I’ll also second Danika’s post to consider what exactly you wanted to use Anki for and whether this scheme of yours makes sense.
But anyway, if you want to do custom scheduling, here’s a whole general how-to that ought to be useful for anyone wanting to do stuff with custom scheduling. Perhaps this could be its own forum thread… I haven’t tried modifying learning steps myself but I’m pretty sure what I describe for changing steps etc. is how it works.
Firstly
You can generally figure anything not mentioned here out by inspecting the states object. You’ll need the Webview Inspector addon to see the browser console in Anki. I had to double-check a lot of things too.
console.log(states);
How learning steps work
- The new state occurs only on the first review. On the next review the state will be learning or review depending on the number of learning steps.
- The base interval the you modify with custom scheduling is what is defined in the deck setting’s learning steps.
If you have at least 1 learning step
- Answering Again at any learning step (including the first) resets the card
to the first learning step. The card’s state will still be learning.
- Answering Hard keeps you at the current learning step. The card’s state will still be learning
- The interval shown for Hard is half of the next learning step, so half of what the interval shown for Good
- If there is no next learning step (the current is last) the interval shown is 1.5 times the current step’s interval
- Answering Good, the next state will be review if this is the last learning step, or learning if there is a next learning step.
- Answering Easy at any learning step will skip all learning steps and the next card state will be review
If you have no learning steps
- All buttons graduate the card to review state, the card’s state will never be learning because the state is new for the first review.
Relearning steps after answering Again
Exactly the same as above but replace learning with relearning
Some helper code
Answer button object you modify
It’s complicated…
This is different depending on whether you’re in a normal deck or a filtered deck + the card state + the answer button itself. And, if in a filtered deck, whether you’ve checked Reschedule cards based on my answers in the this deck and if so, which answer buttons return the card to the original deck…
For example, for the Good answer button:
states.good.normal
- You’re in a normal deck.
- Or, you’re in a filtered deck and answering this returns the card to the original deck
Filtered deck with Reschedule cards based on my answers in the this deck checked (the default):
states.good.filtered.rescheduling.originalState
- You’re in a filtered deck and answering this keeps the card in the filtered deck
Filtered deck with Reschedule cards based on my answers in the this deck unchecked:
states.good.filtered.preview containing { scheduledSecs: <number>, finished: true/false }
Filtered deck
When Reschedule cards based on my answers in the this deck is unchecked
- learning and relearning steps don’t exist. Cards just stay in the filtered deck and repeat according to the intervals defined in the filtered deck settings.
- You could make those intervals grow with custom scheduling!
Again
The object you modify to change what the Again button does for a review card changes depending on whether
- you’re in a normal deck
- in a filtered / custom study deck
- and whether there are relearning steps or not
And then also whether the card is currently in learning or releaning.
// Card is currently review or there are no learning steps
const againRevObj = states.again.normal?.review // normal deck, no relearning steps
|| states.again.normal?.relearning?.review // normal deck, have relearning steps
|| states.again.filtered?.rescheduling?.originalState?.review // filtered deck, no relearning steps
|| states.again.filtered?.rescheduling?.originalState?.relearning.review // filtered deck, have relearning steps
// Card is currently in a learning step or new
const againLearnObj = states.again.normal?.learning // normal deck
|| states.again.filtered?.rescheduling?.originalState?.learning // filtered deck
// Card is currently in a relearning step, there are TWO objects to modify
// Review object, whose scheduledDays is what the eventual interval will be after
// the last relearning step. This is lowered a again if you answer again in relearning!
const againRelearnRevObj = states.again.normal?.relearning?.review // normal deck
|| states.again.filtered?.rescheduling?.originalState?.relearning?.review // filtered deck
// Learning object, since answering again returns the card to the first relearning step
// you have this. Modifying this can change what relearning step it goes or the relearning
// interval. Deleting this would make the card not go back to relearning but instead go
// back to being a review card!
