Show "Evaluate" parameters change after optimizing preset

Anki FSRS
1

I usually forget to check the Evaluate parameters (Log loss & RMSE) in the Evaluate button before optimizing a preset to check in which way the new FSRS parameters are better than the old ones.

A way to fix this could be to show a pop-up window showing the change in these Evaluate parameters after optimizing the FSRS parameters.

For example, you press Optimize Current Preset, the FSRS parameters are optimized and a pop-up windows appear with the following information:

“Optimizing the FSRS parameters resulted in
· Log loss: 0.4195 → 0.4170
· RMSE(bins): 2.26% → 2.20%
(Smaller numbers are better)”

1 Like
2

What goal are you trying to achieve by knowing the old values?

1 Like
3

Seeing how much better the last optimization was vs the actual.

4

They’re trying to ask as to why you want to know the previous parameters in the first place. What exactly are you trying to do with that?

5

It’s probably a language issue. OP wanted this:

Optimizing the FSRS parameters resulted in:
・Log loss: 0.4195 → 0.4170
・RMSE (bins): 2.26% → 2.20%
(Smaller numbers are better)

Expertium actually suggested this before. But then some of us want Evaluate to be removed and Dae doesn’t want any change.

6

Anki exists for one main reason: to help users learn effectively. That’s the whole point. In my opinion, developers shouldn’t waste time on requests that don’t move the needle in that direction. Who really cares if some internal value changes by 1% or 56%? It’s irrelevant to the learning experience. Effort should go into features that actually improve retention and usability—not cosmetic stats that don’t matter.

7

It’s not irrelevant and it does matter. RMSE and log loss are metrics showing how well the FSRS algorithm performs. It should be as good as possible.

How detailed they should be shown comes down to preference. This user clearly would like it to be more detailed than it currently is. Others want it removed, as long as the model is as accurate as possible (meaning they don’t really care about the values as long as it works).

Also: developers / contributers can work on whatever they deem worthwhile. And as opinions differ, the opinion regarding what feature is worthwhile or not differs too.

4 Likes
8

I understood that, my question was what they’re trying to do by knowing the numbers before optimization. These numbers don’t directly correlate with your understanding of the cards that you have (AFAIK), so I was just trying to clarify what they’re aiming to do.

9

The purpose of Evaluate isn’t to measure the user’s understanding of the material – it’s to measure the effectiveness/accuracy of the algorithm.

2 Likes