Hello Oisín, Thank you for your feedback.
> Have you done any statistical analysis to see whether this is true? No, just my personal experience. > but that's not good enough without actually performing experiments and > verifying one way or another. You are right, this is not good enough! > but if not, being so sure based simply on an assumption or feeling is not a good enough reason > to say "the algorithm should be changed" Keep reading... > "why do you think you will be LESS likely to answer it correctly after another 50 days?" For a very simple reason. In 50 days the grade of the card went from 5 down to 2. If starting from 5, in 50 days I almost forgot it, so what will I get if I start from grade 2 and increase the time to 70 days? It seems a bit logical for me, and Peter gave credit to it when he said "It is quite possible that what you propose makes more sense". > According to the material I've read on spaced repetition systems, > "about to forget the card" is the OPTIMUM time to review it, so that > the knowledge moves into your long-term memory quickest. I'm not an expert about SRS, I just read an interview from Wozniak and I think this picture is very helpful to understand the concept: [image: How Supermemo Works] I totally agree with you that the best time to review the card is when you are "about to forget". I don't think this is on grade 2 or 3 tough. I always thought it was on grade 4. I will explain you why: from that picture, the best time to review the card is when the chance of remembering it is 90%. So it's more likely to be on grade 4. Speaking of probability, when you say a chance of an event occur is 90%, it means that, **on average**, for every 10 cards, you will remember 9 and forget 1, and it sounds pretty reasonable. Probably grade 2 might refer to 50% of chances to remember, and grade 3, maybe 70%. Don't know about then, but 90% is absolutely more likely to be grade 4. I might be totally wrong, as I said, I'm not an expert, it's just some thoughts and reflections. I just think that, if I'm working on a card for maybe one year, I would rather decrease the interval if I'm grading it 2, then to increase the interval and risking forgetting it. I just want to add another thought I had yesterday about this subject: I think that depending on the subject you are studying, your performance will depend on how familiar you are with that subject. For instance, if you are starting to study Japanese, you will probably forget cards very often. But if you already master Chinese and 2 other oriental language, you will learn Japanese faster. I've been watching this list for several months, and very often I see people complaining of the increased interval when they grade it 2. They don't feel comfortable with it because they said they eventually forget those cards. Thinking about it, I would like to suggest a feature for 2.0. It would be a check box at the preference window saying "Pessimistic mode". The help text for this option would be something like this: "With this option checked, Mnemosyne will show you the cards more often then it normally does. Select it if you are not familiar with the subject you are studying or if you feel you are forgetting many card very often." If pessimistic mode is on, the multiplying factors should be: Grade 2: 0.7 Grade 3: 0.9 Grade 4: 1.2 Grade 5: like normal mode I think this option won't clutter the interface, is easy to implement, the underlying mechanism will stay transparent to the end-user, and will help a lot of people. I don't know what would be the best multiplying factors. I just think grade 2 and 3 should decrease on pessimistic mode. What do you think Peter? Best Regards, Frank On Fri, Feb 20, 2009 at 12:14 PM, Oisín Mac Fhearaí <[email protected]>wrote: > > 2009/2/19 Francisco José Fiuza Lima Júnior <[email protected]>: > > It doesn't make that much sense now. > > > > Let's say the interval for a card is 50 days, and I grade it 2 because I > got > > it barely correct. > > Will it be better to: > > > > 1) increase the interval to 60 days and let me forget it on the next > time, > > > > 2) or decrease it to 40 days and give it a grade 3 next time because I > > didn't leave it for too long? > > > > I think the option 2 is better, because if the interval always increases, > > I'll eventually forget it one day, and then, I'll have to start all over > > like a fresh card. Is it how it works? I know that when I forget a card, > it > > starts all over from interval of 1, 2 , 3 days... > > You're making an assumption here that you will eventually forget the > card by grading it 2 under the current scheme. Have you done any > statistical analysis to see whether this is true? > > The big problem I see with your conclusion is that you SUCCESSFULLY > answered the card, even if it was difficult. If you successfully > reviewed the card once after 50 days even with difficulty, why do you > think you will be LESS likely to answer it correctly after another 50 > days? This is the basis of spaced repetition systems! > > > It doesn't seem reasonable to increase the interval if I'm not > comfortable > > with the card... > > In my opinion, the algorithm should find an ideal interval for a card, > > increasing or decreasing it so that you give it a grade 4. > > That's what it already attempts to do. You shouldn't declare the > algorithm incorrect on this issue without providing some proof. > > > Another example. I grade one card 5 and the next interval was set to 100 > > days. When that time comes, I was about to forget the card and grade it > 2. > > Why do you think that if in 100 days a grade 5 card dropped to 2, > increasing > > it to 130 days will do any good? I'm 99% sure that after those 130 days > I'll > > forget the card... > > According to the material I've read on spaced repetition systems, > "about to forget the card" is the OPTIMUM time to review it, so that > the knowledge moves into your long-term memory quickest. > > You may be "99% sure" that you will forget the card after 130 days, > but that's not good enough without actually performing experiments and > verifying one way or another. If this has really happened to you many > times, you should document it carefully - but if not, being so sure > based simply on an assumption or feeling is not a good enough reason > to say "the algorithm should be changed". > > Oisín > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/mnemosyne-proj-users?hl=en -~----------~----~----~----~------~----~------~--~---
