aleksejrs/anki-2.0-card-time-forecast

In the card-specific statistics table (which I use with Damien's "Card Info During Review", see below),

shows how much review time may be needed to retain the memory for 5, 10, 15 years, assuming the "Good" button is pressed every time; it does not consider forgetting index (which is assumed to be about 90%); calculating with it directly every time would be slow (if it affects the relative numbers, maybe the effect could be approximated differently?); as of 2013-06-18, it uses only the mean average of the recent ≈5 review times; it used also the median in the past, but now the functions are used in add-ons for multi-card forecasts, which need speed (calculating a median with NumPy is almost 10 times slower than calculating a mean manually);
highlights low ease with red, considering its distance from the first ease (only for cards with at least 4 ratings). Also highlights high ease with green, change opt_use_green_for_ease in the code to disable that.
highlights forecast time with red, considering its distance from 60s, with #FF red at 330s+ (5.5m).
I haven't considered Interval modifier; using it may make this add-on show incorrect forecasts and highlights.
The add-on replaces CardStats.report, so if that is updated or replaced by a different add-on, something about card info output might break (but probably no dataloss).

System requirements:

Links

Source code repository (GitHub): github.com/aleksejrs/anki-2.0-card-time-forecast;
Old repository location (Gitorious): gitorious.org/anki-2-0-card-time-forecast-and-ease-warner;
another Ease-related add-on: Suggest Starting Ease for the deck's options group.
another forecast add-on (depends on this one): Study screen time forecast.
You may be wondering if there is a way to show a forecast for each card in the browser. I have an add-on for that, which is a module for an old alpha version of Advanced Browser. I've renamed the files, so it works alongside current version of AB.

Ideas

I am not a programmer, so these are ideas, not plans.

Return to using mean for single-card calculation when NumPy is installed?
Younger cards have longer forecasts, so they look harder, and their forecast is colored red. ** When coloring, consider total time of the reviews. *** Alone, that will confuse cards whose learning has been restarted (after editing, or acquiring the knowledge necessary to memorize them).
Make the configurability of the medium ease more prominent (currently it's the Starting ease).

Feedback? If you want me to notice your feedback soon, please use either:

Changelog (by date of upload)

2015-10-16: increased the max number of reviews in a forecast from about 998 to about 99999. I hope it does not affect performance (haven't tested).
(2015-03-25: source code moved to GitHub, because Gitorious will be closed in May)
2014-12-25: for coloring, use the card's first ease, not the average of it and 250.
2013-10-23: from Anki 2.0.15: add Note ID and Card ID to the info table.
2013-06-17: stop using Numpy and calculating median, use average only. It is one dependency less for this add-on, and at least 1.5 times faster for multi-card forecast by other add-ons which will depend on this one (also added two functions for them). This made my forecast times about 10% bigger. A problem this has is for cards which have accidentally been left for a minute. However, that is good if you consider cards that have been left for a minute 2 of 5 times not by an accident. For filtered decks, the source deck is also shown now.
2013-04-04: colors forecast times over 60s with red (gradient with upper value of 330s); removed my deck-specific tweaks and 1-year forecast; fixed a bug (probably division by zero)
2013-03-27: added an option to disable green colouring of Ease numbers.
2013-03-06: first version published here. Unlike the really first version, it highlights ease and has a better code structure.

aleksejrs / anki-2.0-card-time-forecast