The Roxetta Stone

Valve has never published the number of capsules they sold for any Major. They have, however, dropped one figure in passing, and that one figure is enough to put a hard ceiling on everything else.

In August 2022, in a blog post celebrating the 10th anniversary of Counter-Strike: Global Offensive, Valve mentioned that participating teams and players had received over $70 million from sticker sales over the previous year. That window covers the two Majors of the period:

• PGL Major Stockholm 2021 (October-November 2021)
• PGL Major Antwerp 2022 (May 2022)

That is the only first-party revenue figure Valve has ever shared for these two Majors. Every estimate on this page is anchored against it.

The Headline

From Teams' Cut to Total Sales

Major-sticker revenue is split 50/50 between Valve and the participating teams. That arrangement has been Valve's standing policy since 2014 and has never changed. Every team capsule listing on the Steam Market spells it out: "50% of the proceeds from the sale of this capsule support the included organizations".

So if teams collected over $70m, Valve collected the same. The total gross sales for the Stockholm + Antwerp window comes to:

That $140 million budget is the starting point for every other number on this page. The capsule supply, the souvenir split, the per-capsule breakdown, the rox (Holo) ceiling: each of them has to fit inside the $140m. None of them can be solved from $140m alone, but every one of them has to be consistent with it.

Why This One Number Helps So Much

It does not give us every other number on its own, but it puts a hard ceiling on what the rest can be. Once we know the total amount of money players spent across both Majors, we also know the maximum number of capsules and souvenirs that could have been produced and sold. Multiply average prices by quantities and the total has to land at $140m. Not a dollar more. Move any single variable too far in either direction and the math breaks. The other inputs (price ranges, share splits, scrape data) still do real work, but $140m is the budget they all have to fit inside.

The next section does exactly that: it splits the $140m into capsules, souvenirs, and the Stockholm versus Antwerp shares.

Not all of that $140m became capsules. Several products shipped at every Major and shared the same revenue:

• Sticker capsules: team capsules and autograph capsules, all released at $0.99 and dropped to $0.25 on later sales, plus a Champions capsule released after the final.
• Souvenir packages: match drops or direct-purchase packages (~$3 each via Viewer Pass tokens) containing weapon skins from the Major's active map pool.
• Viewer Pass and patch packs: a small but real slice on top, around $5m combined for SH and AW. The Pass unlocks the Pick'em challenge, souvenir token drops, and a few patches for sale.

Splitting those is the first thing we have to do.

Step 1 · Souvenirs vs Capsules

The souvenir side is the trickiest part of the model, but the count is more solid than it looks. Each souvenir package on Steam carries the name of the major it came from. Stockholm 2021 Mirage souvenirs are tagged "Stockholm 2021 Mirage Souvenir Package", and so on. The skin that drops keeps that link via the gold sticker on top, plus a hidden tournament tag that does not vanish even if the gold sticker is later peeled off.

So in 2022 and 2023, when Stockholm and Antwerp were the only majors that ever shipped these specific Riptide-era souvenir collections, you could query CSFloat DB for the souvenir versions of those skins and get the full SH+AW total at once. The number landed at around 13 million packages opened, including the ones with peeled-off gold stickers.

Today the same query is murkier because later majors reused the same Riptide collections inside their own souvenirs, so a "souvenir Mirage skin" is no longer purely SH+AW. The live count you see at CSFloat DB has shifted to filtering by the SH or AW gold sticker on top, which is why it now reads ~10.9M (a subset that misses every package whose gold has been peeled off and re-applied to a knife). On top of that, the public monthly scan series only covers the first months of each major. After the initial wave the per-month snapshots stopped getting taken, so the months that came after were never recorded as a separate per-map breakdown.

