I had a really excellent first week. I got the 100 gems top task twice. but after seeing everyone’s boards for this week. im feeling a lot less stoked about the probability of that task and the rewards in general.
Id say i def still like the adventure board but the lack of event keys is discouraging for sure.
So—at least for a first pass—you’re assuming that tasks are drawn in two stages. First the rarity is selected, and then within that rarity a task is drawn. This is how troops are chosen from chests, I believe, so it isn’t implausible.
Otherwise, if all tasks were thrown into a pool with appropriate weightings, getting one of the three legendary tasks one day could (if no repetition is allowed) reduce the chance of a legendary by 33% the next day.
I logged in and completed all tasks on Monday. I just forgot to take a screenshot. I only remember it was another kill 100 troops for 20 gems. Sadly I don’t remember the bottom tasks.
I don’t have these, I’m sorry. I was a bit of a derp and just forgot. I can try tracking another week and grabbing all 7 before I upload if you want/need more data.
July 17th (day 7):
Happy to keep posting these for as long as you’re collecting data @Mithran, thanks a lot for the effort you’re putting into figuring this out 
@Mithran the picture from Account 4 made it onto the post shortly before the reset (sorry for that—my brother just played a different game all day)
And you’re right to ignore the Purple troop task the second day—I just forgot to attach a “Disregard” label
More pictures to come when l find the time! Good work so far, all 
Promised pictures:
July 17
Account 1
Account 2
Account 3, Disregard Top
Account 4
Since we are parsing live data to estimate appearance rates and not datamining the “actual appearance rate” where we need to know the special rules to apply that to a calculation (like we would, for example, to get the average value of a legendary task reward starting with mined data), in a set of appropriate size, this would work itself out. For example, a “33% lower the following day” would be represented by the rate of overall legendary boards being recorded at lower than their “actual appearance rate”, but the amount of legendary boards we get over a given period of time would still be accurately represented by our “incorrect” appearance rate estimate. The same can be said for any tasks that are “locked to a slot” or whatever hidden mechanic is preventing 3 of the same rarity from appearing on a given day. I’m not really looking at “what is my rate to get this specific task today”. I’m looking to glean the average weekly value of the entire rewards table. So those things might make the path more bumpy, but will take you to the same place.
The part where we’d run into potential trouble is if these “bumps” were based on rules that would explicitly thwart the type of data collection you are doing - eg., (hypothetically speaking) “You can only roll a mythic task on a Saturday”. If we didn’t know about this “Saturday” rule, even if we recorded weekly sets, we’d assume we have 7 times as much relevant data as we have to make an estimate on that parameter, which is going to lead to a greater degree of inaccuracy. Also, we might have sets with more Saturdays than others if you were just polling “randomly”, which would skew our data. But if we have enough full weeks, including equal amounts of Saturdays, the overall mythic task rate would still be close to accurate even if you weren’t aware of the rule - “1% on a Saturday” would just be recorded as approaching “0.143%” (eg., with an equal amount of Saturdays in a set to any other given day of the week, 1 in 700 total task samples would be expected to mythic task, which is still accurate). Task repeat restrictions would further lower the rate recorded, which would then be reflected when you multiply it back out to calculate the weekly average.
In our case, any hidden “next day” or “locked slot” rule would only skew short term by relatively small amounts, and their influence would be at least somewhat reflected in the data recorded given the data sets requested (a week or more, and consecutive). Part of the reason I want parse this in sets only of a given size is to smooth out any trouble hidden factors like this would cause.
I know part of the concerns people have with this new system is just how “bumpy” the path can get, so it is still helpful to know if these factors are present, though. This looks like one of those things that would be impossible to “prove” from this end, except with evidence of the contrary.
That being said, it looks like “a set of appropriate size” is going to be pretty big, if you want to approach accuracy of within, say, a few gems worth of accuracy compared to average weekly “actual rewards”. The amount of data we have so far shows a clear trend, but we still don’t have great on Legendary boards (over 100% relative divergence between the “daily” and “all” sets), K100vsK150(another 17% relative divergence) or especially mythics(one set doesn’t even have a single sample, and the other only has one total) and the gem/wk difference between the two sets is currently about 11, but that is under the current assumption of equal weighting of task types within a rarity tier. Planning a simple count for this sometime in the next few days to see if we are ballpark, would be at least another week before I have time input data for then set up a table to calculate for every type/rarity combo to use recorded type ratios as well as rarity ratios for table value. At the moment, though, I know it would skew things significantly if I used “hard data” with the sample size we have within every rarity, particularly for legendary.
Here is the preliminary recorded data - needs a verification sweep to ensure everything was recorded correctly.
Daily column indicates whether or not the data is daily posted. Y is yes, N is no, X is omitted.
Queenofevrythng, BruisedBull, and Kris11 are included in the “daily” sets despite having < 7 days worth of total data so far because they have been posting daily.
Fleg and Aurora_x are currently in the “all” set because even though I have 7+ days worth of data, some of it is from before the trail started and batch posted. Voq/Arelana were dropped from daily to “all” due to missing days. eXDe was place in the “all” set due to only having two screens and missing the last several days. (Thank you all still for your contributions, having this data allows for an alternate data set to show how much it affects the outcome)
The one omitted @Julli-of-north, as I have only the one screenshot and no way to record the other data reported on my sheet.
The columns for S1, S2, S3, and ST are slot 1, slot 2, slot 3, and slot top, respectively. Data for the bottom slots was recorded using the standard one-letter rarity indicators and color coded. Data for the top slot was recorded as LT (‘legendary top’) or CT (‘common top’) so they would have a different value during table-wide counts. An X means the data is omitted.
This next set was recorded a bit differently, since I got these n bulk, but parsing remains the same for the results. I might reorganize these later to save space:
Based on this, we currently have:
FYI, my initial estimate was:
So overall, this table should be set to pay out slightly better than my initial estimate (with a few things, like chaos shards, are a bit lower, but still very present).
Reward table estimates once I have a chance to give task type ratios a once-over to see if they are at least trending toward “even” and do a verification sweep of the current data.
If you still need posts…
18.07.2019 
Still taking any data submissions, yes. Might be a while before I add more to my table, though, and might start a new set (those that have yet to finish out their first 7 days will go on the first table as soon as I get them).
And so it looks like I get the honor of the second mythic task to grace this thread, and the first one to actually pay out keys.
July 18 - Day 8
Acct 1 (disregard top)-
Acct 2 (disregard top)-
Acct 3 (disregard top)-
Acct 4 (disregard top)-
I’ll start the new cycle with all 3 accounts, when I get back from work.
I did Kill 100 troops on both alts yesterday and I’ll start doing that every day too, so we have better representation.
I didn’t want to post them last week, since I some times either don’t finish top task or not even log in all day on one or both of them.
I put seven consecutive days in my original post here.
Looks like I’ll be of for about two days and the I could start new seven day cycle on Sunday or maybe better Monday.
