I have been reading webpages about crypto and passwords
until my head swims. I think I have a scheme; I would
really like to see any comments about whether I am being
an idiot.

My purpose is to have hard-to-guess and unique passwords on
all of the different systems in life that ask for passwords.

Here is my plan: I propose going to these webpages to get
somewhat randomish data:

https://www.fourmilab.ch/hotbits/secure_generate.html
https://www.random.org/integers/
https://www.grc.com/passwords.htm

and put together a couple of thousand characters worth
of what I think are hard-to-guess character strings
with different sets of characters (because some systems
limit the kinds of characters you can use).

My sets of data would be:

Set NUM: 0-9 (10 possible characters.)
Set LET: a-z (26 possible characters.)
Set LETNUM: a-z+0-9 (36 possible characters.)
Set UCLCNUM: a-z+A-Z+0-9 (62 possible characters.)
Set PRNASC: Printable ASCII Characters from Character 33
("!") to Character 126 ("~") (93 possible characters.)

I think I can creat some of the above from others by, for
example, getting a new big a-z+0-9 set and throwing away
everything that isn't a number.

I plan to store the randomish data in a text file encrypted
with Axecrypt. Whenever I need a password. I will cut some
characters out of the data, always using as many characters
as the site or program will allow, and always using the
maximum length password that the site or program will allow.

I will then store the list of which passwords go with which
sites or programs in another file, again encrypted with
axecrypt, and back it up to 5 different places once a week
or so.

Am I right in thinking that this is the most secure way of
dealing with multiple passwords?

BTW. is there a formula that lets me plug in the length and
the number of possible characters and tells me how big a brute
force search would be?