How to Use CS2 Trade-Up Probability and EV Wisely
The numbers below are illustrative examples.
CS2 trade-up probability measures which output you can hit, while expected value is the probability-weighted average of outcome prices. Expected P/L is that average minus input cost. Use both with the share of outcomes above cost: a contract with +$36 expected P/L still has a 35% probability of losing money if the output distribution is volatile. Live board figures are estimates after fees. A trade-up can lose money.
Sort live trade-ups by Above cost % to see probability in practice, or run scenarios in the CS2 trade-up calculator before risking inputs.
How Trade-Up Probabilities Work
When you submit 10 inputs spanning multiple collections, the game picks one output skin from the next rarity tier. The probability of each outcome is proportional to how many inputs came from that skin's collection.
Say you use 7 inputs from Collection A (which has 2 skins at the next rarity: Skin X and Skin Y) and 3 inputs from Collection B (which has 1 skin at the next rarity: Skin Z). The weights are:
- Collection A total probability: 7/10 = 70%
- Collection B total probability: 3/10 = 30%
Within Collection A, each next-rarity skin gets an equal share of that collection's share of the 10 inputs: Skin X = 35%, Skin Y = 35%. Collection B has only one option: Skin Z = 30%. Those probabilities sum to 100%.
This weighting is deterministic. There's no hidden RNG on top of it — the game generates a random number, maps it to these weights, and you get that outcome. Over enough trade-ups, your results will converge on these exact probabilities.
Expected Value: The Formula
Expected value is the probability-weighted average of outcome prices. Prices and expected P/L on the live board are after sale fees. The formula:
EV = sum(probability_i * output_price_i)
Expected P/L subtracts total input cost from that average:
Expected P/L = EV - total_input_cost
Using our example above, suppose Skin X is worth $120, Skin Y is worth $40, Skin Z is worth $200, and your total input cost is $80. This example omits sale fees, so the prices and expected P/L below are before fees.
EV = (0.35 * $120) + (0.35 * $40) + (0.30 * $200)
EV = $42 + $14 + $60 = $116
Expected P/L = $116 - $80 = +$36
Positive expected P/L. On average, this trade-up returns +$36 per attempt. Run it 100 times and you'd expect roughly +$3,600 across those attempts. That's the theory. Dollar examples later in this post omit sale fees unless they say the prices are after fees.
Why EV Alone Is Misleading
That +$36 expected P/L hides important information. Look at the individual outcomes:
- Skin X ($120): 35% probability, P/L = +$40
- Skin Y ($40): 35% probability, P/L = -$40
- Skin Z ($200): 30% probability, P/L = +$120
At these before-fee prices, outcomes above cost are 65% (Skin X + Skin Z). Outcomes below cost: 35% (Skin Y). The expected P/L is positive, but you lose money more than one-third of the time. If you run this trade-up once and hit Skin Y, you're down $40. The +$36 expected P/L is real, but it only materializes over many repetitions.
Now consider a different trade-up: one possible output, estimated at $84 after fees, for $80 of inputs. Its probability is 100%. EV = $84. Expected P/L = +$4. Boring, but there's no outcome variance: every result is the same skin. It can still lose money if the output's price drops before you sell. For someone doing one trade-up, expected P/L is one average number, and it does not show how far the individual outcomes spread above or below it, so the single-outcome case may suit better than a +$36 expected P/L contract with a 35% probability of losing $40.
Outcomes Above Cost: The Practical Metric
The share of outcomes above cost measures the probability that the trade-up produces an output worth more than your total input cost after fees. It counts only those outcomes, weighted by their probabilities. It doesn't care how much you profit or lose — just whether the result is green or red.
A trade-up with 90% of outcomes above cost is one where 9 out of 10 outcomes are worth more than your inputs. You might be up $5 on those and down $30 on the other one, with expected P/L of +$1.50. The expected P/L is tiny, and the one outcome below cost costs as much as six of the good ones make.
