Bug in Practice scoring

Here is a list of words for which I receive zero experience points when I give correct answers for them. Most, if not all, of them come up more frequently than other words during each practice session:

そこ
体操
信じる
応じる
にきび
俗流
点滴
手前
招待状


I can't figure out any pattern, but I thought you might be able to. It might depend only on the information in your lessons database, but it might also depend on information in my user records.
posted by fuonk

Comments 7

  • mog86uk
    mog86uk
    You receive 0 EXP for correctly answering any question that you've already answered correctly 29 times.

    This is not a bug. This limit was introduced intentionally, roughly 4 years ago (this site is 7 years old). I think the original purpose of the 29 answer limit was to prevent users repeatedly answering only one question through exploiting a bug. That bug no longer even exists, and I don't know whether anyone even exploited it, yet this limit is still in place penalising ordinary users.

    Anyone who has learnt all the questions and kept their Oblivion list empty for two and a half years should no longer be able to earn any EXP at all from Practice/Lessons... I get 0 EXP for EVERY level 1-6 question and also for some of the lv7 questions. If I made my own list of 0 EXP questions like you've done, it would be around 3,500 items long. ^^;

    The reason for your list having those specific questions is because you have answered them incorrectly in the past. If you answer the majority of your questions correctly but answer one or two questions wrong, then this causes Practice to spam showing those questions you got wrong.

    Luckily, Beeant made some improvements to the encounter algorithm a few days ago. This should slow down how quickly you end up with more of these 0 EXP questions. And now Beeant is looking at introducing an SRS system, this could help even more at slowing down the pace of 0 EXP questions.

    Also luckily for you, since you made your account before this 29 answer limit came upon the site, you may have earned EXP that hasn't counted towards this 29 answer limit (or answered questions back when they awarded higher EXP per question). So you (like Medyrius in rank 1) will likely be able to earn a higher maximum EXP than those of us who joined the site in 2012 onwards.

    In short: There is no bug in Practice scoring. Nothing to worry about...
  • mog86uk
    mog86uk
    Also, this 29 correct answer limit applies to all questions on the site (not just the ones in Practice). It applies to Practice, Vocab test, Kanji test, Readings quiz, Notes quiz, Hiragana test, and Katakana test.
  • fuonk
    fuonk
    mog86ukさん,
    Thanks for the historical information about this problem. I still consider it a problem. The only reason I have answered questions about 体操 correctly more than 30 times is that I was asked questions about it well over 100 times in the course of two days of practice sessions. The fact that I score 0 each time now is appropriate, but the fact that I am asked questions about 体操 so frequently (as well as most of the other words on that list) is a bug, in my opinion. You can describe it as "the encounter algorithm is working very badly", if you want, but I would simply say "the encounter algorithm has bugs"-- that is, its behavior is extremely different than any behavior one could reasonably want or expect from it.
    Once some kind of spaced repetition algorithm is implemented, this bug will no longer be relevant, so it may not be worth finding and correcting it. It does, however, underline the need for implementing a spaced repetition algorithm; the "encounter algorithm", whatever its conception may have been, simply is not functioning reasonably.
    I've become addicted to the site's practice function as a kind of game the last few days; I hope some memorized information will be reinforced as a result, but I expect that I will soon get tired of it, and return to lang-8.com and readthekanji.com, where I do my real practicing; at Lang-8 by having text conversations in Japanese, and at readthekanji.com by reading words in the context of sentences and typing in their pronunciations in hiragana. Both of those activities are in my opinion more useful for my learning. Multiple choice questions are easy to implement, but not nearly as useful. I will not be around long enough to exploit my unfair advantage concerning the maximum number of experience points I can earn here.^^
  • mog86uk
    mog86uk
    You wrote in the title that there is a bug in the "scoring", not in the encounter algorithm. I wasn't saying that there're no bugs in the encounter algorithm.

    I made a test account in February -- mog_0xpjanai. http://japaneseclass.jp/user/statistics/64230

    Within 6 days of making that test account I reached "241 correct : 1 wrong" on a question. See the evidence in the crazy screenshot in the following post, which very clearly demonstrates how bad the encounter algorithm problem was:
    http://japaneseclass.jp/feeds/comment/6005

    The reason it was so bad on that account is because I answered almost everything perfectly except that one question, so therefore the encounter algorithm went absolutely nuts with spamming that question endlessly. That question reached 328 times correct before I stopped using that account, and I only used that account for little more than two weeks... So I can quite agree with your thoughts on the Practice encounter algorthim. Hopefully the recent algorithm adjustments has already dealt with this problem a bit.

    As for rtk.com and lang-8, I've used both before and they're both very effective systems. However you have to pay $5 a month to use rtk.com, and it doesn't really have the same competitiveness as JCJP and lacks proper community feeling. If every user on JCJP was donating $5 a month, this site would be more slick and polished with much more content. Lang-8 isn't something I'd use regularly as it feels tedious trying to think of something to write about, but of course it is an extremely useful site. JCJP is a very different style of site and focusses on different things to those two sites. This site is more casual and more focussed on community and fun game-like competition.
  • fuonk
    fuonk
    mog86ukさんへ。
    I was lucky enough to be one of the first users at readthekanji.com, before it became a pay site, and my account was "grandfathered in". Even though the monthly fee is modest (unlike iknow.com, when they made it a pay site), I'm not sure I would continue to use readthekanji.com if I had to pay for it.
    I'm glad to hear that japaneseclass.jp has a good community; I miss the community we had at studyjapanese.org, which the creator unfortunately sold to another website which was only interested in making money. (If the owner of studyjapanese.org had asked the users first, we probably would have chipped in and bought it from him instead.) One of the nice things at lang-8.com is that it develops little pockets of community-- it has far too many users for the entire site to function as a single community. When I write entries, I usually write about grammar and usage issues in English, trying to make them of use to the Japanese and Chinese users I interact with there. I do far more correcting than entry-writing, though.
    Is the community here centered on competition, though? The nice thing about the community at studyjapanese.org was that it was centered on sharing what we were learning and helping each other out.
    I did originally report the problem with those vocabulary items to beeantさん as a problem with the repetition, but as he or she didn't respond directly to that part of my message, I thought the issue of the scoring might make it seem more like an issue that needed to be addressed. I didn't know, of course, that the 0 scores after 29 repetitions were by design.
  • kevinlame
  • tonnyken
    tonnyken
    thank for share
    http://slitherio-o.com
fuonk

Share