Saturday, February 13, 2010

Supermemo History

Before I decided to use Supermemo as my primary flashcard system, I created many paper flashcards. As the number of paper flashcards increased, so did my struggle with them. Managing hundreds of cards began to feel overwhelming. Many words were forgotten. Hopelessness started to set in; next came anger, which led to the discarding of flashcards and dismissal of their relevance. "I'm better off without them," I thought to myself, only to realize that my language skills had become even more dismal. Indeed it was a very discouraging cycle. Now that Supermemo handles the scheduling, this is no longer a problem.

Recently I began to wonder, "What were the older versions of Supermemo like? Are the newer versions really better than the older ones?" Some web sites (Scroll to the bottom) mention that Supermemo has changed in how it calculates future repetitions (Relying on E-Factors, Optimization Matrices, and other complicated-sounding words). Some claim that the change was for the better, some say that it needlessly complicates things.

On the Supermemo web site, there is a history of the algorithms used for the various versions of Supermemo. One has even been adapted for paper flashcards. Both to satisfy curiosity and to (hopefully) more fully understand how Supermemo works, I am going to be involved in an ongoing experiment: I will use paper flashcards to learn completely new information (Greek, Arabic, Hebrew and Korean characters), using older Supermemo algorithms to calculate the next review date.

I plan on learning about each iteration of the Supermemo algorithm and how it calculates the next review date. Once I think I fully understand how one version of Supermemo works, I will make a post about it.

Right now I am using the paper version of Supermemo. After the fifth repetition using paper flashcards, I will put the information into my main Supermemo database.


  1. Great idea, Littlefish! I appreciate you are doing this test. Do I understand it right you are going to make real paper flashcards or you put them in some other application with correct algorithm? I believe years ago I used some spaced repetition program which allowed to input your own learning algorithm. I am looking forward to your reports.

  2. I am at the moment that I feel I have to start using Supermemo or Anki, or maybe both to see how it goes... any preferences, advice?