LZMA filtering documentation very poor

I'm trying to optimise compressing some documents, so I'm using LZMA's filters. These are really badly documented: https://docs.python.org/3/library/lzma.html#filter-chain-specs

General problems:

• The filters are using short-hand for variable names lc, lp, mf, which is really bad practice for a publically available interface (and even for internal ones!). (ok, so this isn't a doc problem, but a library one).
• It's not at all clear what any of the variables do. There's no links to any documentation (even third-party) which explain what any of these things do.
• It's not clear what values are acceptable for some of them.
• • preset is documented elsewhere, and I assume the value range is the same?
• • lc doesn't say the maximum is 4, you have to read lp to see the combined maximum is 4, and then deduce.
• • lc, lp, pc, nice_len, don't state minimums
• No defaults stated other than depth (0),
• It's not at all clear what the different Filters do.
• The limited blurb that does exist about Filter chains seems to assume a high degree of expertise in these things. For example: The last filter in the chain must be a compression filter, and any other filters must be delta or BCJ filters. - that has cleared nothing up.
• What's a MODE_FAST or MODE_NORMAL do?
• What's the difference between MF_HC3, MF_HC4, MF_BT2, MF_BT3, or MF_BT4?
• What does lzma.PRESET_EXTREME do? There's no documentation for lzma.PRESET_DEFAULT at all (which my IDE exposed to me via autocomplete).

Etc etc.

Some specific problems as well:

• 'filters': [{'id': 33, 'preset': 6, 'dict_size': 262144, 'lc': 1, 'lp': 1, 'mode': 2, 'nice_len': 1, 'mf': 3}]
  That creates an LZMAError. It's only after generating thousands of combinations of these that I notice a pattern: nice_len: 1. It seems there's a minimum nice_len size. This is not documented.

• 'filters': [{'id': 33, 'preset': 2147483653, 'dict_size': 1073741824, 'lc': 1, 'lp': 0, 'mode': 2, 'nice_len': 201, 'mf': 20}]
  Generates a MemoryError (as do lots of others of my tests). No idea why as I have plenty of RAM. I presume some combination of those settings are causing an issue, but there's no documentation warning about this.

Start with the LZMA filter-chain specifications section at the linked Python documentation page and review the descriptions of filter options, presets, modes, and filter chains. Document the parameter meanings, valid ranges, defaults, filter differences, and relevant error or memory constraints so the examples and failure cases are explained.