Contrats Trade-Up CS2 — la mécanique complète, le float, l'EV

What a trade-up contract actually is

A CS2 trade-up contract is an in-game function that takes ten weapon skins of the same rarity tier and returns one weapon skin of the rarity directly above. The function lives in your Steam inventory under the wrench icon labelled "Trade Up Contract" and has been in the game, with the same core mechanic, since August 2014.

Three rules govern every contract:

1. All ten inputs are the same rarity tier (e.g. ten Mil-Spec, or ten Restricted, or ten Classified).
2. The output rarity is one tier above the inputs — never two tiers, never the same tier.
3. All inputs are non-StatTrak together, or all StatTrak together. You cannot mix. Souvenir skins are excluded entirely.

The contract is destructive: the ten input skins disappear from your inventory, and the single output skin is generated and dropped in. There is no undo, no refund, no cooldown. If you mis-clicked a $200 input and the contract eats it, it is gone.

The whole interesting math sits in three places: which collection the output comes from, which skin within that collection you land on, and what float value that output skin carries. The rest of this guide walks each layer end to end.

The five rarity tiers, and which ones trade up

CS2 weapon skins use a seven-tier rarity ladder, but only the middle five participate in trade-up contracts:

You cannot trade up to a Covert and then trade those Coverts up again to a knife — knives are not on the weapon-skin rarity ladder. The only path to a knife output is the two-step covered later in this guide, and it only works on specific legacy collections.

The rarity colors are identical to the case drop tiers explained in the cases pillar, but the trade-up system uses them differently: a case picks one of five tiers with fixed percentages, a trade-up always promotes by exactly one tier.

The collection weight — what determines your output pool

Every CS2 weapon skin belongs to exactly one collection: Dust II, Cache, Mirage, Cobblestone, Italy, Train, Inferno, Norse, the various case collections (Kilowatt, Gallery, Fever, Revolution, Recoil, Dreams & Nightmares, Snakebite, Operation Riptide, Operation Broken Fang, Glove, Clutch, Shattered Web, Spectrum, Spectrum 2, Falchion, Shadow, Wildfire, Phoenix Weapon, Operation Bravo, Operation Vanguard, Operation Hydra, eSports 2013, eSports 2013 Winter, eSports 2014 Summer, Chroma, Chroma 2, Chroma 3, Revolver, Gamma, Gamma 2, Glove, Vertigo, Office, Nuke, Lake, Safehouse, Aztec, and dozens of others). Each collection contains a fixed list of skins per rarity tier.

When you submit ten inputs, the contract counts how many inputs came from each collection. That count is the weight for drawing the output collection:

• 10 inputs from the Cache Collection → 100% chance of output from Cache Collection.
• 5 inputs from Cache + 5 from Dust II → 50/50 chance of output collection.
• 7 inputs from Cache + 3 from Mirage → 70% Cache, 30% Mirage.

Once the output collection is picked, the contract picks one skin uniformly at random from the output-rarity skins in that collection. If Cache has 3 Restricted skins and the output is from Cache, each of those 3 skins has a 1/3 chance.

This is the key controllable variable. By choosing your input collections, you decide which output skin pool the contract draws from. Collections with one Restricted skin give you a 100% chance of that specific Restricted (conditional on the collection being drawn). Collections with five Restricted skins dilute the output across five possibilities.

The float averaging formula

The output float is fully deterministic given the ten input floats. The formula:

Output float = (average of 10 input floats) × (skin_max − skin_min) + skin_min

Where skin_max and skin_min are the float clip boundaries of the output skin — the upper and lower wear values that specific skin can actually exist at. These are per-skin: AK-47 Redline starts at 0.10 and never reaches FN; AWP Asiimov runs the full 0.18–1.00 BS range and never appears Factory New; AK-47 Case Hardened maxes out around 0.70.

