Skills fading too fast (according to my tree) :/

I was wondering how Duo's team determines the algorithms that cause skills to fade in the tree, and if they would consider changing them in order to avoid too much user frustration.

Every day I start the morning by clicking on any skill that is not golden and strengthening it. However, this is becoming too time-consuming and I differ in opinion from what the tree is naturally doing.

Yesterday, for example, after turning my whole tree golden again, I went on and learned two new skills. That's a lot of work, considering each skill has multiple lessons. And the vocab and grammar is all recycled from previous lessons, if it is not new.

This morning, I see that there are NINE skills that have changed from gold to various other colors. That is pushing my limit of user-frustration. I want to go on and learn new materials. I don't even agree with the supposition that I need to review all that past material. I timed how long it took me to do it the re-goldening process: 23 minutes.

And it would be nice if the generic "strengthen skills" would strengthen a few of the deteriorated skills, but it doesn't seem to do anything. You have to go into each skill one by one, and strengthen it.

Any comments from the developers? Can you change these algorithms, or give a defense for the fact that so many skills deteriorate even when a person is actively putting in a lot of work on the lessons?

Thanks in advance.

January 25, 2015


Two new skills a day seems like a lot to me, so maybe that's why there is fade. Every day I do 1-2 new lessons (not skills), and strengthen 2-4 old ones. Then some immersion. I tried to do more than this at one point (ie one skill a day), but I didn't retain it and skills started fading.

Also - you probably know this, but if you "hover" the cursor over words to check meaning they un-gold that skill faster than if you do not. So I reserve that for things where I am really stuck, and lately only do it for new words and otherwise try not to do it at all.

January 25, 2015

My goal is only one lesson a day on average, but since I have less time on weekdays I do more on weekends. My stream says I learned the two skills yesterday but I think one was yesterday and the other on Friday. But still, learning more should not make previous skills deteriorate, because all the materials are recycled. On the contrary. Perhaps if there is a limit on what our brains can assimilate, at the most this should effect the new lessons and make them deteriorate quickly in the future.

You didn't say... with your system (1-2 new lessons + 2-4 strengthening skills) does your tree remain golden?

And thanks for the fact about the hover feature -- I did not know that. Nice to know!

January 25, 2015

Update: I don't know when I posted this, but today (02-28-15) I timed how long it took me to re-gild my tree. Over the week, that is Tuesday to Friday, I had only strengthened one skill a day, because I'm too busy with my work. This morning there were 12 skills in need of strengthening. It took me 43 minutes. Definitely too much, imo. Also, I figured out I've been spending on average two hours a day learning Swedish since the course came out. Fun, but also too much. I don't know why I haven't tried Greg Hullander's suggestion. Definitely starting that tomorrow. I'll do just two or three strengthen skills (picking those lowest in the tree), I won't do it timed, and I won't touch the hover feature. I'll post a response to this comment soon... see if I feel a bit more contented!

February 28, 2015

I have the impression (which may be incorrect) that scoring around 17 on the untimed practice keeps skills golden longer than getting only 3 or 4 with the timer. Certainly, it gives me more practice. So maybe the question is: how many times do you get the 'strengthen skills' sentences wrong?

January 25, 2015

On one of them I did score an 18. That is one of the reasons why I think their algorithms need adjusting. It also takes longer if you do manage to stay in the game to the end. :) But you are probably right; I say this because it used to be my habit of repeating the strengthen skills over and over again until I could consistently score 18-20 points. I have stopped doing this because others have already completed the Swedish tree whereas I'm not quite half way through it. Maybe I was learning better but I was certainly much slower (of course Idk if perhaps they are native speakers who tested out of much of the tree, but still, I do want to finish in a reasonable time)!

January 25, 2015

I know what you mean. Because I want to learn to translate, I spend a fair amount of my time doing Immersion. Then I see people who have "finished" their trees at much lower levels that I have, and it makes me feel I should spend more time learning skills, and less time translating. It's hard not to become competitive (one of the reasons I spent so much time in Immersion is that I was trying to do half as much per day as royalhalo does, but s/he is a beast! Take a look at the profile.)

January 25, 2015

Btw Majory, thanks for your comment. I didn't even know that you could make 17 points in untimed practice. I thought the max was ten, as in a lesson. I'm going to go try that right now! Edit: Yeah, in untimed you only get ten points. But if it makes the skill stay golden longer, that's great. Definitely I feel I learned something whereas if I make five or six points in timed practice I don't feel that way, even though the skill will then turn gold.

January 27, 2015

Yes, I've chatted with her. She's tier 195 in French - incredible. Others high up there are Luigi.MCMLXXIV (Italian, now has started German) and Sanlee (tier 118 in Spanish). Obviously these people really enjoy translating and spend several hours a day doing it. I'm sure there are many others. ;)

January 27, 2015

I wish I could find people who would check my translations! The numbers of upvotes in both Spanish and Portuguese have remained constant despite the fact that I do a fair amount of translation in each every day. Of course, the goal is learning the language(s), right? Not increasing translation tiers! (But both would be nice! ;)

February 28, 2015

I'd do it, but I'm waaaaay behind you! :)

I always like your translations though...they usually read very elegantly.

