Advanced Statistics

If Advanced Statistics are enabled, it changes the Forecast graph so that it shows the estimated number of reviews that will be due on a given day in the future taking into account future reviews, learning new cards and failing cards. The bars and the left axis show the number of cards due on each day if you study all cards each day, while the line and the right axis show the number of unseen (shown as learn), young and mature cards your deck or collection will consist of if you study all cards each day. The forecast graph does count reviews that are currently overdue. It assumes that the overdue cards will be reviewed according to the maximum reviews/day deck option.

Advanced Statistics can be enabled in Settings -> Advanced -> Advanced Statistics (in plugin section).

The outcome of future reviewing, learning or failing cards affects reviews after that future review. To take this into account, the probability of each outcome is computed from the review log. Then the outcome is randomly chosen, such that an outcome which is more likely according to the review log is more likely to be chosen than an outcome which is less likely according to the review log. The settings all affect how the effect of the outcome of future reviews on subsequent reviews is taken into account.

Compute first n days, simulate remainder

If this setting is set to a number greater than 0, instead of randomly choosing an outcome, each possible outcome is taken into account in the simulation, together with its probability. The probability is taken into account for the graph and for future reviews in which it results, which also affect the graph. One review has a couple of possible outcomes (say 4), which all result in a review. That review also has a couple of possible outcomes and so on. If many reviews are simulated this way, many reviews (4 x 4 x 4 x … ) have to be taken into account which increases the time it takes to compose the graph. Therefore, for reviews later than n days from now are simulated by randomly choosing an outcome.

In summary, higher n gives a more accurate graph, but it takes more time to compose the graph.

Precision of computation

Reviews which occur with a probability smaller than 100% minus the configured precision of the computation are simulated by randomly choosing an outcome rather than taking into account each possible outcome. This setting is only applicable if the first n days are computed. If Advanced Statistics are disabled, the Forecast graph shows the estimated number of reviews that will be due on a given day in the future if you do not review cards, learn no new cards and fail no cards.

In summary, higher precision gives a more accurate graph, but it takes more time to compose the graph.

Number of iterations of the simulation

Composes the graph several times and then displays the average of these graphs. Each time the graph is composed, another outcome might be randomly chosen. If we average many outcomes which are randomly chosen taking into account the probabilities from the review log, the average outcome will likely be close to the average of the probabilities from the review log. If we average many graphs, the average graph will likely be close to the graph which is generated by taking into account all possible outcomes. If the number of graphs which are averaged is not too high, it will be faster than taking into account all possible outcomes.

In summary, a higher number of iterations gives a more accurate graph, but it takes more time to compose the graph.