Fermi Questions (fermiquestions.org) is a Wordle-like game where you try to guess the answer to Fermi estimation questions with 6 or less tries. After each guess, you'll see if your answer was too high or too low. You win if your guess is within ±20% of the correct answer.
Example questions:
- How many new cars were sold in the US in 2024?
- How many humans have ever lived (including those currently alive)?
- How many chickens are slaughtered for meat every year?
The skill of Fermi estimation is also extremely useful if you want to perform better here on Manifold. As Philip Tetlock shows in his book Superforecasting, many of the best forecasters break down complex questions into smaller, more manageable components which is exactly what you can practice when playing the game.
In the first 24 hours after going live it had some 1,100 unique visitors, which primarily came from Hacker News. Resolution source is the unique visitors count on the publicly available analytics dashboard: https://plausible.io/fermiquestions.org
Update 2025-08-04 (PST) (AI summary of creator comment): In addition to the default bot-filtering by the Plausible analytics tool, the creator reserves the right to manually exclude any suspicious activity from the final unique visitor count.
Update 2025-08-04 (PST) (AI summary of creator comment): A unique visitor is defined as a distinct visitor to the website within a 24-hour period, based on their IP address. A user who visits on different days will be counted as a new unique visitor for each day.
1,000
3.00I've greatly appreciated people here sharing screenshots of their game statistics; please continue to do so! This is great fun!
Another clarification that might be important: For the purpose of this market a unique visitor is defined as a distinct visitor to the website within a 24-hour period, based on their IP address. So if the same user returns on a different day, they are counted by the analytics dashboard as a new unique visitor.
I love how my guess distribution looks approximately normal.
I want to take a crack at the game difficulty:
I imagine the intended way to play this is to do a fermi estimation to get within an order of magnitude of answer, then change my guess up or down an order of magnitude based on the higher/lower hint. A correctly done Fermi estimation should put your two guesses as red above and below the true answer in the worst case, and you can guess the lower number * 3.2 to guarantee a win if both are red. If your first guess is yellow, you can win in the next guess, and if your second guess is yellow, you can win on the next turn. Therefore the fair expectation for someone talented at the game to consistently be at or below would be 3 guesses.
I think it is fair to give people one additional guess to account for mistakes , but each guess you give someone past the third pretty much expands the target they have to hit with their initial estimation by an order of magnitude in each direction from the correct guess. I recommend 4 guesses, but if you think my viewpoint is very different from the average person who would play this, I wouldn't complain about 5.
Another idea:
Your questions are all very "global" in that someone from anywhere in the world would have a good chance at knowing enough to estimate the answer. This is a good thing, but it could limit your options. What if there were more specific question (think piano tuners in Chicago) that came with a few pieces of information (Chicago is a city in the midwestern US and is the 3rd most populous city in the US, in the piano tuner example) to allow people from anywhere to have a shot at answering them?
Love it. Others have said a similar thing, but it seems like a shame that it feels like 95% of the game is coming up with the first guess, followed by trivial (but tbh still quite satisfying) binary search.
I think one challenge is that users might be quite bimodal between people who know how to do Fermi estimation (for whom the first guess is where the fun is) and people who don't (who might spend a few guesses just to get the order of magnitude and then might run out of guesses for the final precision).
Suggestion: all users keep going until they reach 6 guesses and then get scored based on how close they are to the final answer. Two downsides: you need more precise source numbers, which might be frustrating if top uses feel the source has false precision, and it makes it less wordle-esque
But ultimately, I think the there's a tension between "wordle-esque and satisfying for mass users" and "really good Fermi estimation game that is satisfying to fans of Fermi estimation". It might be that your current middle road is already optimal!
@Fion The idea of always playing 6 rounds and being scored on proximity is so smart! Kind of reminds me of Geoguessr.
@TimothyJohnson5c16 I’m really conflicted about changing the difficulty based on the feedback I receive here because I think that few audiences perform better in answering Fermi questions than the Manifold/prediction market crowd. But some way to disincentivize doing binary search would probably be good.
Fermi Question of the Day: 4 August 2025
"How many iPhones has Apple ever sold?"
I won using 2 out of 6 guesses. Can you beat me?
🔴🟢
Fermi Question of the Day: 5 August 2025
"How many iPhones has Apple ever sold?"
I won using 3 out of 6 guesses. Can you beat me?
🟡🟡🟢
my first time. We take those. (1st was under, 2nd over)
the one I got wrong was because I didn't look at the arrows closely, so I didn't notice when I went from red ↑ to red ↓ and so I kept trying even higher numbers.
Typing out the whole number feels a bit pointless when only the number of 0s and one or two digits actually matter. Except for the Earth surface question, all my responses were multiples of 2.5 (like 100, 2 500, 50 000, 750 000, etc.)
Another issue with the order of magnitude being way more important than the leading digits is that you will usually settle the former before the latter. But when typing out the number, you write the leading digits first and the zeroes after. If after writing the number you erase the leading digit to replace it, the whole input field resets. This resulted in me having to retype the number several times, because "fix the order of magnitude first, worry about the leading digit later" is such a natural way for me to think about the problem that I kept doing it.
A different input method would be really nice, maybe a slider for the number of digits?
These are just nitpicks that won't stop me from coming back for more, but I thought you would be interested in hearing it anyway.
@Wott Thanks for the extensive feedback! I’ll try to implement an easier input method for large numbers in the near future. I do have to be somewhat careful as to not take up too much space on smaller screens but I guess I’ll manage to do something here.
Something else I noticed in your screenshot is that there is no “Share” button next to the “Close” button. This shouldn’t be the case, would you mind looking at the site if you now can see this additional button? And if not, what device are you using? Is your browser blocking scripts?
@danftz yeah, I'm running ublock origin with cosmetic filters turned on. The Share button shows up if I turn it off.
I tried out the existing questions. I thought it was neat, but I was a lot less interested in the repetitive guess and check portion than just making a really good guess the first time.
I think I'd rather be tracked by the (log?) difference from my first guess to the correct value. Because now when I want to see how I did, it just says I got it right on the first guess X number of times, etc.
@Eliza I would probably come back, but not every day. I'd rather come back once a month and answer 30 of them.
@Eliza I had at least one time, maybe two, where I accidentally typed the wrong number of zeroes, wasting a guess. It would be nice if the input method allowed me to just select this in an easier way.
@Eliza If you enable the game to allow two or more new and related rounds in a single day, you can sometimes demonstrate some very interesting things with these kinds of numbers. There is a lot of creativity available as the curator here and that would be one way to show it and demonstrate that you are having a dialogue with participants.
One example of this is I saw many of your questions specified US, you can often find surprising results extrapolating from US to worldwide. Some figures will be astoundingly small or large by comparison.