Even with those gaps, the months that were scanned tell a clear story. The breakdown below shows, per map, how many packages were opened in each monthly snapshot in the public dataset. It is the largest dataset publicly available for SH and AW souvenirs:

Stockholm 2021 Oct 2021 to Jul 2022 · 10 months 10.07M scanned

Month Mirage Dust 2 Ancient Inferno Nuke Vertigo Overpass Month total

Oct ’21 441k 184k 148k 59k 84k 51k 93k 1.06M

Nov ’21 1.51M 750k 535k 123k 182k 90k 103k 3.29M

Dec ’21 836k 312k 148k 19k 48k 47k 34k 1.44M

Jan ’22 625k 187k 70k 9k 32k 20k 10k 954k

Feb ’22 594k 152k 66k 5k 12k 35k 8k 872k

Mar ’22 555k 172k 39k 4k 16k 31k 9k 825k

Apr ’22 359k 148k 51k 2k 10k 35k 5k 610k

May ’22 242k 125k 74k 8k 9k 43k 2k 503k

Jun ’22 148k 86k 20k 2k 5k 16k 2k 280k

Jul ’22 156k 39k 16k 2k 3k 4k 4k 224k

Map total 5.47M 2.16M 1.17M 233k 401k 371k 270k 10.07M

Antwerp 2022 May 2022 to Jul 2022 · 3 months 3.04M scanned

Month Mirage Dust 2 Ancient Inferno Nuke Vertigo Overpass Month total

May ’22 668k 367k 109k 144k 91k 51k 46k 1.48M

Jun ’22 449k 191k 117k 77k 70k 23k 20k 948k

Jul ’22 352k 121k 31k 41k 30k 31k 9k 615k

Map total 1.47M 680k 258k 261k 191k 105k 76k 3.04M

Cell saturation · log scale

1k 10k 100k 1M peak

Three things jump out. First, Mirage absolutely dominates: it tops every single row, with totals roughly twice the runner-up Dust 2. Second, the opening curve falls off fast: November 2021 alone (3.29M) is more than the last six months of Stockholm put together. Third, the captured monthly data already adds up to 13.1 million packages opened, which matches the historical CSFloat DB query for SH+AW souvenir skins almost exactly. Few packages have been opened since, so 13M is the floor of the real opened count, not a ceiling.

Now layer in the unopened tail still sitting in inventories four years on, plus a small allowance for packages opened during months that were never scanned, and the realistic SHAW total lands around ~27m souvenir packages at an average of $3 each (Viewer Pass tokens push the price slightly above the $2.99 listed). The remaining ~14M between the scanned 13M and the modeled 27M is mostly unopened tokens still sitting in storage units, plus the slack you would expect from a four-year-old market. So the breakdown looks like this:

~$86m Souvenirs + Pass ~27m packages · $3 avg

~$54m Capsules ~192m capsules · ~$0.28 avg

Combined Stockholm + Antwerp gross sales $140m

A couple of notes on that bar. The souvenir + Pass count comes from CSFloat DB, but with two corrections layered on top of the live read. First, souvenirs whose gold sticker has been removed (the most popular thing to do with the gold sticker is to peel it off and re-apply it on a knife) drop out of CSFloat DB. Second, a small unopened tail still sits in inventories four years on. Both corrections push the realistic total into the mid-twenty-million range. The capsules side, in turn, comes out smaller than people typically expect because most of the budget went to souvenirs, not capsules. Almost every capsule was bought during the recurring 75% off sale at $0.25, so the average price across all capsules sold lands at about $0.28, just above the sale floor.

Step 2 · Stockholm vs Antwerp

Now we split each pool between the two Majors. They did not share evenly. Most capsules went to Antwerp, most souvenirs went to Stockholm, and both pools tilt by similar amounts in opposite directions:

Capsules ~192m units · ~$54m · mostly Antwerp

Stockholm 2021 ~66m capsules ~$17m capsule revenue

Antwerp 2022 ~126m capsules ~$31m capsule revenue

Souvenirs ~27m units · ~$81m · mostly Stockholm

Stockholm 2021 ~19m souvenirs ~$56m souvenir revenue

Antwerp 2022 ~8m souvenirs ~$25m souvenir revenue

The capsule split comes straight from the post-sale Steam Market paper-listings ratio between the two Majors. Antwerp ended its sale with ×1.89 times more papers listed than Stockholm, which is the cleanest cross-major proxy we have for total capsules opened. The souvenir split comes from CSFloat DB applications directly. Both numbers point in the same direction with the same shape.

