Show HN: Fermi – A Wordle-style game for order-of-magnitude thinking

Why do I have to pick the factors first? Do I have to use all of them? Why shouldn't I use all of them? If I do, why aren't they already picked?

Why are all of these sliders? Why don't they start at zero? Why do all of them at the default setting result in me winning?

Speaking of, am I winning? Failure and success give me the same feedback. I can play this game over and over until I win? That's not really in the Wordle style.

imo you should be picking from a palette of maybe-relevant factors and increasing/decreasing their order of magnitude and order/operation, then when you submit you're locked into a win/loss state like wordle. This would be much more of a game than what you've got here.

Some feedback:

1. Why are factors reorderable? axb = bxa

2. Why are factors pickable? Is it coincidence that all of todays choices are useful (or did I mess it up)

3. The ranges/sliders ruin it for me IMO, they pick the order of magnitude for you

I echo what the others say about the current interface.

I think there is an opportunity for "normal" and "hard mode" like Wordle. In normal you are given the names of the factors but have to enter the expressions for their value (no default values given). In hard you have to come up with your own factor names as well as the values (the name is purely aesthetic for the user to keep track of their ideas). The win condition is the same (be within an order of magnitude of the answer) but maybe there is a special "super" win for being within 10% (log scaled, so lower bound n^0.9 and upper bound n^1.1).

My honest opinion is I have no clue what's going on. To some extent it is because traditionally you wouldn't consider this a game like Worlde, other than the fact you presumably release a new setup for each day.

To some extent, the idea could be improved by always including all factors by default and only allowing sliders to be changed. Or allow me to use less factors.

The biggest issue is I have no clue what happened when I clicked guess. Going back I see the reference estimate, but frankly its unclear what that even means (again average person is uninformed). I want something to clearly tell me how close I got, and maybe what percentile I'm in for users who have played today. It's simply not addicting in the way that the other recent Worlde like games are meant to be.

Other nitpicks:

The example popup does not help me understand anything

Its not apparent why the sliders have an initial value nor what the significance is

I am not sure I entirely understand the game. In typical Fermi questions, the goal is to arrive at the appropriate order of magnitude. In this game, the order of magnitude appears to be provided for you. The extrema are ~220k and ~1.8M for the question regarding passengers in the air, and these extrema are less than an order of magnitude apart. Am I misunderstanding something...?

The idea has potential, but needs polishing. When I read the tutorial, I thought that I had to guess the target number using completely unrelated values. Was a bit disappointed to see the "options" were obvious and not really optional, and the scales too limited. Seemed too easy. It would be more entertaining to guess the number of people currently in flight based on x number of full football stadiums, the avg number of eggs laid by y hens per month, etc.

A visual queue of reaching within the success range would do a lot too. Was a bit confused whether I was right or not after submitting the answer.

I like the concept.

However the UI is a little confusing (on mobile anyway).

I might suggest you don't present the pre-set range values, as it makes it a lot easier.

Have you considered - perhaps in a way to gamify it a bit more - giving a first hint, and hiding the next 3 unless the player asks for them? It would add an element of 'getting it in 1 go' etc. I'd imagine one or two-shot winners are also more likely to share their results. More potential for your app to go viral.

I'm on my phone, trying this. It's a question about how many people are in the air right now.

- when I tap on a factor, I can edit the slider, but it doesn't do anything.

- when I tap another factor, the previous editor stays in place.

- how do I choose my factor?

- where do I actually do the multiplication?

- where do I check my answer?

- why do the factors have "grab icons" (the six dots)? Why would I drag one?

https://imgur.com/a/QSRBbYR

Etc.

Feedback: The units on square feet per acre are inverted. Should be acres per square foot.

I recommend having the solution check unit cancellation to confirm that it works.

I'd also like to see a choice between multiplying/dividing in the equation.

Neat idea, reminds me of this clip[1] from a Twitch streamer guessing the Costco price of sliced cheese.

[1] https://www.youtube.com/watch?v=LKoYrpFn-Ls

If I were doing something vaguely similar, I might emphasize:

- Bounds rather than point estimates. Hard bounds (definitely more/less than this) and soft bounds (likely). With associated payoffs/penalties. A point estimate too, but bounds first. Like error bars in a plot.

Perhaps even "bet your certainty" custom hardness. Perhaps an advanced "draw your payoff curve", so one might draw a hard step-function low bound with some soft slope on the high bound. So one might draw a big-payoff pointy spike, at the risk of missing entirely. A pretty UI might even multiply the curves for you. Perhaps one gets points for each component estimate, not just the final result - though might need an LLM to cope with the variety of possible steps in reasoning.

