CS2 Trade-Up Contracts Explained: Rules, Odds and Floats
How CS2 trade-up contracts work in 2026, including Covert-to-knife trades, Souvenir inputs, output odds by collection and the normalized float formula.
A trade-up contract turns several skins you own into one skin of a higher rarity. For a decade it was a niche tool for crafting traders. Then two updates, one in October 2025 and one in May 2026, made it one of the most important mechanics in the CS2 economy. Here's how contracts work now, how the odds and floats are calculated, and why most of them lose money.
The standard contract
The basic rules:
- Ten inputs of the same rarity produce one output one rarity higher. The ladder runs Consumer Grade, Industrial Grade, Mil-Spec, Restricted, Classified, Covert.
- The output comes from the collections of your inputs. Mixing collections is allowed, and each collection's share of the inputs sets its share of the odds.
- StatTrak can't be mixed with normal items. Ten StatTrak inputs give a StatTrak output.
- Each input's collection needs a higher tier. An item that's already the top rarity in its collection has nothing to trade up into in a standard contract.
The collections directory shows the rarity ladder for every collection, and the Classified rarity page lists every skin at the tier just below Covert.
Covert contracts: knives and gloves
Valve's October 22, 2025 update extended the contract to the top tier:
- Five regular Covert items can be exchanged for one regular knife or one pair of gloves from a collection of one of the items provided.
- Five StatTrak Covert items can be exchanged for one StatTrak knife. There are no StatTrak gloves.
The possible results are the knives and gloves in the cases your inputs come from, so the inputs you choose decide which knives are possible. Knives made this way could not be traded or sold for their first week; Dust2.us reported the first batch reached the Steam Market on October 30, 2025.
Souvenir inputs
The May 22, 2026 update added another rule: Souvenir items can be selected alongside normal items of the same rarity. Their Souvenir attributes, including the tournament stickers, are removed, and the result is always a single normal item. Souvenir and StatTrak items can't be mixed.
That turns Souvenirs into ordinary ingredients. As of October 14, 2026, the cheapest Souvenir M4A1-S | Knight in Factory New was listed at $2,228.33 on cs2skins.gg, below the cheapest normal one at $2,502.59.
How the odds work
Each possible output's chance is its collection's share of the inputs, divided by how many items that collection has at the next tier.
Example: 7 inputs from Collection A, which has 2 items at the next tier, and 3 inputs from Collection B, which has 4.
- Each Collection A outcome: 7/10 × 1/2 = 35%
- Each Collection B outcome: 3/10 × 1/4 = 7.5%
Collections with fewer items at the next tier concentrate the odds, which is why traders favour them when targeting a specific skin. For Covert contracts, the same idea applies: each case's share of your five inputs sets its share of the result, spread across that case's knife or glove pool.
How the output float is calculated
Valve doesn't publish the formula. Community testing since the October 2025 update shows it works in three steps:
- Turn each input's float into its position within that skin's own float range: (float − minimum) ÷ (maximum − minimum).
- Average those positions across all inputs.
- Map the average onto each possible output's range: minimum + average × (maximum − minimum).
Older guides simply averaged the raw floats, and that shortcut no longer matches the game.
Why it matters: a worked example
The M4A1-S | Knight (Classified) and the AWP | Dragon Lore (Covert) both belong to The Cobblestone Collection, where the Dragon Lore is the only Covert. The Knight's float range is 0.00–0.10. The Dragon Lore's is 0.00–0.70.
Ten Knights at 0.05 each sit at position 0.5 in their range. Mapped onto the Dragon Lore's range, that gives 0.35: Field-Tested. The old raw-average method would have predicted 0.035, Factory New. To get a Factory New Dragon Lore under the normalized formula, you need Knights with an average float below 0.01.
That difference is worth thousands. As of October 14, 2026, the cheapest normal Dragon Lore was $10,385.00 in Factory New and $5,900.00 in Field-Tested, while ten Knights at the cheapest Factory New Souvenir price would cost about $22,283. That's a deeply negative contract even before fees.
What trade-ups did to prices
Covert contracts turned red skins into knife ingredients overnight. Ars Technica reported that the P90 | Asiimov's median Steam sale price jumped from about $10 to well over $100 within a day of the October 2025 update. As of October 14, 2026, the cheapest P90 Asiimov on cs2skins.gg ranged from $146.48 (Battle-Scarred) to $213.86 (Factory New). A Battle-Scarred MP7 | Bloodsport, another Covert input, started at $23.51. Browse the whole tier on the Covert rarity page.
Knives and gloves moved the other way. Pricempire's estimates, reported by Ars Technica and Dust2.us, put the drop in total CS2 market cap at a third to a half within two days, before a partial recovery.
Running the numbers
Before any contract, calculate expected value:
- List every possible output with its probability.
- Estimate each output's float from your inputs, and look up the price for that exterior.
- Subtract the selling fee you'll pay: about 15% on Steam, less on most third-party stores.
- Add up probability × net price for all outcomes, then subtract what you paid for the inputs.
If the result is negative, the contract is a gamble, not a trade. Compare input and output prices across stores on the marketplace comparison page, because buying inputs a few percent cheaper is often the difference between positive and negative.
Practical limits and risks
- Trade Protection. Items received in a trade can't be consumed for seven days, so fresh inputs can't go into a contract right away.
- Fees and spreads. Selling the output costs money, and rare outputs can take a long time to sell at the price you modelled.
- Valve can change the rules. Both major changes in the past year arrived in ordinary patches. A contract that looks profitable today can look very different after the next update.
- Most contracts lose. Popular profitable recipes get copied quickly, which pushes input prices up until the edge disappears.
