Best Way to Make Your Tree Turn Gold and Stay Gold
For the past several months, moomingirl and I have been doing some experiments to find the best way to "regild" trees. Since I thought a lot of people might have New Year's resolutions along those lines, I carefully wrote up our conclusions:
The short, answer, though, is do three skill-specific strengthens per day. Never do a general strengthen unless all skills are already gold. Never do timed practice, and never use hints. Regild from the bottom up if you're still comfortable with the language (e.g. you just finished the tree). Regild from the top down if you're effectively starting over. How long it takes will depend on how decayed the tree is and how well you do the exercises. As little as a month. Maybe as much as three or four if all skills are at one bar.
While I was at it, I went through my older post about the best way to use Duolingo and updated it to reflect the big change that did away with hearts.
As always, I'd welcome any feedback.
Thank you, very interesting! Kudos to both you and Moomin for keeping up with that for 90+ days. It's definitely timely, because I've been ignoring some trees and I've been wanting to get back into. I have 22 decaying skills in French, 15 in German, 20 in Italian, and 7 in Russian (to English). (I'm not actively working on Spanish or Portuguese). That's nothing compared to the hundreds of words I'm behind in Memrise. :(
But if I have to do 3 exercises per language per day, that's 12 for the four trees, which is a lot for someone with a full time job, so I don't think I can swing that much. Also, I prefer not to do any new lessons at all whenever I have any decayed skills. I like having the tree gold before moving on, so my progress is quite slow. From time to time I'll do an intensive weekend where I re-gild everything (though I've never had this many), but it seems not to be worth it at all, because after a day or two I always have tons of colorful skills again. It's so de-motivating.
I'll definitely incorporate some of your tips though, like never using the general strengthens while I have ungilded skills, starting from the bottom up (I always practiced the easiest ones first), not using timed practice (I never have, so now I know it's not worth it to start), and never using the hints. I used to use the hints a lot to make sure the word was right so that I wouldn't lose a heart and have to re-do the whole lesson. I thought that would help me pass the lesson and keep the skill gold, but it was probably hurting me.
The problem with binge-reviewing is that you end up so sick of Duolingo that you don't want to touch it for days. I still think you'd get your best result if you just picked one language and let the others go until you finished that one. When that one is done (even six months from now), you could come back to one of the others and start regilding it from the top. That's a bit like starting over, but it's faster and it does take advantage of the work you did before. When it's regilded, you can start doing new lessons. When you're done, you start another one.
I feel the same way about not wanting to proceed unless the tree is gold. If you follow my suggestions, I think you'll find that's not so hard to do.
Have you thought about just focusing on one language and letting the others go for a while? I only have two incomplete trees (French and German) and that's challenge enough. I can't imagine what four would be like.
I often go through periods where I work on just one language. Right now it's Russian :)
The problem is the skills decay so fast (both duo and memrise) when I ignore the other languages and jumping back in gets harder and harder every time. Even a day off leaves me screaming, and now it's been like 6 weeks of basically just Russian. So I was thinking that for the new year I should try keeping up with all of them (not growing them, just not losing them) but it seems so impossible.
Did you complete Italian and Spanish before starting French and German? That's patience. :)
Wow, I'm just stunned about the fact that you are lvl 20 in French, but have not yet completed the tree? I have completed French and I am only lvl 14 now. I've just started German, as I want to finish a tree before starting a new one. But I now have to try and keep French in Gold too.
Congrats on the Italian!
I'm thinking maybe - if I have time over New Years - I'll try to binge-review and get everything as gold as I can, and then try do the two general strengthens a day (which is still 8 lessons so still kind of a lot). Usually whenever I do finally get it gold and I wake up the next morning and it's still gold, my first thought is "woohoo, don't have to study language X today!" and then the next day it's "oh, only 2 skills turned, let me review one and save the other for tomorrow" and then within days it's all decayed again. So maybe if I follow your advice of keeping them gold and doing the strengthens, I can at least keep the trees managable. I like your idea of durable trees, where you can on vacation (even if vacation just means working on another language) and the tree doesn't fall apart. That is really appealing.
I added it to the duowikiFAQ as this is an incredibly FAQ (can't imagine why we didn't add it earlier).
Anyway, it is interesting that you advise people to re-gild the skills from bottom to top rather than top to bottom as it would theoretically go against the Spaced repetition principles. The oldest practised ones would generally be last.
In addition, the regular practice should in theory be more effective for the same reasons. My guess is that what is good for the tree is not necessarily what is good for the memory, and the fact that most of the questions will be familiar(since they come from one skill rather than the whole tree) means that a student is generally more likely to be more successful and also cheat because they can see the words before the practice begins. I'd wager that's one of the reasons it is better.
Also, a small mistake in your article. A strong skill or word only requires 80% of "strength", 79% and it has decayed. That's why skills decay often, as most people only strengthen it until it reaches the 80% mark and then leave it alone. So the next day one or two words lose their strength and it effectively loses a bar.
Anyway, great article!
Yes, but the claim is that 2-bars on words (out of four max so 50%) equals four bars on the skills (so 80%). More accurately, you need to average 2.5 bars on the words to get 4.5 bars on the skills, which turns them gold.
I'm also curious whether the new skill bar will make a difference. I looked at the thread you linked to, but it's all a year old.
