Just to point out, in the latest beta that was just released a little while ago (28.05b), the Help Me Decide button’s ratio graph is still displaying (t_end - t_start) / R_end , which is, once again, not useful information for the user, as opposed to what was discussed in this thread of (R_end - R_start) / (t_end - t_start), which is, once again, the correct formula to allow a user to calculate the DR which will maximize their amount of information retained per unit time.
I do have… a bunch of math for how to optimize and implement and help the user simply choose the optimal DR for a flat DR policy, (which Anki could really benefit from at this moment in time), and hide all the scary math away from him, but it seems that SSP-MMC-FSRS is the future and also fixes a bunch of problems with calculating an optimal flat DR (of which there are numerous).
@acbraith I’ve been playing around with low DR and how Anki interacts with it. As it is at the current state, even a 70% DR is not in a suitable state for mass-adoption due to the scheduling system and the time it takes to calibrate the FSRS parameters to such values. (I think I mentioned something about this earlier in the thread about how an entire overhaul of the entire scheduling system would be necessary to fix this issue.) Even if the Simulator indicates 70% or lower as being optimal, that isn’t necessarily the case as the FSRS curve itself might not be properly calibrated in that region. Let alone the additional problem of slowly lowering the DR down to such a value without possibly losing tons of ΣR from the user due to a miscalibration.