1. Forum
  2. >
  3. Topic: Duolingo
  4. >
  5. Has Duo adjusted the rate of …


Has Duo adjusted the rate of decay of the skills yet again?

For the last 3 or so months the skills on my French tree turned color at a reasonable, easily completed rate. My Spanish tree degraded so rapidly it was impossible to keep up with.

Now for the last week my french tree hasn't changed at all. I just do a couple of strengthenings and move on. Once I got the Spanish tree gold again, now it turns at a MUCH slower rate and has since been easier to complete.

I am pleased but confused. Obviously my language skills haven't dramatically improved that much. But I must say that I'm more motivated to stay gold now that it's obtainable without spending hours a day. And now still have time time to practice the reverse trees as well.

July 11, 2017



I've noticed the exact same thing this past week - in fact, I was looking at the forum to check if anyone else reported something similar. ;) For the past few months I've had the impossibly rapid degilding (although it's happened for both my trees, in my case), and now suddenly the rate of decay is much, much slower.

Since I've still seen some people complaining in the forums about the decay rate being too fast, my guess would be that this is another A/B test, where a certain group of people have the decay algorithm adjusted to be slower. I also noticed that, when I looked to see which A/B/ tests are true for me (as instructed at https://www.duolingo.com/comment/5023812), I found something called sqrt_skill_strength_experiment:true I'm going to guess that this is related to the slowed decaying I've noticed, with sqrt (square root) referring to some key part of the updated algorithm. I can't know for sure, though, of course.


That method finds some A/B tests. I think there might be many more. Going to my user page (www.duolingo.com/users/piguy3), a very significant part of it is entries that begin with "treated" (the typical word for assignment to a treatment or control group) and then parameters that sure look like tests. Doing Ctrl-f on "treated" indicates there might be as many as 549 of them in progress, in addition to the couple dozen that show up using "ab_options". Certainly, a lot more than one of these could be about the skill strength algorithms.


Oooh, if this is truly an A/B test... then I hope one day the slowed decay rate will be more common!


It seems so. My German tree used to be impossible to maintain gold, at least 15-20 skills decayed every day, but 3-4 weeks ago they started decaying much more slowly. I even had a feeling that they strengthened a bit by themselves, but I'm probably wrong about that. Since then only a few skills decay every day. I really like it, I manage to practice a bit the German tree, work on the reverse tree and I even started learning another language. If they didn't do that, I think I would have given up.


Duolingo has definitely been messing with the decay rates. My English to Spanish tree stays gold with little effort, but my Spanish to English decays at an absurdly fast rate. It is the same language! My error rates (and peeking) are the same for both courses. I've pretty much given up trying to keep up with the decay rate. If you haven't experienced it, you may not understand how bad it is. BTW, I do about 200XP a day all on my rapidly decaying tree, so it is not like I'm not putting in the time.


This is definitely true. When the tree was decaying so fast, at first I tried to keep it golden, but even with 300 XP per day I couldn't. And I didn't do that many mistakes or peaked at words, I usually had 18-20 correct answers and 0-2 peaked words. I guess it was an A/B test, which is why some people who never had such high decay rates, can't understand how it was. I know they need to do testing to improve the site and everything, but with this they are losing people, myself almost and/or partially included. I know a couple of people who gave up duolingo completely because of it and they probably weren't the only ones. I came back after giving up for a couple of weeks, but since I lost my streak, I just really don't care about it anymore and I'm using way less duolingo and way more other, more stable, resources.


I just posted his same thing before seeing your reponse. Absurd is right. As I mention, I'll be strengthening them in a single session, and one or more I just strengthened will decay again.


My Swedish tree is still decaying at an absurd rate and I find that very frustrating. As someone else said earlier, I don't care about getting the tree golden because that looks pretty, but because I need a system to help me know which words/grammar I need to practice. I used to have a paper flash card system, but this is (was) more fun. Now it is getting so frustrating that I switched to Memrise.


I noticed that with my sons trees.


For me it feels exactly the opposite. My trees decay so fast that I have almost no time to golden them all...


Frustrating. I applaud Duo's continued efforts to improve what they have, but these "adjustments" can be difficult on this end of things if you are in the group affected.

At it's worse I had sometimes more than 25 turning every day. I recommend not attempting to reguild everything when it's turning that fast. Instead concentrate on 2or 3 skills and either redo the lessons, or do the skill review until you are doing it perfectly every time. Then move on. Skills renewed in this way will stay gold longer. (Seemed to work even when my Spanish was decaying rapidly) eventually you will get that tree gold. What's important is what you're learning, not a bunch of colors.


What's important is what you're learning, not a bunch of colors.