A trade-up with 20% of outcomes above cost but high expected P/L is the opposite: most attempts lose money, and one rare, expensive outcome pulls the average positive. It's positive expected P/L with very high variance. Mathematically sound, emotionally brutal.
Neither metric alone tells the full story. You need both.
Risk Profiles: High-Variance vs Low-Variance
High-variance profile: a low share of outcomes above cost (15-35%), high expected P/L. These trade-ups have one or two expensive outcomes and several cheap ones. Most outcomes land below cost, and the few above it are large. A knife trade-up where 4 out of 5 possible outputs are worth less than your inputs, but the 5th is a $3,000 Karambit Fade, fits this profile. Expected P/L might be +$80, but you're losing money 80% of the time.
This only fits if you can afford repeated losses and do enough volume for the average to show up. Ten attempts at $200 each ($2,000 total) with 20% of outcomes above cost and +$80 expected P/L have an expected P/L of about +$800 across all 10. That's an estimate after fees, and the real result can be negative: 6-7 outcomes below cost in a row is a normal run. Can your budget absorb that?
Low-variance profile: a high share of outcomes above cost (75-100%), modest expected P/L. These trade-ups have most or all outcomes above breakeven. Individual profits are small — $3 to $15 typically — but losses are rare. A Classified-to-Covert trade-up where 8 of 10 outcomes are profitable and the other 2 lose only a few dollars fits here.
This suits smaller budgets and traders who want lower variance. Ten of these at $50 each ($500 total) with 85% of outcomes above cost and +$5 expected P/L per trade-up come to about +$50 across all 10, with a narrower spread of results. Expect small numbers, and a trade-up can still lose money.
When Negative EV Trade-Ups Make Sense
This sounds contradictory, but trade-ups with negative expected P/L can be rational under specific conditions.
Suppose a trade-up has −$5 expected P/L but 60% of outcomes above cost, with this distribution: 60% probability of +$15, 40% probability of −$35. Expected P/L is (0.6 * $15) + (0.4 * -$35) = $9 - $14 = -$5. The math says don't do it.
But what if the 40% loss outcome produces a skin you actually want to keep and use? Or what if the $15 profit outcome produces a skin with high trade velocity that you can flip immediately, while the loss outcome is a skin that will eventually recover in value? Context matters beyond raw expected P/L.
The board includes trade-ups with negative expected P/L. The Min above cost % filter narrows the list to ones where more outcomes land above cost. A trade-up with −$3 expected P/L, 70% of outcomes above cost and a worst-case loss of $8 has limited downside and mostly outcomes above cost. Its expected P/L is still slightly negative.
The opposite is also true: positive expected P/L doesn't automatically make a trade-up worth doing. A trade-up with +$50 expected P/L, 5% of outcomes above cost, and $500 input cost means you lose money 95% of the time and need to do it 20+ times for the average to converge. Unless you have $10,000+ allocated to this single trade-up type, the variance will eat you alive.
Putting It Together
The best trade-ups score well on both metrics: positive expected P/L and a high share of outcomes above cost. These are rare, which is why discovery engines exist — manually finding trade-ups where most outcomes are profitable AND the expected P/L is meaningfully positive requires evaluating thousands of listing combinations.
When you have to choose between the two metrics, let your budget decide. Large budget and high volume? Weight expected P/L more; you can absorb the variance. Small budget and a few trade-ups a week? Weight the share of outcomes above cost more. Lower variance matters more than expected P/L when you can't afford several losses in a row.
TradeUpBot lets you sort by either metric. Sort by Expected P/L to find the highest expected P/L. Sort by Above cost % to find the trade-ups where the largest share of outcomes lands above cost. Expand any trade-up to see the full outcome distribution — every possible output, its probability, and whether it's above or below breakeven. That distribution is the trade-up. Everything else is just a summary of it.
Published 2026-03-21 by TradeUpBot Team.