It doesn’t because one bin presets five days here.
Oh, I missed that. But then, including >1 day in a bin is confusing. Probably, that is why you said that a line chart will be better.
But, I think that including these charts will be useful. Maybe add a button saying something like “Plot a graph showing workload against desired retention (slow)”?
But I thought the simulator would also do what “Compute minimum recommended retention” is doing, no?
This was meant for LMSherlock
If the simulator doesn’t do what Compute minimum recommended retention is doing, then that functionality will be lost. That’s a bad idea. I think you should add an extra button that will make the simulator run the same minimization procedure as Compute minimum recommended retention, but with new settings.
The simulator is not an alternative of the “Compute minimum recommended retention” feature. It is a different feature and is meant to be a version of Anki Simulator that supports FSRS.
The only thing that connects them is that “Compute minimum recommended retention” needs to perform multiple simulations to calculate the minimum value.
Fine. If this idea is bad, is hard to implement and doesn’t satisfy anyone’s need, I would just give up. I can just make it open to Anki’s Python backend, which will allow add-on developer to use it.
No, don’t get me wrong, the simulator looks promising. I just think that it should be able to do what CMRR did AND more. Computing minimum recommended retention should be one of the features of the simulator.
Is the MRR here going to be more accurate? I’m also struggling to understand what would be so useful about a simulation other than you knowing how the workload is going to look like for some particular settings. Is the graph you showed useful in some cases? I personally wouldn’t care how workload:knowledge changes with desired retention [but playing devil’s advocate, wouldn’t you say something like knowledge acquisition rate is much better to use here?].
Are you sure about that? You don’t want the best bang for your buck? You don’t want a graph instead of just a single number?
Also, yes, CMRR is going to be more accurate due to the following reasons:
- Using real new/review limits
- Using real deck sizes
- Using memory states of existing cards
CMRR already does what I want. Tell me this, why would I want a graph with workload:knowledge values plotted against desired retention? I just need the desired retention at which this is lowest, what will I do with everything else?
If its going to be better in simulation, then just improve CMRR why bother putting this graph over there?
@L.M.Sherlock what do you think? Personally, I think a graph is better than a single number, but I’m fine with either. Just please add a “Compute minimum recommended retention” button somewhere in the simulator, so that the users won’t lose that functionality.
1 Like
Many will prefer to set the value higher. The schedule will be useful for them.
The graph will also visually explain what the function does.
1 Like
Just please add a “Compute minimum recommended retention” button somewhere in the simulator, so that the users won’t lose that functionality.
He has not removed any existing functionality. The simulator is an extra feature; it doesn’t replace anything.
To reword what @Keks said, the graph will allow the user to know how much their workload will be affected if they set a higher retention rate. If it is not too much, they may want to use a higher desired retention even if it is slightly less efficient.
Uhh… are you against this or for this?
I support adding the graph between workload (and/or workload/knowledge) and desired retention.
I wonder which comment of mine caused you to think that I am against it. I just said that it is not the MAIN function of the simulator.
I misunderstood you. Anyways, you might have misunderstood me too. I am fine with workload/desired retention graph. The other one (workload:knowledge one) doesn’t even make sense to me. It’s honestly the least intuitive thing you can add. Let’s say workload:knowledge is the same for 80 to 85 percent. Then it increases by some unit (what?) from 85 to 90 per cent. Now how much of that is because of denominator change and how much due to other? What does a person even make out of it: 90% is okay or not okay?
I think you misunderstood the graph. It’s not “workload on the Y axis, knowledge on the X axis”, it’s “Workload divided by knowledge on the Y axis, desired retention on the X axis”.
He has not.
Your argument makes sense. When retention changes from 85% to 90%, if the W/K ratio changes from 5 reviews/card to 7 reviews/card, it doesn’t mean a 40% increase in the workload because the knowledge will increase too. This graph would be very difficult to interpret.
So, we should just add workload vs desired retention graph. @Expertium, thoughts?
I think workload/knowledge makes it more clear that there exists an optimal value of desired retention, but whatever. I’m fine with it either way.
workload/knowledge It better explains how the optimal value is selected. It also shows more clearly the loss of efficiency with a decrease in desired retention. It will also be useful for those who want to set a value higher than the optimal one, provided that they can figure out this graph.
1 Like