How the forecast works

Four outcomes of one forecast

• Low: a cautious outcome. Most simulated futures end above it.
• Expected: the middle outcome. Half of the simulated futures end above it and half below.
• High: a strong outcome that few simulated futures beat.
• Viral: a breakout outcome that only the strongest futures reach.

All four come from the same set of simulated futures, so they never cross: on every day Low is below Expected, Expected below High and High below Viral.

What it learns from

• The game's units sold on each day since launch, updated with every sales report.
• The game's own weekday shape: which days of the week sell more.
• The launch discount, if there was one, and the quieter days after it ends.
• Planned discounts you add and Steam's seasonal sale windows. A deeper discount, or one inside a Steam sale, lifts more, and a quieter week follows the window.
• The game's price and your payout settings, which turn units into net to studio.

Before launch, the wishlists at launch and a first-week conversion range give the starting point. The range is wide until real sales arrive, and then the game's own days take over.

How it is computed

The forecast is fitted to the game's own sales days and simulates many possible futures from them. Low, Expected, High and Viral are read from those futures for every single day and for the running total.

A planned discount starts from the typical lift for its depth and for whether it falls inside a Steam sale. Once the discount has run, the game's own result counts instead.

Adding a discount or changing a price changes only what that change does; the rest of the forecast does not jump.

How it gets better

Each run is compared with the days that followed, and the range is tuned by how far earlier forecasts missed. The more runs a game has, the more its range reflects its own track record.

Past forecasts are kept

Every run is stored as a new record and never overwritten, so the question what did the forecast say last week always has an answer. On the chart, each closed day shows, faintly, the range and Expected the forecast gave the day before, next to what actually sold.

What the Forecast view shows

• Net to studio from launch to 31 December as one total, split into what is already booked and what is still to come, with its Low–High range.
• Next 30 days, Next 90 days and Rest of year 1 beside it.
• Actual against Expected for each closed day.
• What moved the forecast since the previous run.
• Planned discounts, editable in place; the forecast refreshes by itself within seconds of a change.

The forecast is part of Indie, Studio, and Publisher, because it needs the game's own sales.

What it is not

A forecast is an estimate built from the game's past days, not a guarantee. It does not know about an event, such as a Steam feature or a streamer, until the event shows up in sales or in your calendar.