FSRS 5 vs 6 – Is there any statictical data that proofs that the new version of FSRS 6 is better than previous ones?

Hi,

Is there any statictical data that proofs that new version that FSRS 6 is better than previous ones?

The SRS benchmark repo: GitHub - open-spaced-repetition/srs-benchmark: A benchmark for spaced repetition schedulers/algorithms

I’m currently working on plotting graphs that will visually showcase the improvements between different versions of FSRS. Preview: Calibration of different FSRS versions - Album on Imgur

EDIT: I finished adding all graphs.

7 Likes

I’ve seen in GitHub - open-spaced-repetition/srs-benchmark: A benchmark for spaced repetition schedulers/algorithms that there are various versions of the FSRS-6 data (default para, pretrain, recency…).

Does this mean that recency weighting has to be enabled manually? Or is it on be default?

It’s turned on by default. No customisation is possible from the user side. The benchmarking is simply testing FSRS with and without some code changes.

1 Like