How to interpret the output from Accuraterip verification?

Verifying a CD rip with accuraterip can be done on Linux using arverify from the cdrip-tools package as this question explains. However, I've struggled to find any documentation that explains the output produced by this process.

The output from arverify looks like this:

17.flac
Calculated CRCv1    : D3293E04             
Calculated CRCv2    : 9CE51529             
Calculated CRC450   : A0D2A24A             
Database entry      : CRC: ABEBB4B7, Confidence:  76, CRC450: 7160BA39
Database entry      : CRC: 11405B1F, Confidence:  27, CRC450: 389F7025
Database entry      : CRC: E82F596A, Confidence:  21, CRC450: A64BF887
Database entry      : CRC: 88BCC481, Confidence:  16, CRC450: 4231C09D
Database entry      : CRC: 56A0A258, Confidence:   5, CRC450: 52CF62C1
Database entry      : CRC: 851CDA29, Confidence:   3, CRC450: A127235F
Database entry      : CRC: 96341314, Confidence:   2, CRC450: 3AA482F3
Database entry      : CRC: 788D6152, Confidence:   2, CRC450: 00000000
Database entry      : CRC: 77A373E0, Confidence:   2, CRC450: 00000000
----------------------------------------------------------------------
Accurately ripped with offset 96 (confidence 27+27/154)                     
Accurately ripped with offset -598 (confidence 2+2/154)
Accurately ripped with offset 84 (confidence 21+21/154)
Possibly accurately ripped with offset 16 (confidence 2/154)
Possibly accurately ripped with offset 96 (confidence 76+76/154)
Possibly accurately ripped with offset -598 (confidence 21+21/154)
Possibly accurately ripped with offset 84 (confidence 27+27/154)
Possibly accurately ripped with offset 106 (confidence 5/154)
================================================================================     
17/17 Accurately ripped with offset 84 (confidence 42)
17/17 Accurately ripped with offset 96 (confidence 54)
17/17 Accurately ripped with offset -598 (confidence 4)
17/17 Possibly accurately ripped with offset -598 (confidence 42)
17/17 Possibly accurately ripped with offset 16 (confidence 2)
17/17 Possibly accurately ripped with offset 84 (confidence 54)
17/17 Possibly accurately ripped with offset 96 (confidence 158)
17/17 Possibly accurately ripped with offset 106 (confidence 5)

How should this information be interpreted?

Specifically, I don't understand why the Calculated CRC values don't match a Database entry. Of those database entries, what is the confidence value? (is it the number of matches for that CRC = the number of people reporting the same CRC as me) ?

Also, what does confidence 27+27/154 mean?

I think 154 is the total of the listed confidence values for the listed database entries (76+27+21+16+5+3+2+2+2=154) presumably the total number of samples in the database.

I think the 27 is the number of matches with the same offset as my drive (96) and it's saying 27 out of 27 of those matched. Is this right?

Why are there multiple Accurately ripped reports and why are there also multiple Possibly accurately ripped reports?

In the summary, it reports that

17/17 Accurately ripped with offset 96 (confidence 54)

I picked that line out because my drive offset is 96. I guess this is telling me that 54 other people (whose drives have the same offset as mine) ripped the 17 tracks identically. I'd interpret from this that my rip is good.

What is the relevance of the other entries for other offsets and also those listed as Possibly accurately ripped?

How does one interpret the overall report to decide whether a rip is good?

The above rip is an example of one performed without sample offset correction. Here is another example of an accuraterip assessment for a rip performed with the correct sample offset (96 in this case) applied:

Calculated CRCv1    : B75C614C
Calculated CRCv2    : DEB788B6
------------------------------
Accurately ripped (confidence 105+101/307)
Accurately ripped with offset -29 (confidence 18+18/307)
Accurately ripped with offset -628 (confidence 20+20/307)
Accurately ripped with offset -947 (confidence 18+18/307)
Possibly accurately ripped with offset -29 (confidence 101+101/307)
Possibly accurately ripped with offset -628 (confidence 20+20/307)
Possibly accurately ripped with offset -947 (confidence 20+20/307)
================================================================================
11/11 Accurately ripped (confidence 213)
11/11 Accurately ripped with offset -29 (confidence 40)
11/11 Accurately ripped with offset -628 (confidence 42)
11/11 Accurately ripped with offset -947 (confidence 38)
11/11 Possibly accurately ripped with offset -947 (confidence 42)
11/11 Possibly accurately ripped with offset -628 (confidence 40)

The main difference is the presence of the line

11/11 Accurately ripped (confidence 213)

which reflects accuraterip matches with the same offset as the rip (which is 96 in this case, but not shown). This is the line which reports the overall accuracy of the rip and that, in this case, 213 other checksum reports match ours.

Here is an example of the same disc ripped by the same software (abcde in this case) but without a sample offset:

Calculated CRCv1    : AF3854B4
Calculated CRCv2    : F5401F9C
------------------------------
Accurately ripped with offset 96 (confidence 101+101/307)
Accurately ripped with offset 67 (confidence 18+18/307)
Accurately ripped with offset -532 (confidence 20+20/307)
Accurately ripped with offset -851 (confidence 18+18/307)
Possibly accurately ripped with offset 96 (confidence 105+105/307)
Possibly accurately ripped with offset 67 (confidence 101+101/307)
Possibly accurately ripped with offset -532 (confidence 20+20/307)
Possibly accurately ripped with offset -851 (confidence 20+20/307)
================================================================================
11/11 Accurately ripped with offset 67 (confidence 40)
11/11 Accurately ripped with offset 96 (confidence 210)
11/11 Accurately ripped with offset -532 (confidence 42)
11/11 Accurately ripped with offset -851 (confidence 38)
11/11 Possibly accurately ripped with offset -851 (confidence 42)
11/11 Possibly accurately ripped with offset -532 (confidence 40)
11/11 Possibly accurately ripped with offset 67 (confidence 210)
11/11 Possibly accurately ripped with offset 96 (confidence 216)

Here there is no line without an offset. You will also notice that the two calculated CRC values (there are two because Accuraterip has two versions of its CRC) differ from those above.

The reported offsets are different to those used for the rip and, because no offset was applied, all of them are different. This is showing us accurate rips for this disc at various offsets (these don't match our rip but are shown for information).

To get an accurate rip it is necessary to rip with a specific offset.

Offsets are reproducible using different software (such as abcde or Whipper on Linux) as long as the correct sample offset is used.