Should a daily puzzle tell you how close you were, or only whether you were right?

Every daily puzzle makes one design decision before anything else: what happens after a wrong guess.
"Wrong" is one bit of information. "Warm" is a lesson. "Here's exactly how far" might be a spoiler.

I keep going back and forth on this, so I want to lay out the options and hear how you'd choose.

The three feedback models

Strip away the art and the share button, and most daily puzzles use one of three models.

1. Binary: right or wrong.
You guess, you get a yes or no. Classic trivia. Clean, fast, and brutal. A wrong answer teaches you only that one option is eliminated.

2. Proximity: how close you were.
Guess a city and get the distance to the target. Guess a number and get "higher" or "lower". The colored-tile word games live here too: each tile tells you something about how close the guess was.

3. Explanatory: why you were wrong.
After a miss you get a reason: "Close, but that's the capital of the neighbouring country." This is closer to a tutor than a game.

What the research on feedback says (briefly)

This isn't a new question. Education researchers have argued about feedback for decades.

• Hattie and Timperley's review The Power of Feedback separates feedback about the task, the process, and the person, and argues the task and process kinds are the ones that help.
• Kluger and DeNisi's meta-analysis is the uncomfortable one: in a large share of studies, feedback interventions made performance worse. Feedback is not automatically good.

Translated to puzzles: "you're bad at this" (person) is useless. "You're 400 km too far west" (task) is useful. "You keep confusing neighbours, try borders first" (process) is the most useful and the hardest to build.

The trade-offs nobody puts on the landing page

Proximity can turn thinking into searching

Give "hotter/colder" with unlimited guesses and the optimal strategy is binary search, not knowledge. The puzzle becomes an algorithm. That's fun once. It's not why anyone comes back daily.

Mitigations: limit guesses, make proximity coarse ("same continent" rather than exact kilometres), or reveal proximity only after a guess is committed.

Binary is honest but lonely

Binary feedback respects the puzzle. Nobody can brute force it. But a player who misses three times learns almost nothing and feels stupid, which is exactly the person-level feedback the research warns about.

Explanations are expensive and can spoil

Writing a good explanation for every plausible wrong answer is a content problem, not a code problem. And an explanation can easily give away the answer ("that's the capital of the neighbour" just told you to check the neighbour).

Shareability pulls in a different direction

The most viral daily games share a spoiler-free grid of results. Proximity maps nicely onto a grid. Explanations don't. If growth depends on sharing, that quietly pushes you toward model 2.

A rubric I'd use

Instead of picking a model by taste, I'd score each candidate on five questions:

The rubric doesn't give you an answer. It tells you which failure you're choosing.

Where I've landed (for now)

My current bias: coarse proximity during play, explanation after the puzzle ends.

• During play, feedback should narrow the space without solving it for you.
• After the round, when spoilers no longer matter, a short explanation turns the miss into a lesson.

That keeps the daily tension and still respects the wrong answer. But I can see a strong argument that any proximity at all cheapens the puzzle, and I'd like to hear it.

A tiny example

Here's a deliberately coarse proximity function for a capitals game. It groups distances into bands so the player gets direction without precision:

function band(km) { if (km === 0) return "correct"; if (km < 500) return "very close"; if (km < 2000) return "same region"; if (km < 6000) return "same continent, probably"; return "far"; } 

Five labels instead of an exact number. Tuning those thresholds is where the actual game design happens.

What a player takes away from a miss

Think about the last daily puzzle you failed. What do you remember? Usually not the answer itself, but the moment you saw it and thought "of course" or "no way".

That moment is the whole value of the wrong guess, and each model handles it differently:

• Binary gives you the answer at the end with no story about your path to it.
• Proximity gives you a path (cold, warm, warmer) but no reason.
• Explanatory gives you a reason, which is what turns "of course" into something you'll recognise next time.

If the puzzle exists purely for fun, the path might be enough. If you want players to get better at the underlying skill, geography, vocabulary, anything, the reason is the part that carries over to tomorrow.

There's also a fairness angle. Players compare results. If feedback is too generous, everyone solves it and the comparison means nothing. If it's too stingy, the results reflect luck. The best feedback model makes the shared result feel earned.

Takeaways

1. Decide the feedback model first. Everything else (difficulty, sharing, content cost) follows from it.
2. Feedback about the task and the process helps; feedback about the person doesn't.
3. Fine-grained proximity invites brute force. Coarse bands keep thinking in the loop.
4. Explanations teach the most but cost the most and spoil the easiest. Save them for after the round.
5. Sharing mechanics quietly push you toward proximity. Notice that pressure.

Your turn

If you were designing a daily puzzle from scratch tomorrow: binary, warm/cold, or full explanation? And which daily game do you think got this exactly right?

I write about practice, feedback and the AI-readable web while building NEXO, a daily learning game.