First interval over 30 days

From SuperMemopedia
Jump to navigation Jump to search


From: Alexandar P.
Country: {{{Country}}}
Sent: Apr 6, 2015
Subject: First intervals over 30 days

Question:

I've just installed SuperMemo 2016 on a new computer with a brand new collection. All of my new items are automatically scheduled with first intervals that are quite long. It seems the average is about 20 days, but many are up to 36 or 38 days. I've read on your site that it can be normal for initial intervals to be up to four weeks, but 38 days is longer than that. I wonder if there's something wrong with the software? After starting the new collection, I never modified any settings, priorities, requested forgetting index, etc. I just started the new collection and added cards. I've done my best to look all over your site, the wiki, all over the Internet, etc., and I can't quite find an answer.

Please let me know if this is usual or if something might be wrong. According to your site, it shouldn't be more than four weeks, so I suspect something is wrong.

Hints

original answer

Most likely your collection is fine. Intervals of 38 days do not happen often, but are possible. The most frequent reason in a new collection would be a series of perfect answers to questions in their first review. If you already know the material well, you may experience this side effect.

If the above diagnosis matches your case, you do not need to do anything. Once you start going through harder material, your intervals will get shorter. However, if you still worry, please send your first forgetting curve graph for inspection using:

Tools : Statistics : Analysis : Forgetting curves : Lapses=0 : Repetition=1

new analysis needed

  • the original answer was not relevant in your case as your optimum first interval is clearly 3 days (i.e. pretty typical)
  • anther explanation might be that you use cloze deletions (which have a higher dispersion factor), or automatic memorization (where you can choose interval), or yet another method
  • to explain it all you might simply submit your collection via mail
  • your A-Factor distribution includes topics (choose items to see item A-Factors distr)

Answer:

Notes

  • the name SuperMemo 2016 is not correct and might mean SuperMemo 2006 or SuperMemo 16. However, the answer is the same for those two version as the spaced repetition algorithm has not changed much in the last 10 years (see also: Algorithm SM-17)
  • the value of four weeks is just a rough estimate of the upper limit that might change from version to version (the limit is artificial as perfect recall, in theory, should result in infinite intervals)
  • in SuperMemo 16, maximum O-Factor is 20, which means that the maximum optimum interval is 20, but random dispersion may result in much longer intervals. The dispersion factor varies a lot across versions. Also the parameters of the normal distribution behind the dispersion vary. 38 days should be rare, but possible

21:36, 13 April 2015 (MDT)21:36, 13 April 2015 (MDT)21:36, 13 April 2015 (MDT)Daijumbo (talk) 21:36, 13 April 2015 (MDT)21:36, 13 April 2015 (MDT)21:36, 13 April 2015 (MDT)Daijumbo (talk)

Response by Alexandar to the Answer/Notes to his original question

--> "* the name SuperMemo 2016 is not correct and might mean SuperMemo 2006 or SuperMemo 16." 
 // Yes, I meant SuperMemo 2016
--> "* the value of four weeks is just a rough estimate of the upper limit that might change from version to version (the limit is artificial as perfect recall, in theory, should result in infinite intervals)" 
 // I suppose I was referring to the latest SuperMemo 2016. In the page on the algorithm, which I would assume to be current and so dealing with 2016's algorithm, it says, seemingly nonarbitrarily, "Depending on the forgetting index, the length of the first interval may range from 1 to 20 days, and is not set arbitrarily." Btw, when I used "nonarbitrarily," it wasn't somehow a comment on your "arbitrarily." Just a coincidence, I suppose. Anyway it looks to me like it was 1 to 20 days. Maybe I'm missing something, but 38 days seems a bit past 20. 
--> You've said, "Intervals of 38 days do not happen often, but are possible"--which sort of sounds to me like, "That's not *im*possible, but...."
--> Lastly: "38 days should be rare, but possible." The thing is, I was getting intervals into the 30s and 30s to the extent that I noticed it being strange all within the first 5 or 6 days  of my deck (I'm on day 10 today). In only a few days, this rare event is happening in a not-so-rare way.

I'm sorry if I've made things more confusing than they should, and I have a hard time grasping how these things work. I'm learning as I go along, but for now I appreciate your patience with me.

21:45, 13 April 2015 (MDT)Daijumbo (talk) 21:45, 13 April 2015 (MDT)

Another observation: Wrong A-Factor distribution in my collection

Another question: How can auto-memorized items produce long intervals?

Picture: Forgetting curve

This picture shows that your optimum interval in the said collection is closer to 3 days:

Retrieved from "http://supermemopedia.com/index.php?title=First_interval_over_30_days&oldid=19701"