Checking a calculator you did not write
You cannot look inside somebody else’s calculator, and you do not need to. Four dates with answers worked out in advance will tell you what any of them is doing, because each date isolates one decision. A wrong figure then points at the decision instead of leaving you with a shrug.
Four dates, each testing one thing
3 October 1992 is the control. Whole date: 3 + 1 + 0 + 1 + 9 + 9 + 2 = 25, then 2 + 5 = 7. Separately: day 3, month 1 + 0 = 1, year 1 + 9 + 9 + 2 = 21 and then 2 + 1 = 3; 3 + 1 + 3 = 7. No held-back total appears anywhere, both routes give 7, and every calculator that is working at all returns 7. If this one comes back different, stop — nothing further is worth testing.
5 March 1965 tests the stop rule and nothing else. Whole date: 5 + 3 + 1 + 9 + 6 + 5 = 29, then 2 + 9 = 11. Separately: 5 + 3 + 3 = 11. Both routes reach 11, so the route cannot be the variable here. A calculator that shows 11 keeps master numbers; one that shows 2 reduces them.
8 February 1976 tests the route. Whole date: 8 + 2 + 1 + 9 + 7 + 6 = 33, held back. Separately: day 8, month 2, year 1 + 9 + 7 + 6 = 23 and then 2 + 3 = 5; 8 + 2 + 5 = 15, then 1 + 5 = 6. A 33 means the whole date was added in one go; a 6 means the parts were reduced first.
9 August 1976 runs the same test the other way. Whole date: 9 + 8 + 1 + 9 + 7 + 6 = 40, then 4 + 0 = 4. Separately: 9 + 8 + 5 = 22, left standing. Here the held-back total sits on the separate route, so a calculator that keeps master numbers on one route and quietly reduces them on the other is caught.
Why testing with your own birthday tells you nothing
For most dates the two routes land on the same figure, so most birthdays cannot distinguish between calculators at all. That is not luck: both routes are digit sums, and digit sums leave the remainder on division by nine unchanged, so two ordinary single digits derived from one date are necessarily the same digit. The routes can part only where one of them stops on 11, 22 or 33.
Which is why the four dates above are chosen rather than convenient. Run your own date by all means — but a match across three sites means those three share a convention, not that the figure has been confirmed.
What the tradition offers as a tie-breaker
Nothing that works as one. In this tradition every source states its own rule and then proceeds as though the matter were closed. There is no registry, no adjudicating body, and no test inside the method that could show one reduction route to be the correct one, because both are stipulations about how to add rather than claims that could be checked against anything.
Read as statements about themselves, the sources are accurate: this school adds the whole date and holds back 11, 22 and 33; that one reduces the parts first. Read as statements about the world — your number is 11 — two of them contradict each other on a large minority of dates, and the tradition contains no procedure for deciding between them.
Where the sources do converge is narrower and more useful than a verdict: they all present the calculation as something a reader can repeat. A calculator that withholds the working is at odds with the tradition it is quoting, not only with this page.
Where this page’s calculator parts from the one it links to
The widget below adds the whole date in one string, and holds back 11, 22 and 33 when the running total reaches one of them. The calculator this page sends you to reduces the day, the month and the year separately, and applies the stop rule to the parts as well, so for that one a day of the 11th stays 11 rather than becoming 2.
Across the 35,064 dates from 1930 to 2025 the two produce different figures on 4,492 of them, 12.81 per cent. The full report at the end of that path prints a pair such as 11/2 or 16/7 — the total it held back and what that total reduces to — and on 3,475 dates, 9.91 per cent, the figure from the widget below matches neither half of the pair.
None of the three is broken. They are three stated rules, and the numbers above are what stated rules cost when nobody writes them down. A page arguing that calculators should name their convention has no business hiding its own, so: whole date, master numbers held back, 1930 to 2025 as the window every percentage here was counted over.
What to require before you believe a figure
Ask for the chain. A calculator that shows 25 and then 7 can be checked in seconds; one that shows only 7 is asking to be believed, which is a different transaction.
Ask which route it took, in words. Whole date or parts first is the single largest source of disagreement, and any calculator knows the answer about itself.
Ask what it does with 11, 22 and 33, and whether it holds back anything else. Some implementations keep a longer list, which is why a report can hand you 16/7 for a date this page would call a plain 7.
One thing you never have to ask about is date format. Both routes add the same digits whatever order the day and the month are written in, so 3/10 and 10/3 give the same life path. If two sites differ on your date, the format is not the reason.
Before you trust a result
- Two sites agree with each other. Does that make them right?
- It makes them the same convention. Agreement between calculators that reduce the parts first is expected and says nothing about the calculator that adds the whole date. Test them on 8 February 1976 instead, where the two routes are known to part.
- A calculator gave me a number with no working. Is it wrong?
- Not necessarily, but you cannot tell, which for a result whose only strength is that it can be reproduced amounts to the same thing. Run the four dates above through it and you will know what it does, working or no working.
- Which figure should I quote if two of them disagree?
- Quote the rule alongside the figure: 11 by the whole date, or 2 with the parts reduced first. A number on its own is the thing that cannot be checked, and naming the rule costs four words.