Different questions fared quite differently in 2020-01 Quantum Country efficacy experiment. The distribution of accuracies without review, after a month 20210318091738:
0.4, 0.43, 0.50, 0.66, 0.69, 0.72, 0.84, 0.86, 0.87.
This jives with a general theme from Quantum Country data: QCVC questions are initially forgotten at very different rates.
These accuracies roughly (but not exactly!) align with other measures of “question difficulty”. For instance, the ranks of these questions among the questions which most drive lapses (Half of all long-term Quantum Country lapses come from just 12% of its questions):
20, 25, 7, 36, 21, 42, 51, 67, 66
Ranks in terms of Quantum Country first repetition session accuracy measures:
11, 14, 4, 35, 30, 38, 46, 53, 75
(It’s surprising how skewed these nine questions are toward the bottom half of the distribution. Suggests there may be some essay-order effect.)