The reference price
Every vault decision rests on one number: what the skin is worth. Get it wrong and the contract will defend a fiction with perfect discipline. So the number is a median of completed trades, never a listing, and the pipeline that produces it is published rather than summarised.
Why a median, not an average
Skin prices have a long upper tail. The same AK-47 | Redline (Field-Tested) has a median sale price around $28 and a maximum around $773 — the expensive copies carry rare stickers or unusual float values, and they are genuinely worth that. An average would chase them; a median ignores them, which is the correct behaviour when the question is “what does this item trade for”.
Never a listed price
A listing is an ask, not a trade. Marketplaces are full of items listed at multiples of what anyone pays, and they sit there indefinitely at no cost to the seller. Only completed sales enter the reference price.
Three windows, blended by volume
A single window is either too jumpy or too slow. The pipeline takes the median over three horizons and blends them weighted by how many sales each one actually contains, so a week with two trades cannot outvote a month with two hundred.
The pipeline, as of the current snapshot
| Stage | Count |
|---|---|
| Items scanned | 21,334 |
| Eligible | 455 |
| In the reference basket | 150 |
| Priced with a fresh quote | 150/150 |
| Stale | 0 |
| 30-day notional | $2.9M |
| Share of eligible notional covered | 65% |
Eligibility
- A real 30-day sales history, not a listing history.
- Enough trades in the window for a median to mean something.
- A resolvable canonical name, so the same item cannot enter twice under two spellings.
Sector split
Weights are damped — notional to the power 0.6 — and capped per sector, so the basket is not three knives and a rounding error.
| Sector | Weight | Items |
|---|---|---|
| Rifles | 40.6% | 32 |
| Knives | 25.0% | 49 |
| Gloves | 12.0% | 47 |
| Cases | 8.0% | 10 |
| Pistols | 7.0% | 5 |
| Other | 6.4% | 6 |
| SMGs and heavies | 1.1% | 1 |
Top of the basket
Source: Skinport. These are the ten heaviest constituents of the CS2X reference index, which is also the shortlist most launches come from.
| Item | Weight | Price | 30d sales |
|---|---|---|---|
| M4A1-S | Fade (Factory New) | 3.06% | $192.67 | 357 |
| M4A1-S | Blue Phosphor (Factory New) | 2.41% | $558.59 | 85 |
| Desert Eagle | Blaze (Factory New) | 1.94% | $624.64 | 51 |
| AK-47 | Fire Serpent (Field-Tested) | 1.90% | $694.09 | 42 |
| AWP | Lightning Strike (Factory New) | 1.90% | $477.92 | 65 |
| M4A1-S | Printstream (Field-Tested) | 1.80% | $189.09 | 147 |
| AK-47 | Vulcan (Minimal Wear) | 1.53% | $364.29 | 58 |
| Desert Eagle | Printstream (Field-Tested) | 1.48% | $35.37 | 562 |
| Number K | The Professionals | 1.46% | $46.42 | 408 |
| AK-47 | Inheritance (Field-Tested) | 1.46% | $36.15 | 546 |
From the index to a coin
The index exists to measure the market; a coin only needs one item's price. The same median feeds both. On chain it arrives as a uint128 with eight decimals, keyed by keccak256(market_hash_name), written by signed reports and aggregated to a median of reporters as well as of trades — the oracle page covers that second layer.
Single source. Today the pipeline reads one marketplace. That is the largest unmitigated weakness in the design: a bad feed produces a confidently wrong reference price, and the breaker can flag an implausible move but cannot invent a second opinion.
Snapshot data last refreshed Sep 23, 2026 UTC.