Interesting. Using this query:
___
SELECT count(*) as TotalReviews,
( ( SELECT count (*)
FROM revlog
WHERE ease > 1 ) * 100.00
) / count (*) as OverallAcc,
(( SELECT count (*)
FROM revlog
WHERE ease > 1 AND lastIvl > 60 ) * 100.00
) / ( SELECT count (*)
FROM revlog
WHERE lastIvl > 60 ) as Older60Acc,
( select ( 100.00 - ( sum(lapses) * 100.00 ) / sum(reps) ) from cards ) as LapsesAcc,
( SELECT count (*)
FROM revlog
WHERE ease > 1
) as TotalCorrect,
( SELECT count (*)
FROM revlog
WHERE ease <= 1
) as TotalWrong,
( select count(*) from cards ) as TotalCards,
( select sum(reps) from cards ) as TotalReps,
( select sum(lapses) from cards ) as TotalLapses
FROM revlog
___
Gives me:
Total accuracy for all reviews: 93.8%
Accuracy for cards older 60 days: 93.2%
Accuracy based on lapses vs. reviews: 97.5%
So the lapses don't seem to be a very valuable information since they ignore wrong answers for new learned cards and multiple wrong answers in a row.
The "old cards" accuracy (I've chosen 60 days) is maybe the most important number since I want to know how many cards I "really" memorized.
Do you know what is supposed to be a good rating for old cards? What accuracy do you have there?