The spreadsheet is downloadable (to open in google docs, microsoft office, or libreoffice). The link should have a "file" menu from google near the top left, which you can save it to zoom in better.

The only patterns that I've noticed are:
►Higher number depths give more crystals per box, and not more boxes (Hypothesis: Better red:green box ratio).
►Certain depth types have better spread of the quantity of boxes. Missions will have higher payouts in comparison to the depth types it enforces. (I'll have to sort them when even more data is collected, since some only have 2-5 entries)
. Current rankings (least to most frequent boxes) -> Graveyard, Compound, City, Grass (Island), Monsters (3Slime), Candlestick Keep (Pawn), Clockwork (Gear), Treasure Vault, Arena
►Crystal and orb types are depth-dependent, not stratum. (D22 drops both shining and radiant fire crystals, while the rest of stratum 5 will not drop radiant)

I have done some missions more than others for a couple specific reasons...
►Dropped the orbs/crystals I needed to collect.
►Could run through hitting all the boxes without attacking.

There was one interesting thing that I discovered at least with farming orbs: It is more time-efficient to farm crns (via clockworks T3, and buy advanced or lower orbs from the shop than it was to run through (with MSI VH) missions looking for orbs.

I'm really looking forward to my new data I'm gathering to hammer out the Red/Green drop ratios. I'm also recording a lot more data now with a primary drop (fire/crns/token/item) and a secondary drop (fire/sol/orb). The other thing that I was looking into collecting data for was the monster quantities per depth... but that became too tedious to count (Only got through one depth before giving up).

I'm not quite sure what is causing the current swing in my data, but the only major thing (besides time, and possibly luck with the drop generator) is I have been NOT going solo. This has increased my orb farming from about 1 in 5 depths to 1 in 2 depths. I've only recorded one T2 run so far with this possible discovery, but have run a couple other times noticing a similar trend.

Think I should record it when the depth starts (when the boxes are generated) or when they are opened (when the items are generated)?

They are definitely working on changing these drops. I'm going to put any data I record on hold for now.

As of the last update (10/16) there have been a couple changes (all of which I am not completely aware of yet). But for people interested in a couple datapoints...

►Crystal/Orb gather depths have changed. Most noticeably, Radiant crystals have only been spotted on D28, and no longer D22-27. There is a similar shift for other FC's, but has not yet been fully outlined.
►Orbs seem more likely to be found out of place. Previously, Elites were dominant in stratum 4. Now they can be found between S3-S6. (Sample of 5 elite orbs on various depths). Previously, they were very rare outside S4 (Only 1/40 seen as noted with my previous post, as opposed to 2/5 found in S3).

I'm interested to see if the shift in Radiants also affected Shadow Lairs, but currently have none planned in the near future.

The fire crystals in my own data represent a single drop. On elite, that means 1 = 3 crystals.

Types of orbs are difficult to follow, as noted with my post #7 in this thread. You normally find 2* orbs in stratum 2 (where SotB is located). It is possible to find them outside the stratum, with a very rare chance. In my case, it was finding a 4* orb in S3.

It looks like the rates from boxes remain the same (~200 boxes show similar averages).
The depth that radiants begin (and possibly other crystals) has changed.

From this post, I've done 2 T3 runs that were recorded, both started getting radiants at D27.
From friends, I've heard radiants begin at D26.

Aah. It used to be about 50-50 on D19. It looks like the shift for radiants was global for all FC then.
Orbs are still confusing and rare enough to predict nicely.

Also D25 is now confirmed to have radiants. This was likely a change at 3/52014 patch.
Source.

Ok, I know I talked to a couple people about showing the stratum payouts for the arcade and how odd/even stratums make a difference.
I just added this sheet to the spreadsheet.

I highlighed in green the columns of interest.
Average payout sorted by depth.

You'll notice, despite the low quantity of datapoints, that Depths 19-22 give less crowns than depths 5-7.
Even Stratums give on average better payout than odd stratums.

*D7 average from arcade is skewed due to 2/3 of the datapoints being Snarbolax.
Edit: Moved Gloaming Wildwoods to mission tab, despite running it in the arcade. Next planned calculation will be depth type sorting [box counts and crn payouts per depth type separated by stratum].

It has to execute when the box is opened, since it checks who all is in the depth and alive. Dead players will not get loot and players can join and still get loot mid-depth. They may use tables of pre-calculated random values for the RNG to reduce execution time, but it is definitely checked when boxes are opened.

