Model accuracy
Every night the site checks the best crafts — 200 regular and 50 knife ones — against live Steam market lots: can the contract be put together now, and will it stay in profit.
A bar is the share confirmed on the night of the check; a blue line at the base is 0%; a grey mark is a night without a share (nothing to count), not a zero.
| Regular | Knives | All | |
|---|---|---|---|
| Checked against the market | 198 | 50 | 248 |
| Excluded: “profit on a price without deals” | 0 | 0 | 0 |
| In the red by the calculation: checked, not in the share | 0 | 6 | 6 |
| Counted | 198 | 44 | 242 |
| Enough lots for a contract | 183 | 35 | 218 |
| Profit on live lots above zero | 0 | 0 | 0 |
| Confirmed | 0 (0%) | 0 (0%) | 0 (0%) |
| Profit on lots minus on average prices, median | -15 010 ₽ | -110 768 ₽ | -36 017 ₽ |
| Same, mean | -19 523 ₽ | -118 008 ₽ | -57 956 ₽ |
| Together with “no deals”: confirmed | 0 of 198 (0%) | 0 of 44 (0%) | 0 of 242 (0%) |
| Check, MSK | Checked | Counted | Confirmed | Share |
|---|---|---|---|---|
| 06.10 | 248 | 242 | 0 | 0% |
| 05.10 | 249 | 245 | 0 | 0% |
| 04.10 | 248 | 236 | 0 | 0% |
| 03.10 | 212 | 201 | 0 | 0% |
| 02.10 | 249 | 58 | 0 | 0% |
| 01.10 | 250 | 13 | 0 | 0% |
| 30.09 | 249 | 17 | 0 | 0% |
| 30.09 | 244 | 9 | 0 | 0% |
| 29.09 | 247 | 6 | 0 | 0% |
| 28.09 | 249 | 11 | 0 | 0% |
| 28.09 | 246 | 188 | 0 | 0% |
| 27.09 | 230 | 164 | 0 | 0% |
| 27.09 | 220 | 178 | 1 | 1% |
A craft is confirmed if, on the night of the check, three conditions hold at once:
- enough lots for a contract — the market has the required number of inputs;
- the promised float is reached — these lots make a contract with the float the calculation promised;
- profit on live lots above zero — at these lots' prices, after the Steam fee.
Only crafts in profit by the calculation count toward the share: it answers whether the promised profit comes true. Crafts in the red by the calculation promise nothing — we check them too, but show them on a separate row.
Crafts marked “profit on a price without deals” are left out of the share: their profit rests on a price nobody has bought at, and live input lots cannot check that. How many would be confirmed together with them — in the last row of the table.
Target: the launch plan — the nightly check confirms at least 50% of the top crafts.
The market listings change every hour: lots are bought and new ones are listed. This is a night-time snapshot — during the day the same craft may come together better or worse. Look at a run of nights, not at one.
When few crafts are counted, one craft moves the share a lot — that is why the percentage always comes with “how many of how many”.