Determining the optimal retention for a significant amount of time spend / new cards done per day

Anki FSRS
1

I used the Colab optimizer and got R = 82% which I assume is the minimum recommended retention (optimal). The values were the default, other than learn_span = 90 days.

However, I heard that it apparently assumes you are only learning 10 cards per day. Once the holidays come up, I’m going to be doing another challenge with a friend for 3 months, where I will likely average 100 new cards per day (with peak days of maybe 200-400). Since the optimizer assumes 10 cards per day, would it affect the optimal retention significantly? And if so, does it place it higher or lower than if it had assumed I do more new cards per day? In short, if I’m planning to do more cards (such as 10k-20k), in a short period of time (90-150 days) would it be better to place the retention higher or lower than what “FSRS4Anki v5.1.0 Optimizer” gives? Or does it not have a significant difference?

There is also “FSRS4Anki v4.14.4 Simulator”, but the results are confusing… a graphical comparison between Anki and FSRS?

2

Since the optimizer assumes 10 cards per day, would it affect the optimal retention significantly?

No because FSRS doesn’t calculate how tired you would be. Also, I recommend using the built-in “Compute minimum recommended retention” if you are using the latest beta. I think you are getting into a habit of using the Google Colab version too much.

1 Like
3

Yeah, I just noticed I tend to get more fitting results from google colab.

I wonder though, it is normal for people to get a minimum recommended retention of 0.87? Most of the reddit posts, etc. I’ve seen had people stating values between 0.75 and 0.80.

The inbuilt computer gave me 0.87 for 10 years (0.85 for 90 days) and google colab gave me 0.82 for 90 days.

Edit: more importantly, when I simulated more days in the previous version of Anki, it bought the recommended retention DOWN. But when I simulate more days in the beta version of Anki, it brings the recommended retention UP. This seems unexpected.