite like as we do have with our drops and globals.
The mean of the fitted exponential distribution is 86, hence in mean you have to wait 86 drops for a global. The median of an exp.dist. is the mean*ln(2) = 60. This was the value already identified from the survival function.
The inverse of the mean = 1/86 = 1.16% is the relative frequency of the event. The sample had 85 globals out of 7360 drops which corresponds to 1.15%. Hence we can conclude that the fit is adequate to assume that the waiting time (drops) for a global do follow an exponential distribution.
Now to the memory part.
As mentioned before, the exp.distribution models independent events. For some freq.distributions this is property is also called memorylessness and independence implies that there is no memory in the system.
The observed globals are from a series of consecutive drops. As their waiting time distribution follows a known random process that assumes “memorylessness” we have to conclude that the triggered globals do come from a system that has no memory. This also implies that there was no "personal loot pool" for the avatar Steffel back in 2008/2009.
Steffels data is several years old and it is always good to have a recent dataset. Therefore FallenAngel’s efforts are very welcome.
P.S: Please note that this is enmatter amped data and hence given times are only valid for this setup. For unamped mining the global of 25 PED would correspond to a 6 (class 3) or 15-20 Pedder (class 4). For ore you can double the values.
