Droid Quality Experiment by Drashk
repost from SWG DE forums, with permission by Ackis
Why run the Experiment?
Recently a number of questions, regarding whether or not the Mechanism Quality of sub-components has an effect on the success/failure ratio of a droid, have occurred. A number of people felt that the sub-components help to increase the chance of better successes during all aspects of the experimentation process. I disagreed with this belief, from past experiences, and decided to run an experiment to back my theory.
The Test Environment and Materials
For the entire experiment, I used a W/G/D private crafting station, with a functionality rating of 34.24 and 3, factory made, W/G/D crafting tools with a tool effectiveness of 12.15. I made a run of 100 factory parts for each of the following, using the same materials in both sets of components. Set 1 was fully experimented on, using as many experimentation points as I could. None of the components in Set 2 were experimented on and consisted of parts that had a Great success on the initial combine.
Set 1 Set 2
Manipulator Arm 85% MQ 11 Manipulator Arm 24% MQ -3
Advanced Droid Frame 47% MQ 1 Advanced Droid Frame 10% MQ -7
Droid Sensor Pack 65% MQ 6 Droid Sensor Package 12% MQ -5
Droid Motive System 81% MQ 10 Droid Motive System 24% MQ -3
Droid Brain 77% MQ 9 Droid Brain 22% MQ -4
Overall MQ 37 -22
For the deed construction, I used the following non-ferrous metal
Copper
CR 291 Mall 734
Con 651 OQ 981
DR 213 SR 460
HR 378 UT 506
I ran the entire experiment on the Test Center during build 65956.67964, which is the same version of code currently on Live servers, in a city that offers a +15% Assembly bonus.
The Experiment
The experiment consisted two separate manual production runs of Deed for: LE Repair Droid. I chose the LE due to the fact that only 20 resources of one type are used in the crafting process, yet it uses 5 sub-components. All parts for the first production run came from Set 1 and all parts for the seconded run came from Set 2. Experimentation points were used only on the Experimentation Effectiveness line for all 200 LEs. Any left over experimentation points were discarded. The Initial Combine refers to the first crafting stage, after all components are added. The First Experimentation refers to the first use of all available points, which were 8. The Second Experimentation refers to the use of the final 2 experimentation points. Null count refers to a previous critical failure that made it impossible to gather further data. The percentage to the right of each result is the percentage of non-Null test results. The percentages were rounded up in the nearest tenths place.
Set 1 Set 2
--------------------------------------------------------------------------------
Initial combine
Greater Success 93 93% 93 93%
Critical Failure 7 7% 7 7%
First experimentation
Amazing Success 8 8.6% 3 3.2%
Great Success 64 68.8% 68 73.1%
Good Success 8 8.6% 9 9.7%
Moderate Success 7 7.5% 6 6.5%
Critical Failure 6 6.5% 7 7.5%
Null Count 7 -- 7 --
Second Experimentation
Amazing Success 6 6.9% 1 1.1%
Great Success 73 83.9% 81 94.2%
Good Success 0 -- 0 --
Moderate Success 1 1.1% 0 --
Critical Failure 7 8.0% 4 4.7%
Null Count 13 -- 14 --
Experimental Effectiveness
98% 65 67
86% 1 0
73% 1 0
71% 12 14
59% 1 0
57% 6 5
0% 14 14
Comparing the results
At first glance, it would appear that Set 2 actually produced better results, even though the MQ rating was far lower. What actually becomes more evident, if you know a bit about statistical mathematics, is that the numbers are virtually the same in both sets of results. With the exception of the number of Great Successes in the Second string of experimentation, all of the percentages are within +/- 5%.
I'll take the comparison one set further, by comparing the overall rate of Great successes and Critical Failures during the entire process.
Great Success Critical Failures
Total Experiments 280 279 280 279
Initial 93 93 7 7
First String 64 68 6 7
Second String 73 81 7 4
Total Successes 230 242 20 18
Overall % 82.1% 86.7% 7.1% 6.5%
Conclusion
Even though this test was conducted with only 200 droids, I feel confident in saying that the Mechanism Quality does not have an effect on the success/failure rate, while building a droid. Looking at the numbers in a closed environment, it might even be suggested that experimenting on the sub-components lowers the chance of a Great Success.
The Challenge
I challenge all of you to try and reproduce the results that I have listed by setting up your own experimentation runs. If I am correct, only the qualities of the resource used have an effect on the end result, currently. I've done this much, I challenge someone to prove my observations wrong or back up the information that I have gathered. I'm hoping that at least one person can take me up on this challenge, so that we can compare data
--------------------------------------------------------------------------------
February 04, 2004
Copyright 2004 by Ackis |