Review count not increasing, unable to optimise FSRS algorithm

Hi Team,

I have been having some issues with my deck presets. This Is happening with a few of them. What I normally do, for each deck I copy the preset I’m happy with and then adjust what I need to suit the specific deck I have added.

On some older decks I have used in the past but stopped and subsequently deleted before Re-downloading. I have some dodgy reviews that could effect the FSRS. I.e in the past when I was a bit more new to this I had pressed easy or hard etc. So, I’m going to the bottom and selecting ignore reviews from insert date. But, even though I have selected a date I’m happy with, say 2 weeks ago and I’ve been doing reviews daily since, the review count isn’t incrementing. Meaning, when it comes to me trying to optimise it won’t allow me due to lack of review history. If I reset the date param to default, it finds the old reviews fine. I’ve attached screenshots. Could you advise please?




Please paste ‘Debug Info’ (Settings > About > ‘Copy debug info’) below:

AnkiDroid Version = 2.20.1 (e32a82c33646807ea19623c9c88be38965bc07e6)

Backend Version = 0.1.48-anki24.11 (24.11 c47638ca36f99dd4f3b81ae82d964aec66e392e0)

Android Version = 14 (SDK 34)

ProductFlavor = play

Manufacturer = samsung

Model = SM-S918B

Hardware = qcom

Webview User Agent = Mozilla/5.0 (Linux; Android 14; SM-S918B Build/UP1A.231005.007; wv) AppleWebKit/537.36 (KHTML, like Gecko) Version/4.0 Chrome/133.0.6943.137 Mobile Safari/537.36

ACRA UUID = 5d0d47fc-8a63-4c95-a333-f7f54ffa315b

FSRS = 1.4.3 (Enabled: true)

Crash Reports Enabled = true 

Sorry the photos have pasted in an different order. They show 0 reviews when set to a jan 2025 even though I’m doing daily reviews, and then 1000 odd reviews when set to 1970

I know it sucks but it’s “Ignore cards reviewed before”, any card which doesn’t have a “learn” rating after the given date is ignored by the optimizer completely.

Hey,

Thanks for the reply, could you explain a bit more? Sorry I don’t understand.

Thanks

That feature is “ignore cards reviewed before” – if a card has been studied at all before that date, none of its reviews past or future, will be included in any optimization. To put it another way – optimization will only look at cards that were introduced after that date. That’s why you are seeing “0 reviews” when you try to optimize.

Are you sure that you have review history that is actually a problem? Unless you were grading answers Hard or Easy when that wasn’t true – FSRS can handle that just fine. Are you seeing bad parameters/bad outcomes when you optimize using your actual history? [If you post your optimized parameters, please post them as text, because we can’t see all of them or copy them from that screenshot.]

1 Like

Hey,

Thanks that makes sense! Yes, unfortunately I did have reviews hitting too easy or hard in the past, that’s why I tried to limit the older reviews.

Log loss: 0.4982, RMSE(bins): 23.52%.s

FSRS parameters

0.4026, 1.1839, 3.1730, 15.6910,7.1949,
0.5345, 1.4604, 0.0046, 1.5458, 0.1192,
1.0193, 1.9395, 0.1100, 0.2961, 2.2698,
0.2315, 2.9898, 0.5166, 0.6621

Those look like the default parameters. What about your optimized parameters when you include all of your history?

There’s nothing wrong with using the Easy and Hard buttons.

The ones I just provided in text are the params I get when I hit optimise with all of the 1184 reviews

That’s a screenshot of running Evaluate – not Optimize.

I have done both? I’m just showing you it’s the same regardless of me clicking optimise.

1000058942
Is this sufficient proof?

0.4026, 1.1839, 3.1730, 15.6910, 7.1949,
0.5345, 1.4604, 0.0046, 1.5458, 0.1192,
1.0193, 1.9395, 0.1100, 0.2961, 2.2698,
0.2315, 2.9898, 0.5166, 0.6621

The above from the gif after clicking optimise with 1185 reviews

FSRS parameters are already optimised. Once it’s optimised, you can’t reoptimise it until more reviews gather.

In your case, FSRS will perform best with the default parameters and nothing needs to change. You can probably just continue to study.

Okay, but would you not expect it to change with the amount of reviews? After 1,000+ reviews it should change from default right? And this has been an issue for a 2 weeks so I’d suspect some form of change again or not?

That probably depend on the reviews. FYI, the defaults were generated from optimising on 20,000 user collections, so they’re not just arbitrary numbers. They might just be working best for you.

If you’re still uncomfortable with it, one thing that you can do is do an optimise with * in search (everything) and use the generated parameters as the starting point when optimising on this preset. It can lead to (very slight) improvements in metrics (use evaluate).

Ok, thanks.

Its just weird I optimise my other decks like every month or so and they change quite a bit, I’ve just increased the new cards per day to 25, so I’d expect to get more wrong if I leave it for another 2 weeks and nothings changed would that indicate anything wrong? I’m just not entirely convinced.

What if I change the retention rate and then optimise would that then change them? I just want to rule out there being an actual issue. Is there a way we could test? If not I could just leave it for a while the thread has a month before closure.

Thanks for all your help guys I appreciate your time.

Can you update to the latest (25.02) version and try again? It’s a bit weird that your RMSE is so high.

There doesn’t seem to be any updates in the app store.

I’m a beta tester would that make a difference?

This is what it’s saying my version is in app

I’m not sure why you’re not getting optimized parameters with 1000+ reviews. The only thing I can think of is if your study patterns are too inconsistent to map – but I haven’t seen that before. Your really high RMSE suggests that something is wrong, at least.

No, the Desired Retention is not a factor in setting the parameters.

You’ve been asking about AnkiDroid – 25.02 is a desktop Anki version.

I don’t think you need to update your desktop Anki, but have you tried this on desktop instead? The apps are using the same Deck Options/FSRS screen and code, so it shouldn’t be different, but that would eliminate one variable.

Other things to try –

  • Reset your parameters to default (click the reset button and then confirm to reset) – and Save. Open the Deck Options again and Optimize. Any different?

  • Create a new options preset (but don’t clone this one, in case it’s the problem), switch this deck to that preset – and Save. Open the Deck Options again and Optimize. Any different?