February 28, 2015

Looks like you have a fan club. :)

March 1, 2015

I also am moving slowly through my trees. I love translation, so I spend a lot of time reading (off DL) and doing immersion, and less time pushing forward with my tree. The one thing I have found is I cannot really substitute immersion for regular skill practice. It needs both. I don't mind moving slowly though.

I find if I let a day or two go by without doing skill development it harms me though, so now I try to do some every day. I was traveling last month and mostly just did immersion while away, which was fun, and kept up my streak, but degraded my knowledge. I'm now back to a more regular routine, and only do immersion when I finish my 2-3 practice skills and 1-2 new skills (no dessert until dinner is finished!).

February 28, 2015

The quicker you complete the skills, the more words will have a similar strength value and will therefore will weaken in a similar time frame.

The algorithm is fairly complex with the skill strength being the mean word strength of all the words in that skill. I very much doubt that Duo will change it just because some find it frustrating.

January 25, 2015

Well, from a quick look at your profile I can see you're learning a lot of languages. So it is a given that you'll end up having a lot of skills decay and become frustrated with it all. Also 23 minutes a day to re-gild the tree is not too much, especially when you divide that by the number of skills you learn. Anyway, on average a person could have at least 5 hours of language studies per week in school.

Also, there are a couple of things that will definitely make a skill tree decay faster [1]. A little known fact is that the faster you progress through the tree the faster it will decay, particularly when :

  • taking the placement test;
  • test-out; and
  • checkpoint test .
January 25, 2015

I'd rather spend the 23 minutes doing and re-doing just one skill, then I really learn it well. To re-gild nine different skills, it's enough to make a few XPs on each one, which really isn't making me learn much. I am not complaining btw about the total time I spend, that is all good fun! At university, courses are usually five hours a week combining lecture and laboratory, then they expect you to put in about two hours' study for each hour of class. So 15 hours total. If I don't do that much on DL already, once you figure in the Pimsleur and other things I do, it's much more. What I'm saying is that the process of having to strengthen so many skills daily risks a counter-reaction on my part where I might just tell myself that I don't care what color the skills are... that I can turn them all gold some day after I've finished the tree (maybe I should already start doing this already). In the end, what we're trying for is to find the best way to learn, correct? Probably my old approach was the best, but it was taking long compared to other users.

January 27, 2015

Well, there's an optimum way to keep the tree golden, see greg's post. The thing is that the tree is supposed to reflect a person's memory. The more you rush through things the easier it is to forget them. Greg's approach works simply because the emphasis is on re-gilding (and practicing) the tree rather than on advancing quickly.

If you're more interested in completing the tree, then you can go for it, and later come back to review. I do agree that every one is looking for the best approach to learn, if you find that strengthening too much is negatively affecting your learning then simply ignore it.

There are, for example, exceptional people with eidetic memory for which Duolingo's model wouldn't work well.

Nice hearing from you too, and happy new year..

January 27, 2015

Dessamator, I've tried all the languages offered here on DL but really I only speak six, and I'm only working on Swedish actively at the moment. I like Anki, where basically you can program the algorithm yourself based on your own ease of recalling vocabulary. I have heard that Duolingo also has in store some sort of more personified learning, so let's look forward to it! Btw, nice hearing from you; it's been a while! Happy New Year!

January 27, 2015

I was wondering the same thing though. A few hours after I finished one lesson it had lost a bar already. And I found older skills fading much faster than usual, too. After strenghtening the entire tree and having it golden again, a day later the exact same skills I just strengthened have started fading again. I'm a casual user, so I do around 10-30xp per day. I can see old skills fading, but I was wondering if something has changed so the tree is fading much faster than usual? I don't remember having this problem before, it started like two weeks ago.

I also didn't know about the hover feature, so thanks for pointing it out. But I am still wondering if that would make a lesson weaken after only a few hours?

January 26, 2015

Yes, thanks that is to Dcarl1. As I said earlier, I think the DL team has already got something hatching along these lines so it's important we put in our voices. Thank you for basically agreeing with me. Maybe their algorithm is really a good one, idk! I heard at least in the future they will be measuring how long it takes us to answer. You're working on Swedish too, right now, correct? I think I remember seeing your name in some of the discussions.

January 27, 2015

Yes, I am learning Swedish at the moment! :) I had been thinking of posting something along the same lines and am really happy that someone else seems to have the same issue. At first I was under the impression that it might be related to the mobile app, as the faster decay started around the time I installed the app. That might have been just a bad coincidence, though.

January 27, 2015
Learn a language in just 5 minutes a day. For free.