A worked example. You feed ten Field-Tested inputs with floats averaging 0.18. The output skin clip is 0.00–0.50 (Mil-Spec MAC-10 Heat as an example). The math:

• Average input float = 0.18
• Output float = 0.18 × (0.50 − 0.00) + 0.00 = 0.090
• That float lands the output in Minimal Wear (0.07–0.15)

The formula is the same for every contract, every tier, every collection. The full derivation, the edge cases (clipped outputs, sub-0.00 mappings, the rounding behaviour), and worked examples on twenty real recipes are in the float averaging formula deep-dive.

The expected value of a trade-up

EV per contract = sum across all possible outputs of (probability × net output price) − total input cost.

Net output price means after Steam fees — 15% on a Steam Market sale, or whatever the seller-side fee is on the marketplace you actually use. Most community calculators forget this and quote gross EV, which inflates the number by ~17.6% and turns negative EV into apparent positive EV.

A worked Mil-Spec-to-Restricted contract, sampled from public listings in early Q2 2026:

Positive EV recipes do exist, especially when you control both the input collection and the float to push the output into a wear bracket that prices well. They tend to live in the lower rarity tiers (Industrial-to-Mil-Spec, Mil-Spec-to-Restricted) where input prices are low enough that one strong output can dominate the average.

The catch is variance. A recipe that returns $12.80 only 1 in 3 times will produce a loss on the other two runs, even with positive EV. You need to run the same recipe 20-50 times for the law of large numbers to converge. Profitable recipes for budget operators are catalogued in the profitable $0.50–$5 trade-ups article.

The Steam fee — the silent EV killer

Every output skin sold on Steam Market pays a 15% fee (5% Steam + 10% game). A skin listed at $10 nets the seller $8.70. EV calculations that quote the buy-now price of outputs overstate by 17.6%, which is enough to flip a real-life −5% EV recipe into an apparent +12% EV recipe on paper.

Third-party marketplaces have different fee structures: Buff163 effective seller fee around 2.5%, Skinport seller fee 12% with payout flexibility, DMarket and CSFloat in the 5-10% range. If you are valuing the output as cashout on Buff163 rather than Steam, redo the EV math with that fee. The marketplaces pillar covers each fee structure.

Knife and glove trade-ups — the two-step path

The standard trade-up rules say outputs are weapon skins, one tier above the input rarity. Knives are not on the rarity ladder; gloves are not on it either. So a single contract cannot output a knife or a glove.

But for a small set of legacy collections, the same collection that contains Covert weapons also contains the case-drop knives or gloves at the rare-special tier. The two-step path leverages that:

1. Step one: trade up ten Classified skins from collections that include knife-bearing case lineups (the Bravo, eSports 2013, Vanguard, Phoenix, Breakout, Huntsman, Operation Wildfire collections all qualify in their original case form). The output is a Covert from that collection.
2. Step two: trade up ten of those Coverts to a single knife from the same collection's rare-special pool.

This only works on collections that pre-date the modern case format and were structured with the rare-special slot tied to the same collection list. Modern case collections (Kilowatt, Gallery, Fever) do not let you knife-trade-up because the knife model is exclusively a case rare-special drop, not a collection-tier item.

The two-step path is expensive — you are stacking 100 input skins on the Classified side to land 10 Coverts to feed the second contract — and the EV is usually worse than just buying the knife outright. The detailed walkthrough and the current list of viable collections live in the knife-trade-up spoke when it ships in Phase 4. The shorter primer is in the existing profitable trade-up recipes article.

What you cannot do with a trade-up

The mechanic is narrow on purpose. The things players try and find are not allowed:

• No Souvenir trade-ups. Souvenir skins are not eligible as inputs or outputs. The Souvenir tag is preserved through pickup, not through trade-up generation.
• No mixed StatTrak. All ten inputs are StatTrak together, or none are. Mixing rejects the contract.
• No skipping tiers. You cannot go from Mil-Spec directly to Classified. Always exactly one tier up.
• No charm or sticker trade-ups. Charms (keychains) and stickers are separate item types with no trade-up mechanic.
• No Contraband trade-ups. The M4A4 Howl is the only Contraband item and it has no input or output role.
• No agent trade-ups. Agent skins are a separate category, not on the weapon rarity ladder.
• No knife/glove direct trade-ups — see the two-step section above for the only path.