This. A thousand times this. Even in an ideal world where Duolingo really knew what it was doing and wasn't constantly messing with people's heads by altering its outcome indicators, skill strength decay could only ever be keyed to some sort of broad-based average that might or might not be well attuned to any given user (I mean average taking account of peeking behavior, accuracy, etc). I'm always a bit amazed that people give more deference to the often-terribly-frustrating results of computer algorithms than to their own common sense.

concentrate... and either redo the lessons, or do the skill review until you are doing it perfectly every time. Then move on.

This is common sense :) [ok, I woudn't say "perfectly"; I'd say 80/90% b/c of diminishing returns]


Thing is that you need to practice and the changing color of the skills and their strength is generally a good way to tell you which skills you should practice because it's very difficult to follow which words are in which skill, when was the last time you used them and how well do you remember every particular word. But if it tells you that you need to practice every skill every day five times, it really doesn't tell you anything. This keeping track of when you should practice which skill is a big advantage of duolingo compared to other resources to learn languages like for example the Deutsche Welle course for German, and if you lose this, you lose a big part of why duolingo is so useful. So it's not about keeping the tree golden because it looks nicer like that, it's about being able to know which skills you don't know that well and should practice them.


Obviously I agree the color-change system can/could be useful (although I think the entire way Duolingo goes about spaced repetition is much less credible than Memrise, for example). But people seem oddly unable to adapt when it's self-evidently not being useful. This is the kind of putting pixel colors ahead of learning that I was referring to.

If the system here going awry is enough to tilt the balance to you toward Deutsche Welle, then props to Deutsche Welle. Duolingo losing its users there will hopefully show it that whatever change drove you away isn't one to be maintained, or expanded to greater numbers of users.


I think most people do adapt in some way (ignoring the decay of skills or just quitting duolingo and switching to something else), but I still perfectly understand the need to complain about something annoying.


I'm sure you're right, but I've read things like "Duolingo won't let me move on to new skills," well, often.


Good point, I guess those people really should learn how to adapt.


The color changing is only useful if it prompts us to practice what we need most. A rate of 20-25 changing a day is useful to no one.


I as well noticed this. A few months ago, my French skills decayed at least 4 skills a day. When I came back (about two weeks ago) and strengthened my skills to gold, my skills never needed strengthening again (as of today).


Just as I was about to post something about this! I've been checking the forums for anything about this. I'm glad the skills aren't decaying so quickly as they used to decay. :)


I've been struggling to keep English > Spanish gold. Now suddenly it's very easy. But my Spanish > English tree decays even faster than the other did before! I'd get 12-14 on a finished English to Spanish tree. Now I am getting 12-14 on a Spanish to English tree that isn't even half way done. Even the very basic first lesson or two de-gilds every day. Sometimes, in a single session, before I can get them all gold, the first lessons I strengthened have already decayed.


Are you getting words wrong, peeking, etc.? All of that can cause them to degold quickly.


Peeking doesn't explain the change in the rate of decay


Honestly I find myself doing more 'practice skills' since this has happened, and I'm satisfied to see my tree all golden.


Yea, it is too fast


I can't keep my skills golden.. I have been doing about 12-15 practice sessions a day for the past couple of weeks ( in Danish) but I'm losing a total of minimum of 11 bars per day. I can't seem to progress because I feel I'm constantly trying to catch up with the old stuff and I don't understand how anyone can keep up with that kind of decay. 11-15 bars a day seems excessive!


Let us define "progress" as "further your knowledge of Danish" as opposed to "get further down the Duolingo tree." If the review helps you progress, then perhaps it is worthwhile. At least I have usually observed myself to need much more exposure to things to have them stick than the presence or not of gold color would seem to imply. If, however, it does not help you progress in sufficient measure to time outlay, blithely ignore it.

Useful rule of thumb: if you can do a timed practice and get 17+ right with no hints, you know that skill for now - and you've demonstrated it in a much more sensible manner than relying on Duo's color choice could yield.


I'm repeating basic phrases that I have learned months ago and know quite well.. the ones that I DO need to keep practicing I can never seem to get to because the basic ones keep decaying too quickly. Especially on the days I can't get in more than a handful of practices as the decay is faster than I can keep up with. I do agree though that I need to expose myself over and over to learn something.. it doesn't come naturally to me, but this is way out to lunch.


Well, it sounds like you know which skills' degilding you'd be better off ignoring. Sounds like you'll enjoy things more (and use your time better) if you did so.


I agree it would be a useful tool if it showed you what you most need to review, and if They would quit playing with the decay rate. It's useless when the colors turn at such a fast rate.

Learn a language in just 5 minutes a day. For free.