I think is is safe to come to the following conclusions:

1. Danger rooms are always profitable until energy increases above 10kcrn/100E with the only exception being the first depth of an odd stratum (D1, D9, D19) where the magic number is closer to 6.5kcrn/100E.

2. Even and Odd stratum wealth distribution. I've been saying this for awhile, now to elaborate a bit more on it. The first depth of an odd stratum pays horribly. Usually close to 300crn average (arenas are close to 1kcrn). The remainder of the stratum pays about an average of 600crns. The first depth of an even stratum pays slightly more than odd stratums (~700-800crn). The remainder of the even stratum pays approximately 900crns. [At some point I'll break down to make a full-SQL call comparison for depth types sorted by depth number for a more thorough comparison. Edit: Thought of a clever solution that would take significantly less work to add into the existing calculations. - This may be done within a week.]

3. Arenas are laughably overpowered with their payout if you can beat them. 31 boxes and 3-3.5k crn payout on an even stratum. This means that the one arena out-farms an entire T1/T2 boss lair. Theyre about 3x the payout of most other depths.

4. Candlestick keep is a very strange depth with the highest variance in treasure boxes that I have seen. I have one candlestick keep recorded that contained 30 boxes (Data #395 on Data#5_Arcade). The funny part is that the boxes on this depth were not at their maximum possible quantity (one of the "rooms" had 4 with a max of 6 possible that spawn in the areas). This is the only depth with the possibility to have more treasure boxes than an arena. That being said, the average is around 13, so not a super place to farm.

5. Deconstruction Zones and Graveyards are the only depths which can have 0 treasure boxes. Despite this possibility, compounds seem to have a lower mean at the current worst ranking box count.

6. Clockwork tunnels have approximately a 50% chance of containing a danger room. Without the danger room, they are worse than most other depths. The danger room is significant enough to raise the payout and box count enough to on average be better than all depths except arenas and aurora isles.

7. Aurora isles aren't randomized depths, allowing for better-than-average farming for the (tied for) 2nd best depth type. The Jelly Farm variation is the best of these depths due to a single 3 E gate opening 8 additional boxes and a battle arena (which you can then backtrack and go the normal way again).

I've started taking advantage of the D1 arena for adding tons more boxes to this research... and I have have found out yet another aspect that was previously overlooked.

Depth sure seems to affect rarity rates.
From the first 200 green boxes, there is a current rate of 5% for fire crystals (12 looted). Yes, I could be hitting a strange part in the random number generator... but the "normal rate" is approximately 4x that. This sheet, if you compare the count and FC columns, gives a fairly good visual representation of this idea. It may in fact just be stratum 1 being different than all others.

Also - I may have evidence of double normal loot from boxes (As I mentioned in the SL notes how some seemed to drop both crowns and heat).
Take a look for yourself.
That location contains 4 boxes, of which two contained crowns, one contained heat, and the last contained both a remedy cap and a pill (shown in the image as stacked on top of each other).

[You can blame Glacies for needing the D3 artillery strike damage and this happening on D3 instead of in the arenas I was gathering data on.]

I'd need more data, but D1 (4.5%) is significantly different than D28 (19.9%) for green boxes.
I just added 2 new columns to calculate the chance for fire crystals on those depths instead of crowding this post with even more numbers.

The rest of stratum 1 seems to be close to D1, while S4-6 seem to match D28. The "normal" rate I was referring to is the global average, which I previously believed to be the same across all depths (which currently doesn't seem to be the case).

I'm now concluding this project of mine.
As I asked a couple posts ago... I did end up setting up the form for arcade payouts.
The thread on this topic can be found here.

This links to the form and data will appear in the first post of this thread as well, as that the general discussion thread will likely get lost quickly.

I enjoyed this project and learned a lot about SQL style data organization. The automation in data sorting was one of the key things that kept me coming back to do more. But now I think I have done the analysis and all that remains is gathering a larger dataset.

You are welcome to continue the research, but I am no longer actively supporting it.

The sheet is downloadable and updates based on the "Arcade", "Missions", and "Prestige" tabs [where the raw data goes].
I also avoided taking 2 datapoints on the same depth [friends have offered to help running with me] because it would make a bias in the box counting on depths.

If you just wanna go on runs with me, send a friend request and try and catch me online.