Page 1 of 1
The bias of Not Created Equally
Posted: Sat Aug 12, 2017 10:53 am
by BTAxis
I've done a bit of a census to see if the random stats created by Not Created Equally show any biases. I started a new game with 95 starting soldiers, so along with the 5 recruits available at the start that makes for a sample pool of 100 soldiers. Results:
Code: Select all
Aim Defense Health Dodge Hack Mobility Will
Default 65 0 4 5 5 15 25
NCE Sum 6436 -58 409 611 497 1498 2625
NCE Bias -0.64 -0.58 0.09 1.11 -0.03 -0.02 1.25
It seems soldiers get short-changed both on Aim and Defense, but they get quite a lot of Dodge and Will back for it.
Re: The bias of Not Created Equally
Posted: Sun Aug 13, 2017 11:03 am
by Zork
Lol that's the default values? At Easy they look more like the min values.
For me the random isn't applied independently to the stats, that is a soldier can't have them all very good. But I suppose it won't change a stat analysis looking at them independently.
Re: The bias of Not Created Equally
Posted: Mon Aug 14, 2017 7:46 pm
by Exquisitor
BTAxis, I appreciate your analysis and effort. However, you do realize that you can still have considerable bias in a random sample size of 100. So randomization tests are often done with sample sizes of 1000 or more.
Re: The bias of Not Created Equally
Posted: Thu Aug 17, 2017 4:43 pm
by caseywills
I've noted this trade off as well. Didn't do anything quite so scientific.
I noted that in 5 or so restarts I was sacrificing aim (most important stat).
Re: The bias of Not Created Equally
Posted: Thu Aug 17, 2017 5:39 pm
by BTAxis
Exquisitor wrote:BTAxis, I appreciate your analysis and effort. However, you do realize that you can still have considerable bias in a random sample size of 100. So randomization tests are often done with sample sizes of 1000 or more.
Yes, but it was tedious enough to do 100, hahah. Plus I'm not ruling out an error in adding up all the numbers on my end, so these figures are by no means super reliable. Still, there you have it.
Re: The bias of Not Created Equally
Posted: Fri Aug 18, 2017 10:10 am
by aureon
Those all look within the margin of error.
Re: The bias of Not Created Equally
Posted: Fri Aug 18, 2017 1:34 pm
by Sparky79
Are psi and hack included in the pool? Because that would kinda suck.
Hack is useful to only specialists and somewhat to shinobi, and psi is only to ... psi.
Maybe those two should be in a separate pool?
Re: The bias of Not Created Equally
Posted: Fri Aug 18, 2017 2:30 pm
by JulianSkies
Sparky79 wrote:Are psi and hack included in the pool? Because that would kinda suck.
Hack is useful to only specialists and somewhat to shinobi, and psi is only to ... psi.
Maybe those two should be in a separate pool?
Hack and Psi are, indeed, included in the pool.
Re: The bias of Not Created Equally
Posted: Fri Aug 18, 2017 5:40 pm
by Sparky79
JulianSkies wrote:
Hack and Psi are, indeed, included in the pool.
Well that kinda sucks.
Seems to me that most of the time they just take away from the rookie.