Yeah, it’s not very meaningful. It’s just something I did in the past, out of curiosity.
You asked here what it means replying to that exact post.
You wanted to know the difference in accuracy of predicted error rates in 4-button vs. 2-button users, that suggested analysis, if it works, would be a way to get at that.
That’s actually the best way to do this. As @nmjkjm said earlier, there may be a fundamental difference between people who use 4 buttons and people who use 2 buttons. So taking a bunch of 4 button users and splitting them into “Keep doing this” category and “Use 2 buttons” category in real life, not by doing data manipulation, is the way.
1 Like
No I didn’t mean that. That sentence doesn’t even make sense. You say, “you wanted to know…” not followed by a question.
Yeah but has that been done before or will anyone even do it? I personally don’t believe in what nmjkjm says. What significant differences can exist between this group that would affect the RMSE?
From a purely theoretical standpoint, and this is something you and sherlock can probabaly do better, what are the user traits that can influence RMSE? (he mentions complexity of material but idk why button use will have a correlation with that)
Yeah but has that been done before or will anyone even do it?
It hasn’t been done before, and it would be a logistical nightmare to do, unless some researcher institution helps us.
From a purely theoretical standpoint, and this is something you and sherlock can probabaly do better, what are the user traits that can influence RMSE?
- Inconsistent usage of buttons, such as using Hard as Fail, or pressing Good because “Whatever, I don’t care”, or remembering a card better after pressing Hard than after pressing Good.
- Leeches. Cards that just don’t “click”.
Idk what else
might be possible that people who use 4 buttons are newer users; people who have been using Anki for a longer time (say 2 years) tend to use 2 buttons more. Is there a way to look into that in the data and rerun the analysis on older users’ data. Other than that, I don’t think older users with that attitude will press 4 buttons. They’ll be part of the 2 button crowd.
Is there a way to look into that in the data and rerun the analysis on older users’ data.
The Anki 20k dataset only has interval lengths between reviews, not absolute datetime values, so the best I can do is find users who’s maximum interval (across all reviews) greater than, say, 730 days.
That is fine. A person using Anki for a few months (before quitting) would not have intervals of more than a year. Even in the most liberal scenario, with settings set to default, an interval of around a year takes 5 months. What interval to choose will be a bit arbitrary, but 730 is a nice number IMO.
Here. This is based on 6747 users with max(delta_t)>730. Aside from the curves looking a little more jagged and the green zone becoming slightly smaller, nothing changed.
1 Like
I think I’m convinced 2 buttons are just better. I myself rarely use Hard+Easy, maybe I press Easy once a month. This is good. Let’s see if others have something to say.
Also @L.M.Sherlock I’m sorry I’m pinging you a lot recently but do you think we might have missed something here: Pass/Fail Grading as Default - #67 by Expertium
Did you write this or was someone else using your account?
Clearly you wanted to know if shifting from 4 button use to 2 button use will benefit the user bc you just said it here. I said you wanted to know it, so not sure why you’re denying it. There’s a difference between 4-button users and 2-button users and 4-button use vs. 2-button use, bc you want to know that changing from 4 buttons to 2 buttons will not adversely affect error rates.
I already said above it was done in SM. I don’t find a link immediately but maybe you should take a look at that analysis if you find it. And to see how it was done. To look at error rates of users that use 2 buttons vs. users that use 4 buttons is a very basic statistical mistake.
This is the issue, you’re already made up your mind and are looking for ways to do the analysis that will confirm your thinking. There is a correct way to do it but you’re not interested. As I pointed out, and I’m not sure the experiment that Expertium suggested is not possible (bc depending on how the buttons are combined the intervals could become longer and shorter, so some assumptions could be made, eg if someone had recall on day 20 then presumably they would also have remembered on day 17), there might be a way to answer your question.
A simpler way to do it is to change a new version of the software to randomly display 4-buttons to some users and 2-buttons to some. But even then it might not work bc probably many 4-button users would change the settings back to 4-buttons if it was possible.
It defies common sense that going from 4-buttons to 2-buttons would increase accuracy. If I have a item I remember very easily and select the highest grade (4), why wouldn’t I want a longer interval than for something that I only recall after great difficulty (2) that clearly needs a shorter interval? Already the highest percentage button for me for young and for learning items is 4 and they come back a number of times, so the intervals already seem too short for many items. But I’m wondering if optimizing the parameters for different presets and changing the optimal retention will change this.
See you’re misunderstanding what I was trying to say. I don’t see the point of arguing with people as if you know them better. I was simply asking Expertium if what he says is true what new conclusion we might arrive at but you don’t need to argue about it.
Don’t know about anything else but it’s very clear to me that you were never taught how to properly talk to people.
Canary release for this would be nice. Not sure how it will work with all the addons though
What I was saying, which you misunderstood bc you asked me why I was pinging you and said I was “wrong” when all I did is quote what you said , is that that analysis, depending on the results, would demonstrate exactly what you’re wanting to know - that having 2-buttons is superior to having 4 buttons.
I was working very hard to try to help you in what you wanted to demonstrate - superiority of 2-button use, and you attacked me for no reason first.
Maybe I just don’t want to be pinged by you. I don’t like people who say things like, clearly you want that…, I see this as narcissistic.
Edit: It’s rude to pretend as if somebody is lying when they clearly said they meant something else while saying a sentence. if I’m a liar better not talk to me.
You’re pinging lots of ppl. I was answering a question you asked when responding to someone else, that’s why you were pinged. I was indicating what your question was (when I said “want”) so it would be clear why you were being pinged, maybe you misunderstood my using “want”. You said I was lying first but I see that this is just a misunderstanding.
Don’t spew bs. When did say you were lying? I just simply asked why you were pinging me. Am I supposed to not respond to that?