Each question is a situation the tool can actually carry out. You lock in a percentage first, and only then does it run that situation a thousand times with secure browser randomness: dealing 23 birthdays, tossing until five heads appear, picking doors, counting runs, chasing six faces. The bar you watch is the share of those thousand rounds where the thing happened. Because it is measured rather than looked up, the number lands slightly differently every time, and that is the point.
Someone tells you 23 people are enough for a shared birthday and you do not believe it. A friend insists the coin is due for tails. An argument about whether switching doors really helps has gone on too long. Open this, both of you lock in a number, and let a thousand rounds settle it in about a minute. It also works alone, as a quick check on how well your sense of chance actually tracks.
It only measures these five fixed situations: you cannot enter your own question, and the result says nothing about how clever or lucky you are or what will happen next. A thousand runs is a sample, not a proof, so the measured share moves by a point or two each time. Your bets and samples are not saved or uploaded and no account is needed; this device only keeps one number, your best average gap. It gives no advice about money, betting or real decisions.
Common questions and answers about this topic.
Press Start the experiment, read the situation on the machine’s screen, drag the yellow needle (or use − / + and the quick picks) to the percentage you believe and press Lock in my answer. The machine runs the situation a thousand times; its screen shows each run’s sample and the green needle settles on the measured share next to your yellow one. Five questions make a round; at the end you get your average gap, a title and why the widest miss fools people.
Run the first question and see for yourself. The measured share lands near half, because the question is about any two people rather than someone matching you specifically, and 23 people already make 253 possible pairs.
No, and the second question measures exactly that. The tool tosses until five heads appear and then tosses once more, a thousand times over, and that final toss stays close to an even split. The coin has no memory of the run before it.
Because it is measured, not looked up. A thousand rounds is a large sample but still a sample, so the share typically moves by a point or two between plays. That wobble is honest: it is what real sampling looks like.
No. It only sums up how far your five bets landed from the five measured shares this round. It is not an intelligence, maths or luck test. The only thing kept between rounds is your best average gap on this device.
Your bets and results only live on this open page; there is no account, no history and nothing is sent anywhere. This device keeps one number, your best average gap, so you can try to beat it. Refreshing starts again from question one.