I bought 17 CS2 skins using skinresale estimates. Here's the log.

Over roughly a month I bought 17 CS2 items on cheaper marketplaces, waited the Steam trade lock, and sold them. Capital was $87.68. Sales came to $98.48 ($10.80 / 12.3%). 14 of 17 closed green. Median hold: 8 days — that is the lock, not a market-timing hold.

The buy filter was the estimate on skinresale Compare: if the projected exit still cleared after the lock, I bought. After the sale I logged fact versus that estimate. The tool was useful as a filter and systematically high as a forecast. Median estimate 1.30x, median fact 1.15x.

The rule

1. Find a skin cheaper off Steam than the exit I actually use.
2. Check the skinresale estimate before sending the order.
3. Accept the ~7-day trade lock.
4. Sell. Write down fact next to the estimate. No deleting losers.

I did not buy because “the market moved.” Random week-old asks versus today’s realized sales are usually red after fees — that is the boring baseline. The whole game is not buying the median. For paths that need Steam balance in the middle, that loop is mapped on Steam Wallet Bridge.

What printed

• USP-S | Spiral Glitch (Battle-Scarred) — $14.49 → $17.72 (1.22x, $3.23) · UUSKINS
• AWP | Black Box (Battle-Scarred) — $2.50 → $4.19 (1.68x, $1.69) · UUSKINS
• StatTrak™ Glock-18 | Shinobu (Well-Worn) — $6.62 → $8.25 (1.25x, $1.63) · LIS-SKINS

Dollar P&L is not the same as a pretty coefficient. Sand Dune did 3.33x on a $0.042 ticket and paid $0.10. Spiral Glitch did 1.22x and paid $3.23. Black Box beat the estimate (1.30x on the screen, 1.68x in the book). Silent Shot matched 1.43x to the cent.

Buff Market book

Three tickets from one $20 deposit: Syrex, Glock Shinobu, StatTrak Shinobu. All three closed green.

What did not

• StatTrak™ USP-S | Jawbreaker (Well-Worn) — $4.89 → $3.41 (0.70x, -$1.48) · Steam
• M4A1-S | Blood Tiger (Minimal Wear) — $5.41 → $5.21 (0.96x, -$0.20) · Unknown
• Sticker | Rainbow Route (Holo) — $0.81 → $0.62 (0.77x, -$0.19) · UUSKINS

Hide the red rows and this is an ad. Jawbreaker was the expensive lesson: bought on Steam at $4.89 against a $7.39 estimate, sold at $3.41. Steam as a buy rail is the only venue in this sample that finished net red — and that one ticket ate the rest of the Steam book. Blood Tiger’s buy venue never made it into the log (???). The Holo sticker is in the UUSKINS order screenshot above — a skin estimate does not transfer to stickers.

Estimate vs fact

This is the finding I would keep even if the +12% had gone the other way.

• Median estimate 1.30x vs median fact 1.15x.
• Typical sale landed at about 89% of the estimated dollar exit.
• The calculator’s dollar profit summed to $25.73. Realized profit was $10.80 (42% of that picture).
• 14 of 17 estimates sat above fact. Two trades beat the estimate. One matched.

If a deal is only alive at a perfect fill, skip it. Use the estimate as a ceiling. Live prices still have to be checked on the skin page and the destination market — estimates move, liquidity dies, and a 7-day lock is long enough for the ask you modelled to vanish.

What I would not repeat

1. Sub-$1 tickets for the screenshot multiple. Sand Dune and Nitro look loud in a coefficient column and do not pay rent.
2. Buying the Steam rail because the estimate looks fat. Jawbreaker is the receipt.
3. Stickers with a skin-style estimate. Rainbow Route (Holo) was the cheapest way to learn that.

Off-Steam rails in this sample — UUSKINS, Skinport, Buff, LIS — were where the actual spread lived. That is not a ranking of marketplaces. It is where the cheaper ask happened to be when I clicked buy.