- Order of magnitude. That's "is it more like 1/10, 1, 10, 100, 1000" etc. One can do "high/middle/low order 10^n". But oom estimation a different mindset than linear. Exponential not linear. Linear slider as UX smell.

- Permit multiple chains of reasoning. An argument that nicely establishes a hard lower bound may be of no help in firming up a high bound. And a bound may need to be hammered with multiple chains of reasoning to tighten it up.

Fermi estimation can be the aggregation of exponential-rough-quantitative reasoning with attention to confidence/uncertainty.

In a group setting, one might have: "Ok everyone, how many cats live on this block? Is it more like 1, 10, 100, 1000, etc? Can anyone suggest a bound?" "I saw one in the window this morning - so hard lower bound of 1." "Unless it's been taken to a vet." "The town human population is 10^6, so it seems unlikely the block has that many cats..." "What does everyone think?" "How confident?" "Can anyone suggest another chain of reasoning? Can we tighten up the bounds?" "Could we do a stepping stone problem, the number of households on the block?"

Perhaps a "bet coins" on the answer before grading? One advantage of bounds vs more-common point-estimates is it pulls up the handling of confidence where it can be seen and improved. So anything which encourages reflection and self-evaluation, like betting, might help with that. Another advantage of bounding is it makes each attempt at narrowing bounds, on each component, into it's own little fractal estimation problem. So an initial estimation question proliferates many others. I wonder how a UI might emphasize that...?

But all that riffs on the Fermi estimation aspect, not the Wordle aspect.

A kid/younger audience version of this would be great!

It looks very nice visually, but I think the input sliders are way too narrow. Point in fact, in this example they have roughly a 2:1 range, which means virtually any combination of inputs 'wins' the game (is within 1 order of magnitude). That's not particularly fun, is it.

Personally, I approach Fermi estimates as order of magnitude guesses for each factor (e.g. if guessing how many ping pong balls fit in a 747, you might say, "well the width of the jet is closer to 10ft than 100ft", rather than guess 22ft), so having such fine-grained sliders feels like it misses the point.

A few thoughts, I tried this on mobile but got a bit confused by the initial "How to play" dialog (besides that it was too large to fit on screen). The dialog says:

  How to play. String together factors to answer the question.

For the question I got ("How many people are flying in airplanes right now?"), the influencing four factors were nicely chosen, although I would refrain from guiding the player too much with (too) narrow "x - y" bands. We're looking for orders of magnitude, so you could think about sliders that also suggest thinking in orders of magnitude, say "1, 10, 100, 1000" (e.g. number of passengers on a plane), and "10 minutes, 30 minutes, 1 hours, 2 hours, 8 hours" (e.g. flight duration).

(Another angle here could be: how many flights and/or passengers does the nearest airport handle during a year, or how many flights/passengers do the global top 10 airports handle during a year, and how does that account for the total amount of flights in a year?). In other words: I don't think you want your players to do exact guesses using these rather precise sliders and narrow bandwidth, but hint them into also thinking in orders of magnitude.

I would skip the clicking on factors (I saw in another comment that you thought about arbitrarily mixing addition and multiplication, and their order in an earlier version, but that seems too difficult), and just give the player a few sliders and their proposed answer directly. Perhaps you can show the proposed answer directly beneath the question, start with 0, and have all sliders set to 0 (instead of the current random values) as well.

Another idea, perhaps even more intuitive, could be to give the player one (easy), two (intermediate) or three sliders (expert), without giving them hints for individual contributing factors:

- (slider 1) 1, 10, 100, 1000, ... for the rough orders of magnitude (e.g. "100")

- (slider 2) 1-9 multiplier for the chosen order of magnitude to give the player a way to say "it's more in the order of 500 than 100"

- (slider 3) 0-9 multiplier for one order of magnitude lower, to give the player a way to say "it's more 550 than 500"

Then the answer is calulated as (1) * (2) + (3). You can then tell the player if their chosen orders of magnitude were correct ("within the chosen order of magnitude"), slightly off (one order of magnitude too high or low), or too far off. Let the player decide if they want to be rather precise (with three sliders), or not.

Or even combine these ideas! Let the player choose a strategy ("guided" as you've already implemented with a couple of factors already given, or "non-guided", for lack of a better term, for those looking for an extra challenge).

Closing nit: "Share results" contains a non-existing domain.

If you need inspiration for future questions, have a look at this recent discussion [0] and linked website [1].

[0] https://news.ycombinator.com/item?id=43389455

[1] https://taylor.town/napkin-math

[dead]