7/18/19 day 8
Not that it can be included here but I’m jealous of a guildmate who has hit TWO mythic key tasks in the last week 
1 Like
Added @Dust_Angel to daily set and @Julli-of-North to all set for week one. @Arg dropped from the set due to only having one post with two entries. Cleaned everything else up, columns now aligned to one-week sets and day tracks which day after the post went up that I started on, or, in the case of full submitted week long sets, which day of the set it refers to (day isn’t used to calculate anything, its just for myself to know what entry I’m on so I can track missed days)
Separated daily submitted entries from weekly submitted ones (@aurora_x switched initially posted a full week, then switched to daily, and the daily set is part of the incomplete second week set which I haven’t even entered yet, which is why it is x’d for now)
Which puts us at:
For those that missed days and want to keep submitting, any time is fine. Prefer a label with current data and which day of your set it is, just makes it easier for me. You don’t need to wait until a specific day, and the more consecutive days, the better for data accuracy.
Also, it is easier for me if you make a new post each day rather than editing your post so I only have to scroll down to make new entries.
1 Like
Sorry—I fear I’ve been the worst offender on this front! Hopefully adding 4 accounts a day has mitigated the annoyance a bit, but I’ll definitely try to do better.
July 18, Day 8
Account 1
Account 2
Account 3, Disregard Top
Account 4
How much data do you think is needed to get an even somewhat accurate estimate of the mythic task rate?
Define “somewhat accurate”. Within 25% is a lot easier than 5%—25 times easier, if I am remembering my stats correctly. (Halving your uncertainty requires four times the data.)
Edit: for rare events, the Poisson distribution is a good approximation to whatever is actually happening. And according to the website I just looked at, the 95% confidence interval varies inversely (but to the first power) with the number of trials. So five times the data results in five times tighter uncertainties. Wonder what I was thinking of.
https://www.statisticshowto.datasciencecentral.com/binomial-confidence-interval/#Poissontable
Since a single appearance at this point would currently skew the recorded relative appearance rate by nearly 100%, I’d say… a lot.
For practical purposes on what I’m trying to calculate here, though, I think if we can get the estimate within half of a percent of its actual value, it will be “good enough”, and within half percent of its actual value at this point looks to be… a 100% margin of error. As in, if the appearance rate of mythic boards ends up falling between >0% and 1% and we estimate it at 0.5%, the overall influence it has on the table at that point is pretty minor (mostly gem keys, VIP keys, and shards, since mythic tier has no gems task). A +/-0.5% difference on mythic task appearance would be a difference of (per week) about half a gem key, a tenth of a VIP key, and 19 chaos shards, barely offsetting everything else on the table (within about 1 actual gem difference for example). Even we go off “gem value” for high value keys at 100% of their purchase price, this would diverge the table “value” by only about 2.5% (less if you assign “value” to all the gold and souls etc), which I think is an acceptable margin of error for the table.
By the way… In a set of the size we’d have by the end of this week assuming equal participants, we’d need… like 10 more mythic tasks to be seen to go over 1%. I’d barely have over 1% at this point if I just did single entry mythic tasks from every post I’ve seen so far on the forums even offhandedly saying they or their guildmate got one.
1 Like
- July 2019