So why does each pool tilt the way it does? Four structural shifts stack up into this mirror split:

Why Stockholm Sold More Souvenirs and Antwerp Sold More Capsules

Putting the two steps together: Antwerp lands at roughly ~120m capsules in total. The next section cross-checks that figure against post-sale Steam Market data across nine Majors, and Section 4 stress-tests it against the $140m budget.

Section 2 lands on a 1 to ×1.89 split between Stockholm and Antwerp. That split, and the Antwerp total it produces, deserve an independent confirmation. Steam Market post-sale paper listings give us exactly that. We compare them against the two Majors with publicly known total supply: RMR 2020 and Paris 2023.

Why Papers, Not Capsules

An earlier version of this section used capsule listings to compare Majors. They turn out to flatten the supply differences badly: Antwerp and Paris show up roughly equal on capsule listings (1,451k vs 1,420k) even though Paris shipped five times more capsules than Antwerp. The reason is more subtle than "capsules are bought to be opened". Plenty of capsules are bought as long-term holds, especially by the bigger investors. The issue is that most retail unboxers open theirs, while the bulk of stockpile-style buyers do not list, and once a capsule sale ends a heavy share of the visible listings comes from people offloading leftovers from short-term flipping, not from anything tied to the original supply. Paris shipped massive supply, but a lot of it sat in unopened bulk inventories and never made it to the listings page.

Papers behave the opposite way. They are the most common drop, almost always treated as low-value by the unboxer, and listed at the floor price pretty much immediately. So listing volume scales with capsules opened, not with leftover demand, and the cross-major paper ratios keep the supply gaps intact. The chart below puts every Major on the same scale, with Paris 2023 set to 100% so the gap is impossible to miss:

0 25% 50% 75% 100%

RMR 2020 anchor

528k

Stockholm 2021

112k

Antwerp 2022 our target

211k

Rio 2022

104k

Paris 2023 validation

1,764k

Copenhagen 2024

257k

Shanghai 2024

94k

Austin 2025

51k

Budapest 2025

167k

Source: post-sale SCM paper listings snapshot, March 2026, compiled by [CygaPb] RoadRunner from the CS Market Forum Discord. Bars normalized so Paris 2023 = 100%.

The RMR Anchor and the Paris Validation

We need two majors with publicly known supply to make this method honest. RMR 2020 is the anchor: the supply is known from leaked team revenue figures (Astralis publicly reported $1m of capsule revenue from RMR, NaVi confirmed similar figures, and Lewis the analyst gathered enough team-level disclosures to put the total at ~250m capsules). RMR also predates Storage Units, so every paper that dropped had to be either applied or listed.

Once we have a single anchored ratio, we still need to know that the relationship holds for other majors. That is where Paris 2023 comes in: it is the second major with a public revenue figure ($110m to teams, $220m total Valve gross via the HLTV leak), so we can plug it into the same equation and check the answer comes back close. If RMR-papers / RMR-supply matches Paris-papers / Paris-supply, the ratio works for the rest of the majors. If not, the method is broken. The rest of the supply numbers in the table fall out of that ratio.

So we have a hard ratio: 528k papers per 250m capsules ≈ 2.1 papers listed per million capsules opened. Apply that ratio to every other major and the implied total supply matches the framework analysts have been building for years on the forums:

Paris 2023 closes the loop on the validation. Plug Paris into the same equation and you get 2.6 papers per million capsules, about 25% more than RMR. The gap is clean and easy to explain: Storage Units take in a slice of the post-2021 paper supply that RMR did not have to deal with. RMR sits at the lower end of the range, which gives the more conservative supply estimates downstream.

