This link refers to a different published snapshot. The page shows the current sample; values may have changed.
Results from my recorded Chaos chest openings. These are observed results, not official drop rates.
Try different setups using the published chest sample. Your choices stay in this browser and do not change the records.
Try recorded chest outcomes or an illustrative Prismatic item mix. Simulated results do not add to the recorded sample.
Enable JavaScript to open simulated chests. Recorded replays whole published openings. Experimental uses assumed Prismatic item weights of 75% non-collectible, 20% Standard collectible and 5% CE collectible, with equal item chances within each group. These are not official drop rates.
Browse item details, weights, and Eggshell Coin values.
This catalog describes reference items. Recorded examples are available in item details and do not establish official drop rates. Pips may follow the assumed 3 kg rule.
Browse approved observations and the corresponding published screenshot copies. Pending and excluded scans are not included.
Choose an end date on or after the start date.
Select an opening to inspect its loot.
Compare Standard and Collector’s Edition items. This display is a reference catalog and shows no personal ownership.
Select a collectible to compare its two editions.
A separate manual tally from my Chaos nests. Counts come first here because this sample is much smaller than the chest study.
Nest glow tip: purple indicates Ruins Aniimo Egg, gold indicates Umbral Aniimo Egg, and pink indicates Prismana Egg. Shadow Sparkling is tracked separately from Prismana Eggs.
A personal dashboard for my Egg Heist findings.
I use an OCR tracker to record loot from screenshots, review flagged results, and share the findings here.
Scanner-accepted observations are distinct from individual visual review. Passing scanner checks does not guarantee an error-free record.
This is one player’s complete-opening sample, not official game odds or a community-wide random sample. Routes, source names, game updates and related outcomes in a run can affect what is represented.
Displayed loot value does not mean every item was collected or successfully extracted. Inventory limits, deaths and other costs are outside the calculations.
Whole-chest totals are combined assuming representative samples, independent chest outcomes, and unchanged same-tier loot inside and outside key rooms. Breaking even counts as covering the key cost.
Recombined rooms are not additional observations. The published default includes a sampling check; custom setups use a labelled approximation where appropriate.
These calculations describe a published sample. They do not establish official drop rates, scanner accuracy, or guaranteed extraction income.
Hit rate asks how often a chest contained at least one Prismatic item. Loot share asks how much of the recorded loot was Prismatic. Multiple items in one opening count separately for loot share, and once for the opening hit rate.
Example: 10 openings contain 50 loot items. Two openings contain three Prismatic items in total. The hit rate is 2 ÷ 10 = 20%; the loot share is 3 ÷ 50 = 6%.
Opening hit rate = openings with at least one matching item ÷ complete openings.
Loot share = matching item occurrences ÷ occupied loot slots.
Loot count per opening = occupied loot slots ÷ complete openings.
Average is total displayed Coin divided by the number of complete-value openings. Median is the middle total after sorting. For totals 10,000, 20,000 and 90,000, the average is 40,000 and the median is 20,000.
Unknown Coin values exclude an opening from Coin statistics. Matched-item reference values can be included. Lowest and highest recorded values are sample observations, not the game’s limits. Displayed value does not mean every item was carried out.
Average = sum of known complete chest totals ÷ known-value opening count. For an even sample, median averages the two middle totals. Weight statistics separately require complete weights. Those denominators can differ from rarity statistics.
No observed hits does not prove the true chance is zero. The 95% hit-rate range uses a Wilson interval to express sampling uncertainty. It does not include selection bias, OCR mistakes or changes between game updates, and it is not a scanner reliability score.
With p = hits ÷ n, z = 1.96 and d = 1 + z²/n:
centre = (p + z²/(2n)) ÷ d
margin = z × √(p(1 − p)/n + z²/(4n²)) ÷ d
Range = centre ± margin, bounded by 0 and 1.
Ranges for separate chest tiers are not a combined confidence range for a multi-chest scenario.
The preset uses 1 Prismatic + 2 Golden chests, counting Gold and Prismatic loot, and subtracts a 100,000-Coin Gilded Key shop purchase. The key’s 70,000 displayed loot value is a different figure.
Average room value adds the selected chest-tier averages. Average after key cost subtracts the purchase cost. A positive average does not guarantee a profit from one room.
Cost coverage recombines observed whole-chest totals and counts how many modeled combinations reach the cost. This preserves item relationships inside a chest, but assumes independent chest outcomes, representative samples, and unchanged same-tier loot inside and outside key rooms. These are modeled rooms, not additional recorded rooms. Breaking even counts as covering cost.
Room mean = Σ(quantity per tier × eligible tier mean). Net mean = room mean − key cost. Coverage = modeled combinations with total ≥ cost ÷ modeled combinations.
The default uses the exported desktop model and its sampling check. Custom calculations enumerate empirical combinations when tractable; otherwise they use 131,072 seeded room draws. Custom sampling checks use 400 stratified bootstrap resamples and 2,048 room draws per resample. The range includes simulation variation. More than 24 chests retains averages but omits coverage.
A complete chest with no selected-rarity loot contributes zero. An unknown selected value excludes that chest. Separate recorded minima may never have occurred together and do not establish a game-wide worst case. Inventory limits, extraction losses and other costs are not modeled.
Repeated-opening estimates assume independent outcomes and an unchanged hit rate. Missing several times does not make the next opening due for a hit.
No hits in n openings = (1 − p)ⁿ. At least one hit = 1 − (1 − p)ⁿ. For mixed tiers, multiply each tier’s no-hit probability before subtracting from 1. Golden equivalence uses the smallest whole quantity reaching one Prismatic opening’s observed hit chance; a value comparison instead uses the ratio of average Coin.
A specified streak probability is not the chance of finding such a streak anywhere in an entire play history. A zero-hit sample does not establish impossibility.
When there is no observed pip reading, the assumed mapping is one pip per 3 kg, rounded up and capped at five: 1–3 kg → 1, 4–6 → 2, 7–9 → 3, 10–12 → 4, and 13 kg or more → 5.
Egg percentages use the current recorded manual egg tally, including corrections and resets. Shadow Sparkling is tracked separately from Prismana. This smaller tally is not a population-wide nest drop-rate estimate.
The spreadsheet and website snapshot may update at different times. The website publication date describes its own exported sample.
This is an unofficial personal project. Website code and exported observations are separate from third-party game artwork.
Dates use the UTC calendar day from the stored source-file creation time. Copied or imported files can have different creation dates, so this is not a verified in-game opening time.
No accounts or uploads are needed. The site uses a published snapshot; it does not connect to my tracker or Google account. Private filenames, machine paths, exact capture times and personal collection entries are not exported.
You can reach Nishin on Discord: nisynn. For a recording question, include its public opening ID and snapshot identifier.
Contact availability depends on Discord privacy settings.
2e77a981a2c1
The snapshot date indicates export time, not when gameplay occurred. Website updates can change presentation without changing the recorded data.
