A recommendation gets sharper when it knows what to leave out.
Most discovery systems are built around attraction. They ask what you like, what you played, what is similar, what is trending, and what other people clicked after clicking the same thing. That can help, but it misses one of the strongest signals in taste: refusal.
The games you bounce off are not random noise. They are part of your player shape.
Taste is not only what you enjoy
A player can love strategy games and still dislike spreadsheet fatigue. Someone can enjoy cozy games and still bounce off chores. A player can admire a beautifully made roguelike and still not want another run-based loop tonight.
Those dislikes are not contradictions. They are useful boundaries.
When discovery ignores those boundaries, it keeps recommending games that are technically adjacent but emotionally wrong. The result feels familiar: a long list of almost-right options that somehow makes choosing harder.
Fit is not just matching desire. Fit is also respecting friction.
Hard no is a recommendation signal
A clear no can be more useful than a vague yes.
"I want something relaxing" is a start. "I want something relaxing, but not farming chores, not daily timers, and not a game that asks me to optimize every square" is much better. Now the system can stop treating every soft-looking game as a match.
The same is true for players who say:
- no reflex-first combat tonight
- no grind-heavy progression
- no long tutorial before the game opens up
- no survival meters
- no multiplayer pressure
- no story that demands emotional homework
That is not negativity. That is precision.
The second character should remember what makes you leave
Snowbll's second character idea is about continuity. It should remember the patterns you should not have to explain every time you search.
If you often quit games when progress turns into chores, that should matter. If you like management pressure but hate feeling assigned a second job, that should matter. If you like deep systems but only when experimentation is forgiving, that should matter too.
A useful persona is not a trophy shelf of favorite games. It is a living map of fit signals and mismatch risks.
The point is not to trap a player in a type. The point is to stop forgetting what they already learned about themselves.
Developers benefit from mismatch too
This matters for game makers as much as players.
A mismatch is not always a failure of the game. Sometimes it is a failure of routing. The wrong player arrives with the wrong expectation, bounces quickly, and leaves behind weak signal. The right player understands the shape of the experience before they enter.
Fit-based discovery can help developers explain who the game is for and who might bounce off. That is not shrinking the audience. It is making demand clearer.
A game does not need to be for everyone to find the people who will care deeply about it.
Better recommendations should be comfortable saying no
The old model of discovery tries to maximize options. The better model should reduce the wrong ones.
That means a recommendation layer should sometimes say: this is popular, but probably not for you tonight. Or: this matches your genre, but conflicts with your stated hard no. Or: this one is less famous, but closer to the session you described.
That is a better kind of help.
Snowbll should not decide what is fun. It should make the fit logic visible enough that the player can agree, reject, or refine it.
Because a useful recommendation is not just a yes. Sometimes it starts with the right no.
Join the waitlist and help build discovery that respects what you refuse, not just what you want.
Join the waitlist