The Direct AW vs SH Comparison

Comparing Antwerp directly to Stockholm is the cleanest read because both Majors share almost every structural detail: same generation of storage units, same regional market constraints, same audience, same currency exchange windows. Whatever bias affects the listings count affects both rows the same way:

Stockholm 2021 112k ~66m capsules

Antwerp 2022 211k ~126m capsules

Antwerp papers ÷ Stockholm papers ×1.89

Stockholm and Antwerp combined sold $140m worth of items. That number is fixed. So whatever the supply numbers are, when you multiply each capsule and each souvenir by its average price, the total has to land at $140m. That is the budget.

Here is what a "fits the budget" check looks like with the realistic ~126m Antwerp estimate that both Section 2's budget math and Section 3's listings ratio agree on. Every $1 in the bar below maps to the same number of pixels:

Souv + Pass ~$86m

SH caps ~$17m

AW caps ~$31m

$140m budget

Total $134m in souvenirs and capsules · the remaining ~$6m sits in patch packs and rounding

Souvenirs and the Viewer Pass take more than half the budget. The capsule side splits roughly 1:2 between Stockholm and Antwerp, in line with the paper-listings ratio. The bar fills almost the full budget; the small gap to the right is rounding plus a few patch packs.

Do the same exercise but pretend Antwerp shipped 250m capsules instead of 126m. That would push Antwerp capsule revenue from ~$31m to ~$70m at the average $0.28 sale price. Stockholm cannot drop too far either, because the listings ratio and the souvenir math bound it from below. Watch what happens. The dashed line marks where the $140m budget ends:

Souv + Pass ~$86m

SH ~$17m

AW caps ~$37m

AW over +$33m

$140m budget

Total $173m · $33m more than Valve actually collected

That is what "the math collapses" looks like. The dashed budget line cuts cleanly between AW caps and AW over: everything to the right of it (~$33m of capsule revenue) does not exist anywhere in Valve's disclosure.

There is no honest way to fit 250m Antwerp capsules plus the rest of the items into the $140m budget:

• Souvenirs would have to nearly disappear, but CSFloat DB still has 13M registered openings on record (and that count is a floor, with unopened tokens still sitting in inventories on top of it).
• The average capsule price would have to drop unrealistically low, well under the $0.25 sale floor.
• Stockholm would have to be near zero, contradicted by its own paper and capsule listings.

None of those work. The same reasoning runs in the other direction: if Antwerp had been much smaller than 126m, the capsule revenue would not fill enough of the budget to leave room for the souvenirs and the Stockholm share we know existed.

The Stress Test

One last way to see this clearly. Hold souvenirs and Stockholm at their realistic estimates. Vary Antwerp from very small to very big. Read what each row implies for the total:

Only one row balances. That is why the Antwerp estimate lands where it does. The $70m teams payout is not just a single data point. It is also a budget that boxes the entire model in from above and from below.

Sections 2 to 4 land on a single number: ~126 million Antwerp 2022 capsules in total. Antwerp did not ship one capsule though, it shipped seven. Each one has a different role, a different sticker pool, and a different drop rate ladder. To count rox holos, we first have to identify the exact capsule they come from.

How the 126m Splits Across the Seven

The 126m total comes out of the budget math (Section 2) and the cross-major paper-listings ratio (Section 3). The per-capsule split is worked out a different way. We use live sticker application data on CSFloat DB, anchored by the public drop rates and the constraint that the seven numbers have to sum back to ~126m.

For every capsule, we count how many stickers from each of its rarity tiers have actually been applied on weapons. Then we divide by the drop rate and correct for the rough share of stickers that get pulled but never applied. That gives the supply each capsule must have had for CSFloat DB's numbers to make sense:

applied stickers ÷ drop rate × pulled-vs-applied correction → capsule supply

Repeated for each tier of each capsule, then averaged. Tiers with more applications give tighter estimates.

Three constraints have to hold at once: every capsule's per-tier estimates have to land in roughly the same range, the seven numbers have to sum to ~126m, and the relative ordering has to match what we know about how each capsule was priced and marketed. The split below is the only set of numbers that hits all three at the same time.