const againRelearnLearnObj = states.again.normal?.relearning?.learning // normal deck
|| states.again.filtered?.rescheduling?.originalState?.relearning?.learning // filtered deck
Hard
const hardObj = states.hard.normal // normal deck
|| states.hard.filtered?.rescheduling?.originalState // filtered deck
// Card is currently review or there are no learning steps
const hardRevObj = hardObj.review
// Card is currently in a learning step or new
const hardLearnObj = hardObj.learning
// Card is currently in a relearning step: two objects
const hardRelearnRevObj = hardObj.relearning?.review
const hardRelearnLearObj = hardObj.relearning?.learning
Good
const goodObj = states.good.normal // normal deck
|| states.good.filtered?.rescheduling?.originalState // filtered deck
// Card is currently review, there are no learning steps, or this is the last learning step
const goodRevObj = goodObj.review
// Card is currently in a learning step and there is a next learning step
const goodLearnObj = goodObj.learning
Easy
const easyObj = states.easy.normal // normal deck
|| states.easy.filtered?.rescheduling?.originalState // filtered deck
// Easy answer always defaults to skipping learning/relearning steps so in all states
// you'll only have this
const easyRevObj = easyObj.review
Getting the current review object
For example, you want to
- get the current
easeFactor so you can modify the next easeFactor in <answer>RevObj.normal.review
const revObj = states.current.normal?.review
|| states.current.normal?.relearning?.review
|| states.current.filtered?.rescheduling?.originalState?.review
|| states.current.filtered?.rescheduling?.originalState?.relearning.review
const currentEaseFactor = revObj?.easeFactor;
const previousInterval = revObj?.scheduledDays;
Modifying learning steps
- If you have
<answer>LearnObj, answering this keeps the card in learning
- If you have
<answer>RevObj: answering this graduates the card to review
As per the above description of how learning steps work and the answer button object variables.
If you have learning steps, you’ll have
againLearnObjhardLeardObjgoodLearnObj (if there is a next step)- or
goodRevObj (if the current step is last)
- and
easyRevObj
With zero learning steps you’ll have
<answer>RevObj for all answers
Modifying interval or the next step
<answer>LearnObj.remainingSteps: which learning step is the next one. If you have 5 learning steps, setting this 5 means going back to the first step.<answer>LearnObj.scheduledSecs: interval to wait, default value is defined by your learning steps definition in the deck settings and the behaviour described above for Hard
When answering gradutes the card to review
easyRevObj.scheduledDays: As described above, answering Easy graduates the card to review state. Review state doesn’t have scheduledSecs, they have scheduledDays. I think making intervals less than a day for review cards is not possible, the scheduler will always set a review card’s interval to 1 day.
Modifying relearning steps
Mostly same as above, replace learning with relearning. But as mentioned in the answer button objects, relearning steps can have TWO objects in them:
review: defines the eventual interval after the last relearning step- and
learning: if there is a next step or you’re not advancing in steps (hard & easy)
The learning object is omitted by default for the Easy answer relearning object.
Reminder on modifying interval
Remember that learning and relearning states have scheduledSecs and review states have scheduledDays. As mentioned, I’m not sure if it’s possible to make review intervals less than 1 day but you’ll find out by trying.
Extra advanced
Modifying whether an answer graduates a learning card to review or not
I haven’t tried but I imagine you could do this with a card in learning:
// Make answering easy skip one learning step instead of graduating to review
const nextRemainingSteps = goodLearnObj?.remainingSteps;
// Skip a step if there are two steps still left
// this will be false, if nextRemainingSteps is undefined (not in learning or current step is already the last)
if (nextRemaningSteps >= 2) {
delete states.easy.normal.review;
states.easy.normal.learning = {
remainingSteps: nextRemainingSteps - 1,
scheduledSecs: 60, // Set some interval
elapsedSecs: 0, // This is always zero, does nothing I guess?
easeFactor: 2.5 // Need to set some factor value for SM-2
};
}
Or the same idea applied for answering Hard for a review card. Making it put the card into relearning could be possible too?
It’s even better than SM-2, and once personalized, it does better at scheduling for 98 or 99% of users (the stat keeps going up, so I’ve lost track). The smart money right now is on: 
Even if you’re trying to mimic a system that has worked for you in the past, you’re overlooking the built-in systems that might be substantially better and/or more efficient at helping you learn.
]
