Currently, my deck presets are set to 70% Desired retention (DR). In the latest Anki, changing this up or down has a new box that indicates the new Approximate Workload (AW) as a multiplier. How exactly is this value calculated? Is “workload” defined by number of cards to review per day, amount of time spent per day, or something else?
Using one deck as an example, increasing DR from 70% to 82% has an AW change of 1.11x. This sounds like an incredible deal; a mere 11% workload increase for a 12% retention increase would be very worthwhile for me.
But if I use the FSRS simulator and compare 70% vs 82% with 10 new cards/day, the workload for 82% is far higher than 1.11x on both the Reviews and Time metric. In fact, it’s nearly 50% higher.
How did you increase it? If you put any intermediate values, then the AW would be incorrect. So, if you 70→75→82 then Anki compares 75 with 82.
If you’re really seeing such a drastic discrepancy between the info box value and the sim, an argument could be made for removing the infobox, or improving it further.
Simulator is far more accurate than what you’re using. The info box just uses rough estimates of time to quickly calculate a value. But the expectation wasn’t that info box would be really off like that.
I was increasing the DR by clicking the up arrow and seeing how much each step increased the workload. I did not save the settings in-between each change. The info box also does explicitly say “(compared to 70%).”
I tried this with my other two presets that were set to 70%. Both said the AW would be 1.10x or 1.11x , but when comparing the values in the FSRS simulator the actual workload difference was about 30% higher. It’s hard to make a perfectly fair comparison though, because there is a lot of noise in the FSRS simulator graph.
If you’re really seeing such a drastic discrepancy between the info box value and the sim, an argument could be made for removing the infobox, or improving it further.
I would say this is warranted. At present, I would argue that the infobox is actively harmful. If I hadn’t cross-checked it with the simulator, and then here with this forum post, I would have unwittingly gave myself a 50% increase in workload (on one deck).
@L.M.Sherlock I believe this is similar to what another user said on r/Anki. Perhaps the idea with approximate workload wasn’t good after all. Then the problem is - what’s next? We can’t use the actual simulator for this because it’s too slow
Here’s a download link for my collection. These are the relevant presets I mentioned that are at 70% DR now, and have ~1.11x shown when I increase to 82%.
FSRS Japanese Vocab (this is the one in the screenshot)
Japanese Sentences Morphman
anime