Why the Numbers Sit Where They Do

The split is not arbitrary. The ordering tracks structural reasons:

• Team capsules outsold autograph capsules on average. The team logos appeal to a wider audience than individual players, and the team capsule is bought as the gateway to the souvenir-eligible team finish more often than for the rare auto drop.
• Legends capsules outsold Contenders or Challengers. Legends contain top-seeded teams and players (NaVi, FaZe, G2 etc.), which gets the most casual interest. Contenders has the lowest-seeded entries, which is why Auto Contenders sits near the bottom of the table, even though that is the capsule we care about.
• Auto Champions is the smallest. It contains autographs from the FaZe roster only, the Major champions, and was sold as a one-off post-tournament item.

So when we put the seven calibrated numbers in order, this is what comes out. The capsule that matters is highlighted:

The capsule that matters sits near the bottom: Auto Contenders, ~12 million units. It is one of the smaller cards in the deck, just 9.2% of the Antwerp total. Every rox (Holo) sticker in existence came from one of those capsules. Now we need to know how often a capsule actually drops one.

The Drop Rate Ladder

The Auto Contenders capsule contains 25 player autographs in four finishes: Paper, Glitter, Holo, Gold. Each finish belongs to a rarity tier with its own drop rate, sourced from the public capsule rarity table on csgoskins.gg and cs2roi.com:

The "rarity scale" column is log-scaled relative to Paper. Each chip's width drops with the order of magnitude: Paper at 1×, Glitter 5× rarer, Holo 26× rarer, Gold 125× rarer. Linear bars would make Holo and Gold disappear entirely.

Each row shows the chance of pulling any sticker in that tier from a single Auto Contenders capsule.

From "Any Holo" to "Specifically rox Holo"

The 1-in-31 number is the chance of pulling any Holo sticker from the tier. Inside the Holo tier sits one slot for each of the players in the capsule, and the tier randomly picks one. csgoskins.gg's published per-player chance is 1 in 40 for any specific player Holo within the tier (one slot per player plus a few decorative variants split the same draw). So the chance of pulling rox specifically works out to:

1 in 31 holo tier × 1 in 40 player slot → 1 in 1,248

≈ 0.0801% per Auto Contenders capsule

Independently confirmed by the csgoskins.gg drop chance calculator for this exact sticker.

That number looks tiny because it is. On average it takes 1,248 capsules to produce one rox (Holo). To get a sense of scale, here is the same odds expressed three different ways:

Closing the Loop: The Supply Ceiling

Two numbers fully define the rox (Holo) ceiling. Capsules ever produced, multiplied by the chance each one drops a rox holo:

12mcapsules × 1/1,248 = ~9,225

absolute rox (Holo) supply ceiling

That is every rox (Holo) sticker that can ever exist. Not "currently". Not "so far". Every single one, period. Once the last Auto Contenders capsule on Steam is opened, no more rox holos can ever be produced. The supply is hard-capped at around 9,200 stickers, set in stone by the original Antwerp 2022 capsule run.

The next two sections look at how many of those 9,200 have already surfaced, how many are still locked inside unopened capsules, and how the supply is split between applied and unapplied stickers.

Section 5 gave us the ceiling: ~9,225 rox holos can ever exist. That figure is the maximum, assuming every Auto Contenders capsule eventually gets opened. The number that has actually surfaced is smaller, and the number applied to weapons is smaller still. This section measures both.

What CSFloat DB Has Recorded

Every hour, ROX CULT scrapes CSFloat DB for new rox (Holo) sticker applications. Latest snapshot:

Why the CSFloat DB Number Is Almost Right

The 5,564 figure is good but not perfect. There are owners with private inventories, accounts with no scan history, and a long tail of stickers applied right after release that never hit the Steam Market. Old supply estimates used to apply a generous "miss rate" of 50% on top of the CSFloat DB number to account for all of that. The realistic correction today is much smaller, thanks to one number we measure directly.

