A forecaster you can audit
Every prediction the app makes can be traced back through three questions: what did it study, what did it notice, and how sure is it? Here are the answers, in order.
1 — What it studies: a 300-round window
Think of the window as the app's working memory: the 300 most recent results from your room, sliding forward each round, with the newest results weighted heaviest. Why 300? Because the interesting habits live at different timescales — a colour run plays out over minutes, but a digit's nap-and-return cycle can span hundreds of rounds. Fifty rounds can't see the slow habits; ten thousand would react to them too late. Testing settled on 300 as the balance point: long memory, quick reflexes.
2 — What it notices: six habit families
Colour runs
How long stacks get, and how runs of each length historically ended.
Swings
Strict alternation — and how many flips typically come in a batch.
Number naps
Each digit's absence versus its own average absence.
Favourites
Digits appearing above their expected share, and for how long.
Violet marks
The specific run-breaks after which violet tends to appear.
Reprises
Short sequence shapes that echo earlier stretches of the window.
You've watched every one of these with your own eyes — the app simply finishes the counting before the timer does.
3 — How sure it is: the confidence level
When a round opens, each family leans a direction. The prediction is the weighted lean; the confidence level measures how much the families agreed. Strong agreement — confident forecast. Families split — the app tells you the round is thin and lets you skip it. Right after the game's rhythm shifts, it holds off a round or two rather than guess through the fog.
That last behaviour is the part most tools fake and we won't: admitting uncertainty. A forecaster that only ever says "sunny" isn't a forecaster, it's a poster. The panel shows the reasoning under every prediction, and your session log keeps score of how confidence levels actually performed.
The foundation, and its limit
None of this would work if Wingo rounds were coin flips with no memory. Watched at scale, they're not — the game distributes results with habits, and habits can be counted. The same fact sets the limit: habits drift. When they do, confidence drops, and the smart move is the one the app is already making — wait it out. Over a session, that patience is worth more than any single lucky call.
