Board Thread:Wikia Discussion/@comment-25283969-20150511171714/@comment-24430017-20150515032120

Homuhomu123 wrote:

Thanks for the concerns. According to my past experiences on testing & data collection, here are my answers:


 * Confirmation bias >>  Could be resolved by collecting sufficient amount of data thru experiments


 * Randomness >>>  Could be largely (if not w/ 100% certainty) controlled by increasing sample size. I guess you've read the blog post from Winzongod already.

> Autehnticity >>>>  This is the most difficult thing to manage. If there's a convenient (& safe) way to do the auto-recording thing (If you have one pls enlighten us). I think most testers will surely use that. Currently, most player reports are simply spreadsheets / graphs uploaded. The others are posts on Shitaraba BBS w/ a few words of description. Most of the Kancolle knowledge on this wiki, however, are using them as basis for conclusion.


 * "Impossible to figure out actual rates with any certainty" >>>>> Right and wrong. It is extremely hard to find the actual rates, but we are able to put a good estimate (or range) for that value. On the otherhand, if you choose not to believe most of those player reports due to authenticity issue... I'll suggest do your own tests and maybe convince others to believe in it someday.

Regarding the drum loading expedition, we've seen an earlier chart stating "3 drum give 50% GS rate". As you may have noticed, recent tests (Kitaraha, zel & me) revealed that the GS rate for #21 w/ min drum is pretty far from 50%. Assuming the old test wasn't falsified, the drum expd should've been changed.

Most experiments aren't conducted like expeditions as you actively conduct them and log the data. The problem with expeditions is that they come in at random times rather than when you are free. I think a safe assumption is that most people try to have expeditions running around the clock. So if you are busy doing other things you can recall and resend and expedition occasionally forgetting to log your results. Now if you expect something like GS you are more likely to be suprised when you don't get GS and this perception can make you more likely to log the data.

Many people are using programs like KCV, KC3kai. These can potentially have the functionality to log data. Even if it is not automatically uploaded having it track 100% of expeditions will provide assurance that people didn't forget about any expediitions. I think if you really care about the accuracy of your results you will want to find someone to implement some type of logging or tracking system.

And yes like I mentioned before 21 used to be 100% with 4 drums and 3 sparkled ships for me in the past, but that is no longer true and i need 4 ships for 100%. People did a lot of tests back then and I think it is safe to say that the rates have changed. I think many people have noticed a change in it?

You are right in that if the rate is really 19% per ship then it should be quite noticable if you use 5 ships (~95%). In past I think we concluded it was around 100%/6 and I think that got rounded down to 16% along the way. If you want to see if it is 19% or ~16.7%, it should be pretty doable. If you want to figure out to <1% accuracy it is going to be more difficult. Now if the 19% includes barrel expeditions then that is probably going to throw things off.

Since rates for 21 have changed its also possible that rates for normal expeditions have changed and even on a per expedition basis. I personally have not noticed anything like that and I am skeptical that this would occur.