Steam Launch Revenue Estimator — Calibrated Cone vs Rule of Thumb
Last updated: 2026-05-06 · Reading time: 6 min
A Steam launch revenue estimator that gives you a single number is wrong half the time. The estimators worth using output a probability cone — P10 (downside), P50 (median), P90 (upside) — so you can plan budget against the floor, not the median.
Run a free calibrated forecast right now
Enter your Steam app ID + current wishlist count. Get a P10/P50/P90 revenue cone with comp-set evidence, Boxleiter cross-check, and what-if levers for marketing decisions.
Free Steam revenue estimator →Why a single-number estimator is the wrong tool
Most online "Steam revenue estimators" do roughly this:
revenue = wishlists × $5
That's the Boxleiter formula — a useful sanity check but a terrible budget plan. Two problems with single-number outputs:
- No risk envelope. An indie launch’s lifetime revenue distribution is heavy-tailed. Your $5 × wishlists median has roughly equal probability of being too high or too low — but the size of "too low" can sink you, while "too high" is upside you don’t need to plan for. You need the P10 floor to size your runway.
- No nearest-neighbor calibration. A flat $5 multiplier averages across the whole indie corpus. Your specific genre + price + wishlist trajectory has a tighter band — or, for novel genres, a much wider one. A flat multiplier hides this.
What "calibrated" means and why it matters
A revenue cone is calibrated if its published interval contains the true outcome with frequency at-least 1−α over a reference distribution — on average, not for every individual game. For an 80% calibrated cone, 80% of games covered fall inside the P10-P90 band. The remaining 20% fall outside — which is the honest answer when your game is unlike anything in the calibration corpus.
Steam Launch Forecaster validates calibration with leave-one-out cross-validation on the 77K-app corpus + a held-out test set of recent launches. Empirical coverage is published on the methodology page, including where the model under-performs (mega-hits, novel-genre breakouts).
Inputs the calibrated estimator uses
| Input | Why |
|---|---|
| Wishlist count at forecast time | Strongest single signal; logarithmic relationship with revenue |
| Wishlist trajectory shape | Linear-growth wishlists convert better than spike-driven; informs cone width |
| Genre + tag overlap | Determines the comp-set; tighter clusters → tighter cones |
| Price point | $10 vs $25 vs $40 dramatically changes per-unit revenue and conversion shape |
| Days to launch | Wishlist trajectory has time to keep climbing or stall |
| Whether you’ve uploaded Steamworks data (paid) | Total Lift Attribution recovers ~75% of campaign wishlists Steam under-reports — tightens the cone meaningfully |
How the cone narrows or widens
The cone is wider for games where the model has less to anchor against. Specifically:
- Tight cone — your game has 5+ strong nearest-neighbors in the comp set, your wishlist trajectory looks like theirs, and your genre is well-represented. Treat the cone as an actual budget plan.
- Wide cone — novel genre, atypical wishlist trajectory, sparse comp set. The cone is honest about its uncertainty — treat it as directional, not load-bearing.
- Cone fails — mega-hits and breakthrough genres where the corpus has zero precedent. The model says so explicitly with a divergence flag, and you should not trust the upper bound.
Free vs paid estimator features
The free single-game forecast uses the same calibrated cone math as the paid $299 launch report. The differentiation:
| Feature | Free | $299 launch report |
|---|---|---|
| Calibrated revenue cone (P10/P50/P90) | ✅ | ✅ |
| Boxleiter cross-check | ✅ | ✅ |
| 5 nearest-neighbor comp launches with revenue | ✅ | ✅ |
| Marketing-lever causal estimates | limited preview | ✅ full |
| Total Lift Attribution (recover ~75% under-reported wishlists) | — | ✅ |
| Re-runnable through your launch window | per-session | ✅ tracked |
For most pre-launch budget planning, the free forecast is sufficient. The $299 report unlocks when you start running paid campaigns and need true cost-per-wishlist tracking.
Run the free Steam revenue estimator
Enter app ID + wishlist count → P10/P50/P90 cone with comp-set evidence and Boxleiter cross-check.
Free forecast →Need the full launch report with Total Lift Attribution? $299 single launch report →
Frequently asked questions
What’s the most accurate Steam launch revenue estimator?
An estimator that outputs a probability cone (P10/P50/P90) rather than a single number. A single-number estimate is right ~50% of the time and badly wrong the other 50%; the cone tells you the spread so you can plan budget against the P10 floor. Run a free calibrated forecast →
How do I estimate revenue from my Steam wishlist count?
The Boxleiter rule of thumb (revenue ≈ wishlists × $5) is a starting point but breaks down on novel genres, mega-hits, and ad-heavy wishlist mixes. A calibrated estimator integrates wishlist count, trajectory shape, genre, price, and comp-set evidence to produce a tighter cone. Boxleiter explainer →
Is a free Steam revenue estimator accurate?
The free single-game forecast uses the same calibrated cone math as the paid $299 launch report. Only marketing-lever causal estimates, comp-set explainer details, and Total Lift Attribution are paid features. The base cone is identical.
Should I use the median (P50) for budget planning?
No. Plan against the P10 floor. The P50 is the median — you have a ~50% chance of landing below it. The P10 is the level above which 90% of similar games land. Sizing runway against P10 is the conservative engineering call.
When does the calibrated estimator fail?
On mega-hits (1M+ wishlists where the multiplier compresses) and novel genres where the corpus has no precedent (Vampire-Survivors-class breakouts). The model tags these with a divergence flag — treat the upper bound as untrustworthy when you see it. Most indie launches with 5K-200K wishlists in established genres get reliable cones.
Built by Greg C. — senior software engineer with production ML experience in calibrated prediction. Steam Launch Forecaster trains a CQR-calibrated model on a 77K-app Steam corpus. See the methodology →