SuperMemo2 incorrect

[mbid]

I checked the implementation of SuperMemo2 here because the specification was unclear to me. I think there are a few problems:

• The specification says "If EF is less than 1.3 then let EF be 1.3.", but I don't think this is enforced somwhere here.
• As I interpret the specification, we have I(n) = EF(n) * I(n -1) if Q(n), Q(n - 1), Q(n - 2) are all >= 3, where Q(n) is the response quality in the nth review, EF(n) is the E-Factor after (or before, this was unclear to me) the nth review, and I(n) is the number of days between the nth and the (n + 1)th review. However, here the number is calculated as I(n) = EF(n) ^ (n - 2) * I(2), which is not the same.

I've written a similar tool (wanted to use my csv files, and of course NIH), which contains an implementation of SuperMemo2 as well, see here. Maybe we could factor out an implementation into an independant library.