open All Channels
seplocked Test Server Feedback
blankseplocked Statistical anomalies in salvaging
 
This thread is older than 90 days and has been locked due to inactivity.


 
Author Topic

Lorelei Lee
Posted - 2010.10.19 13:56:00 - [1]
 

The number of times one has to cycle a salvager before succeeding is not distributed quite uniformly. Has anybody else noticed this? I have been experimenting with it a lot, and I think what I am observing is a bug. Here is a histogram of salvaging 58 identical Serpentis Hired Muscle (BS) wrecks:

Cycles: Number of wrecks that succeeded after that many cycles
1: 5
2: 10
3: 3
4: 5
5: 5
6: 6
7: 3
8: 3
9: 2
10: 3
11: 0
12: 0
13: 1
14: 1
15: 0
16: 3
17: 4
18: 2
19: 2
20: 0

No wreck required more than 19 cycles. If all cycles have equal chance of success, then I would expect the histogram to look like exponential decay. This does not look like exponential decay. Note that 1 cycle is not the most common result. Note the spike in the 16-19 attempt range.
In the above run, all salvaging was done with a single Salvager module on auto-repeat, no interruptions, no other modules cycling at the same time. Here is another run of 47 Serpentis Hired Muscle wrecks, all salvaging still done with a single salvager on auto-repeat, but this time while fighting, tractoring wrecks, and generally being active:

Cycles: Number of wrecks that succeeded after that many cycles
1: 11
2: 8
3: 1
4: 5
5: 2
6: 1
7: 5
8: 2
9: 1
10: 3
11: 3
12: 2
13: 0
14: 0
15: 0
16: 0
17: 1
18: 0
19: 1
20: 0
21: 0
22: 0
23: 0
24: 1
25: 0

This looks more like exponential decay.

Hypothesis: the use of random numbers in salvaging is subject to some kind of cross-talk with other effects. The calculation itself is buggy; the cross-talk, when it occurs, jolts it in unpredictable ways, yielding results more random than the success calculation on its own.

CCP has not published how their success calculations work, so it is really hard to prove that it's a bug without knowing what the expected behavior is. That is why I am asking the rest of you for hard data. There is only so much salvaging I can do before I go insane.
I am also seeing anomalies in what I find: looks like Fried Interface Circuit is overrepresented in that 16-19 attempt spike. Overall, I found more than twice as many Fried Interface Circuits when salvaging while fighting vs. when doing nothing else. If confirmed, that means you can affect what you find in a wreck by changing when you look -- cool, huh?

Denidil
Gallente
The Graduates
Morsus Mihi
Posted - 2010.10.19 19:13:00 - [2]
 

you need a sample size about 20x larger to be meaningful.

/computer scientist.

Lorelei Lee
Posted - 2010.10.20 00:36:00 - [3]
 

I agree that 20x more data would get me a useful degree of confidence. Help me make it so :)


 

This thread is older than 90 days and has been locked due to inactivity.


 


The new forums are live

Please adjust your bookmarks to https://forums.eveonline.com

These forums are archived and read-only