What happened before your crash?

Any two columns look a bit related by luck, and the more pairs you check the more luck you find. So it does three things most tools do not.

Compares ranks, not values. One extreme day cannot invent a pattern.

Shuffles your own days in blocks, two thousand times, and asks how often chance beats what it found. Blocks rather than single days, because good and bad spells come in runs.

Corrects for how many pairings it checked — using the strict correction, the one that assumes the checks are tangled together, which these are.

Tested on made-up data with no pattern in it at all, at four different levels of "good weeks and bad weeks come in runs". The ordinary correction invented a finding in 7 to 17 runs out of 100. The strict one invents one in 2 to 5. That is why it uses the strict one, and why it will sometimes stay quiet about something that is really there.

Code: /js/what-came-before.js · its own checks: ?selftest=1

It cannot tell you what caused anything. Two things moving together is not one causing the other.

It cannot diagnose or treat. It is not a medical device and not a doctor.

It needs enough days. Under about three weeks of overlap, almost nothing clears the bar. That is the maths being honest, not the tool failing.

It only sees what you wrote down. If the thing that matters was never a column, it cannot be found.

No. It cannot tell you what caused anything. Two things moving together is not one causing the other. What it does is narrow the list. It checks whether anything you wrote down lines up with your worse days at a delay of one, two or three days, and it says plainly when nothing clears the bar.

Often because the two days were never lined up. If the hard day was Monday and the bad day was Wednesday, comparing Wednesday with Wednesday lines up the wrong pair. Some trackers can look at the days after a factor — Bearable's report goes out to seven days — but that is a different question from whether what you are seeing is more than chance. This page checks the day before and the two before that, as well as the same day, and tests each one.

It was built for that shape of problem: a bad day that follows the hard day by one to three days — post-exertional malaise, or PEM, in the words most people here use. That gap is easy to miss by eye and just as easy to see where it is not there, which is usually what makes pacing so hard to get right. It is not a medical device. It does not diagnose. It makes no claim about any condition. It reports associations in numbers you already recorded.

Yes. Bearable exports one row per entry rather than one row per day, and its dates look like 22nd Jan 2022. Both are handled. The page turns it into one row per day, keeps the item names you chose in the app, and tells you on screen that it reshaped your file. Any CSV works: in Google Sheets or Excel it is File, then Download or Save As, then CSV.

No. Your file is read by your own browser. There is no server, no account, no email box. Close the tab and this page has nothing left of it. You can check it yourself: load the page, turn your wi-fi off, then run it. It still works. The honest limit is that reloading the page with wi-fi off will fail, because the page has to be fetched once. What never travels is your file.

About twenty days at a minimum, and it works properly from about three months. Below twenty it will tell you so and name the number, rather than pretending.