Inside the Craft Insights panel on the dashboard, there is a counter for the total number of historical ownership transfers across every known rox-stickered weapon:

The 9-trades number is specific to rox crafts, not a generic Steam stat. rox (Holo) trades for $130 to $200 most of the time, and a craft built around it typically lives on a $200 to $20,000 weapon. The combination is high-value enough that owners list, trade, gift, sell, and re-buy these items constantly. Every transfer is a chance for CSFloat DB to scan the item:

pulled

1scanned

2scanned

3scanned

4scanned

5scanned

6scanned

7scanned

8scanned

9private

today

This is why the 9-trades figure works for rox specifically and not for, say, a $0.05 paper sticker. Cheap crafts trade rarely, sit in inventories for years, and may never appear on the Steam Market. That is where a 50% miss rate makes sense. For rox holo, given this trade volume and price tag, the realistic miss rate is closer to 5%:

5,564 applied CSFloat DB × 1.05 miss correction → ~5,842 applied (real)

A correction of 5% rather than the historical 50%, anchored on the directly-measured 9 trades per craft.

Applied Versus Unapplied

Once a rox holo comes out of a capsule it lives in one of two states. Either it has been applied to a weapon (permanent, CSFloat DB sees it), or it sits unapplied across one of four buckets: Steam Market listings, third-party marketplace listings, raw inventory, or a Steam storage unit. We model the unapplied share at ~30% of total pulled stickers, which gives:

~5,842 applied to weapons 70%

~2,500 unapplied 30%

Total pulled ~8,346

The 8,346 total is what we treat as "rox holos that have surfaced so far". Out of the 9,225 ceiling, that leaves roughly ~879 still locked inside unopened capsules. Most of the supply has already left the box, but a meaningful tail remains.

Where the Unapplied Sit

The 30% unapplied (~2,500 stickers) is where most of the uncertainty lives. Direct measurements only give us partial visibility:

• ~17 active listings on the Steam Community Market as of 15 May 2026, counting raw stickers and Sticker Slabs. Steam Market is the only place where anyone can buy without a third-party account.
• ~111 active listings cross-marketplace, per the csdb.gg aggregate across Steam plus CSFloat, Buff163, Skinport, CS.MONEY, DMarket and the rest. That is the visible supply any collector can find for sale at the time of writing.
• Low single-digit weekly sales on Steam Market for the raw sticker and slab variants combined. The market clears that volume before new listings show up.
• Sticker Slabs launched in November 2025 and absorb a chunk of the unapplied pool by design. Slabbing a rox holo is a way to display it without committing to a weapon. Some slabs are listed (already inside the 111 number), others sit in inventories.

Putting those measurements against the 2,500 estimate, here is where the unapplied stickers actually live as of the latest snapshot:

Steam Market listings ~17 0.7%

Third-party marketplaces ~94 3.7%

Held but not listed (inventories, storage units, slabs) ~2,389 95.6%

Listed for sale at the time of writing: ~111 across all marketplaces · The vast majority sits in inventories, storage units and slabs we cannot scan from the outside.

The 4.4% list rate is normal for a high-value collectible that most owners hold as an investment rather than active inventory. The long tail of "held but not listed" is the largest single bucket and the hardest one to count. Storage units in particular are completely invisible from the outside, even on public profiles, so a rox holo placed inside one is effectively hidden until the owner takes it out.

The next section turns these numbers into the final answer the page started with: how many rox holos will ever exist, when supply runs out, and what that means for the people holding one today.

Sections 1 through 6 carved out the inputs. This one closes the loop. We work the chain backwards from 5,564 applied stickers on CSFloat DB all the way back to capsules consumed and rox holos still locked away.

