Where the three part company

Stack the three addresses one above another and read down each column. At every position there are three characters, and exactly three outcomes are possible: all three the same, exactly two the same, or all three different. Count each outcome across every position and you have the bar tile.

The first tile picks out the earliest position where the three are not all identical: position 6. Everything before it is the shared opening covered on the previous panel. Everything after it is where the addresses actually differ from each other.

The count of positions where all three agree is 6, which is more than the length of the shared opening. That means agreement is not confined to the front. There is at least one position further along where all three happen to carry the same symbol.

One of those is structural rather than coincidental. The final character of an address in this format is not free: it encodes the version of the format, and every address of this generation ends the same way. That is covered on the panel about the last character, and it is worth knowing because it is a place where sameness is guaranteed rather than lucky.

The remaining agreements, if any, are coincidence. With three strings, 56 positions each and only 32 symbols available, the occasional accidental match is expected. Reading meaning into it would be exactly the error this board keeps warning about.

If you are comparing an address in front of you against one you trust, the positions before divergence are doing no work. They will match whichever of the three you are holding, and they would also match a string built to look like one of them.

The comparison only starts earning its keep at position 6. That is a useful thing to know, because it means a check that reads only the first few characters is checking the part of the string that is guaranteed to look right.

It does not follow that you should start reading at position 6 and skip the front. The front is cheap to read and a mismatch there is still a mismatch. It follows that the front cannot be the whole check.

The gauge shows the proportion of positions where all three published addresses differ from one another. It reads high, which is what you would expect from three independently generated keys.

A low reading would be the interesting result. Three addresses agreeing at many positions would suggest they were not independently generated, or that they share more structure than the format requires. That is not what these show, and this board is not going to dress up an ordinary result as a finding.

All of this compares the three published strings against each other. It has nothing to compare them against from outside, because this site holds no other copy of any address and never fetches one.

So the divergence point tells you how the three relate. It cannot tell you whether all three, or any of them, are the strings the market actually publishes. That question needs a second source, and having a second source is a completely different kind of site from this one.

If you ever want to do this yourself, the trick is to stop reading the strings as strings. Put them one under the other in a fixed width font, cover everything except one column with two fingers, and move down the page rather than along it.

Reading down a column is slow and dull and it is the only manual method that does not skip. Reading along a line invites your eye to jump, and jumping is what produces the confident feeling of having checked something you did not look at.

For a comparison against a source you trust, two columns are enough: yours and theirs. The three way version on this panel exists because the board holds three strings and comparing all of them is the fullest thing it can do with what it has.