The collection economics that decide whether a recipe is profitable

Three structural factors decide whether a given trade-up recipe makes money in the long run:

Input price per slot

The cheaper the input, the higher the ROI percentage for a given output mean. A $0.50 input × 10 = $5 stake; a $5 output average is +0% gross EV (and −15% net after Steam fees). A $0.40 input × 10 = $4 stake with the same $5 output averages a +20% net EV. Twenty cents per input matters more than the headline output number.

Output dilution

A collection with one Restricted skin (conditional on being drawn) gives you 100% certainty of the specific output. A collection with five Restricted skins dilutes across five different outputs at 20% each. The dilution is what kills most cross-collection recipes — adding a second collection looks like it diversifies the output, but if the second collection has more Restricted skins than the first, you actually increase dilution and reduce the expected output value.

Float distribution of input pool

Float-controllable input pools (skins that exist in tight float bands and trade cheaply at the bottom of their range) let you push outputs into clean exteriors. Float-uncontrollable pools (skins that only ship in Field-Tested or Well-Worn) cap the output exterior at the dirtier end of the output skin's clip. Pushing a Battle-Scarred output to Factory New is impossible if the input pool itself does not produce sub-0.07 averages.

For the per-tier list of which input pools actually have float control, see the float averaging formula deep-dive.

Where pricing data lives

The Steam Community Market is the canonical source for input and output skin medians, with 7-day history visible on each listing. Third-party aggregators (Buff163, Skinport, DMarket, CSFloat, CS.Money) carry the same items at different prices and fee structures — Buff163 is the most liquid by volume globally, Skinport tends to lead on EU availability, CSFloat has the cleanest float-conditioned pricing. The marketplaces pillar covers each one.

For ongoing EV calculations on specific recipes, the /tools/trade-up-calculator page (shipping in week 19) will let you paste an input list, pick output collections, and see live EV with the Steam fee subtracted. Until then, the math here plus a spreadsheet is the workflow.

How trade-ups plug into the broader inventory system

A trade-up is the second of three ways to generate a CS2 skin (the first is opening a case, the third is finding one in your weekly drop). Each has its own EV profile:

• Case opening: high variance, almost always negative EV at retail. Headline knife pull, but 80% of openings die at Mil-Spec.
• Trade-up contracts: lower variance per attempt, occasionally positive EV with the right inputs. No knife output without the two-step.
• Match drops: free, but rate-limited to ~1 skin per week. Drop pool is the active map collections, mostly Consumer/Industrial/Mil-Spec.

The economic insight: trade-ups make sense when you have a positive-EV recipe and the patience to run it 20+ times. Otherwise the math is variance-dominated and you are better off direct-buying the output skin from a marketplace. The skin investing pillar covers when trade-ups belong in a broader portfolio.

Why this pillar matters for the rest of the site

Trade-ups are the only deterministic mechanism in CS2 for generating a specific skin from inputs you control. The float averaging formula is the deepest piece of skin-economy math that has no equivalent in any other game's item system. Three downstream things depend on the rules here:

• Per-skin pages in the items encyclopedia link upstream to the collections that can produce the skin via trade-up, so buyers can see whether a target skin is reachable from a contract.
• The trade-up calculator (/tools/trade-up-calculator, shipping in week 19) reads the formula here as its specification.
• The investing pillar's skin investing view of trade-ups as an asset-generation strategy uses the EV math here as the baseline benchmark.

The ten-input rule, the collection weight, the float averaging formula, the StatTrak parity, the Souvenir exclusion, the knife two-step — every other piece of trade-up analysis on this site builds on those six facts.