The FSRS Simulator graph & tooltips are wrong

1. The number of reviews in the tooltips seem to be off by one day

The easiest way to see this is by mousing over a peak, looking at the number of reviews, then move one to the right. The number will go up instead of down.


2. The number of reviews tomorrow is sometimes significantly less than it should be.

N.B. I’m deliberately looking at “in 2 days” to account for the first issue.

e.g.:


Interestingly the number is correct if you simulate 30 days:

4 Likes

I have also seen a very stark mismatch between what the graph shows and what the tooltip shows (which might be related to #1 in OP).

A simulation for 3 days:
Screenshot-2024-12-17-00-46-17-53-a9eef3a2a561b80d5c76daebd0f9a14c

Look at the y-axis and it should be 30 cards for tomorrow but the tooltip shows 6 cards. The number for two days later is 30 cards which seems correct.

My educated guess is it’s showing me the number of learning cards I’ve left for today.

1 Like

Steps

  1. Run a simulation using the defualt values.
  2. Run a simulation for 3650 days.
  3. Click on a part of the graph after the point where 1st simulation ended.

Behaviour

The tooltip shows you number of reviews for both of the simulation but as the first simulation ended at an earlier day it should show nothing. Currently it gets the data for the last simulated day instead and shows it to the user.

1 Like

Already mentioned here: The FSRS Simulator graph & tooltips are wrong

1 Like

I was talking about this:

Screenshot-2024-12-17-02-13-50-08-a9eef3a2a561b80d5c76daebd0f9a14c

The first simulation ends at 30 days. After 53 days, there should be nothing to show me from the 1st simulation.

It’s a different issue from the two I mentioned before.

Poor Jarrett, it feels like we are dogpiling on him with all these issues now.

That’s why I was orignally not mentioning it while it said “(experimental)”.

Yes, it’ll be nice if someone else looks into this and sends a PR. Jarrett might have too much on his plate.

Reason for merging different posts? @Danika_Dakika

I merged these topics – whether the graph is wrong or the tooltip is wrong or both are wrong is all closely related.

I wasn’t reporting anything being wrong. The tooltip was just showing a previous data-point. Well, anyways. I thought it’s a more visual bug than anything.

If the graph and the tooltip don’t match – doesn’t one of them have to be wrong? :wink:

1 Like

Relevant: Fix Fsrs simulator input problems by Luc-Mcgrady · Pull Request #3642 · ankitects/anki · GitHub

That PR doesnt touch the graph data itself, just the surrounding inputs.

I guess the focus line doesn’t match the data points well:

Perhaps it’s caused by moving average:

The current window size is daysToSimulate / 50.

The #1 will be fixed in Fix/dataPoint index is off by one day in simulator by L-M-Sherlock · Pull Request #3645 · ankitects/anki · GitHub

4 Likes