The Whole Chain in One View

Before unwinding the math step by step, here is the full funnel from Valve's $140 million disclosure all the way down to the rox holos applied today. Each step shrinks the number by an order of magnitude or more:

$140m Valve disclosure · SH + AW combined

subtract souvenirs & pass

~192m total SH + AW capsules ever sold

×0.71 to AW share

~126m Antwerp 2022 capsules

×9.2% are Auto Contenders

~12m Auto Contenders capsules rox source

×1/1,248 rox holo odds

~9,225 absolute rox holo ceiling

×~91% reached

5,564 rox holos applied today

Working Backwards: From Applied to Capsules Opened

Three multiplications. Each one inverts a factor we already calibrated:

1. 1
   CSFloat DB miss correction 5,564 × 1.05 = ~5,842 applied (real) +5% to account for the small share of crafts that never appeared in any of their nine ownership transfers.
2. 2
   Applied / pulled ratio ~5,842 ÷ 0.70 = ~8,346 pulled 70% of pulled stickers end up applied. The other 30% sits in inventories, listings, slabs, and storage units.
3. 3
   Capsules per rox holo ~8,346 × 1,248 = ~10.42m Auto Contenders opened Each rox holo costs an average of 1,248 capsules of trying, per the per-player drop rate from Section 5.

Out of the ~12 million Auto Contenders we estimated in Section 5, the math says about ~10.42 million have been opened. That leaves a small remaining pool:

~10.42m opened ~91%

~1,100k left

Of an estimated ~11.5m total Auto Contenders supply

The Three Headline Answers

The page started with three questions in the eyebrow. Time to answer them straight up:

The Application Curve

The shape of the rox holo application curve is the visual proof of a supply that is running out. A small first wave of crafting at $0.99 right after release in May 2022, then heavy crafting through the summer once the capsule went on sale at $0.25 (when the 75% off discount started in September 2022 and stuck), a steady decline as the pool drained, and small bumps each time the price spiked:

May 2022release Summer 2022peak ~640 / month Now~30 to 100 / month

The Capsule Price Tells the Same Story

If the supply estimate were wrong and there were plenty of capsules left, the Auto Contenders price would be flat. It is not. Capsule price has been climbing through 2025 and 2026, with the supply curve catching up to demand. People still want to crack rox holos (the sticker still trades north of $130) and there are fewer capsules left to crack:

Sept 2022 sale$0.20 floor Mar 2023$1.24 peak Now$0.78 climbing

The Numbers in One Place

Five anchor numbers, each with its source:

The Limits of This Analysis

Honest uncertainties, so the math does not pretend to be cleaner than it is:

• The $140m figure is rounded. Valve said "over $70m" to teams, which could be $71m or $75m. Does not move things much in either direction.
• The souvenir total (~12m units, $3 average) is read directly from CSFloat DB but still has a small error bar from the unopened tail. A 20% shift here would move the Antwerp capsule supply by ±10m.
• The applied/pulled ratio of 70% is the least certain number in the model. Direct measurements only see ~17 listings on Steam Market and ~111 across all marketplaces; the long tail of stickers held in private inventories and storage units is hard to count. Realistic range is 20% to 35% unapplied.
• The Auto Contenders supply at 13m has ±1.5m uncertainty. The split between the seven capsules comes from the same dataset, so a shift in one moves the others.
• Drop rates published by csgoskins.gg are calculated from observed unboxings, not from first-party Valve data. They are internally consistent and match cs2roi's independent calculation, but not officially audited.

The conclusion holds even if all of those move at once. Whether the rox holo ceiling sits a thousand stickers higher or lower than 9,200, whether 91% or 85% of the supply is already opened, the picture is the same. The capsule is running out, and the applied count is closing in on the maximum. There will not be many more new rox holo crafts than what already exists.

Reading the Stone

You should now be able to look at the chalkboard at the top of this page and read it like regular text. The Stockholm and Antwerp blocks split the $140m. The seven Antwerp capsules add up to roughly ~126m units. Auto Contenders sits at ~12m, and rox lives inside it at one in 1,248 odds. Paris on the right adds a sanity check at a much bigger scale. The little spelling slips were never the point.

The data behind this page will keep getting refined. As CSFloat DB adds new applications, as the unopened capsule pool slowly drains, as Cologne 2026 changes the way Major capsules even work, the numbers above will move slightly. The page stays live and updates as new evidence comes in. The estimates will tighten, never loosen.