I just remembered something related to practicing and keeping skills strong. If, from the beginning you always practice until you see the image below, then your tree will always be gold. :)
P.S. That's particularly hard to achieve, I've only ever managed to see that when I am starting a course and there aren't many words to practice in the first place.
Well I have heard the 80% number quoted, but 50% is what I measured, so I'm sticking to it. :-)
Spaced repetition only applies to independent facts, but the questions Duolingo asks are highly interdependent. The skills at the very top of the tree are heavily exercised in all the skills below. The skills at the very bottom are usually rather neglected. Starting at the bottom causes you to refresh those skills with some delay. As you move up the tree, the ones further down decay and need to be refreshed so you actually do get some spaced repetition.
Therefore, redoing the top third or so of the tree has far less value than redoing the bottom part. If someone redoes the top third and then quits that's worth almost nothing. Redoing the bottom third has enormous value. If someone finishes the whole tree from the bottom up, that person will have redone many of the lower skills--getting a good bit of spaced repetition. If they did it from the top down, they'll end up with lots of practice of the basic skills (which they didn't need) plus another brief glance at the advanced ones.
The only argument for top down is for someone who is thinking of deleting the tree and starting over. It would make far more sense for such a person to just strengthen all the skills from the top down--even skills that are still gold.
Actually, the 80% is related to the bar. There are 5 bars so 4/5 of it are 80%. That's were it comes from. I do agree that bottom up makes sense since older skills are refreshed too.
Though I expect that if one diligently practices using general practice everyday the skill tree is unlikely to deteriorate unless one fails often.
Also, the new strength bar feature may make things change somewhat. So time will tell.
Just to clarify that, you only do three skill specific strengthens in total, don't try to completely regild three entire skills.
Some of the larger skills (with lots of lessons within) take more than one strengthen to bring back up to gold, especially if the strength has dropped by more than one bar. So some days you may manage to regild three skills in a row, other days you may have to do three strengthens just on the same skill.
Personally I find it much more interesting to practice different skills each day. One lesson worth of study on a single topic seems to be just the right length for me; beyond that I start to lose interest. I always start with a single general skill review each day to warm up, then practice one skill from every fourth row moving down the tree (again, to avoid too much repetition), then one of each skill unlearned at the bottom (if any). If I still feel like practicing more that day, I go to Immersion.
Fair enough. We are all different, and different strategies keep us motivated.
In your case then, do one strengthening each on three different skills. Sometimes you will regild one, sometimes not. You will still be doing three practices, and getting a bit more variety. :)
I only use the web site, so I really don't know how the apps work. For a partial tree, I advise two strengthens and one new lesson per day. (Three strengthens on a day when you do no new lessons.) It doesn't matter whether some skills have full strength or not. Hope that helps.
Thank you for your very prompt response. I think the app works in a very similar way, so I will definitely be able to adopt the strategies as you have described them. I'm already doing a lesson and recently about 3-4 strengthens a day depending on how many skills were weak. So: I'll do 2 specific/1 specific and 1 general/1 general practice a day as well as 1 lesson, change the order in which I do specific practices, avoid hints and revise my daily goal from 10 to 20 that'll increase the effectiveness of my learning. Merci beaucoup!
I think I understand. The information at the top of the page is intended for those with fully completed trees, and the reference to 'a month or two' is how long it would take for a completed tree to become 'durable gold', e.g. fully gilded and able to stay that way with minimal effort from the learner.
Thanks for your suggestions Gred! Have a lingot.
It is 2017....and your article from 2014 is still true...
Even for the new DuoLingo portal (Scala code) where the global strengthen button still chooses skills from top to down of your tree and does NOT strengthen like 5...10 or even 19 non-golden skills at the same time (I tried it with 19 non-golden skills three times in a row: only 1 skill was strengthend in a full normal session (typing) on the web interface).
With the global strengthen button I had max. 2 skills strengthend (timed practice) at once, not 5, not 10 or more...
Use the "Skill strength viewer" user script - already working for the new portal for EN source courses as of V0.2 - to check skills with the lowest percentage (word strength).
Do you guys remember the article about the Decay A/B test groups?
I was in a test group with a "slower decay" and I can only remember one time where ALL my skills decayed over night and one/two/three strengthen excercises brought them back alltogether.
The DuoLingo message at the end of the session was like "You strengthened skill X and....xx (I believe it was very high number like 59???)...more skills.
BTW: There are have been at least 2+ threads for the new DuoLingo (Scala) portal lately where skills seem to very quickly fade over night although they have been strengthend one day to gold before.... (I do not care, as Memrise always had their shorter review intervals 4-5h/12/24h....)
SO i have yet to be 100% gold. I now have all but the last section gold. I keep them there and work the last section first with timed practice. then go back and those not gold if under 80% without timed. I have rarely used the strengthen bar. Would it be wise for me to continue to strengthen the units that decay each then do a few strengthen bar lessons. I do like the timed as it makes me think quickly ( like it is spoken).As well I am not a great typist but seem to get thru most of each unit every time and on occasion with time remaining.
I think the issue with timed practice is whether you do the long sentences or not. Some people just skip those. If you're fast enough to really do everything, then I would expect it to work just as well.
It's hard to believe it's actually better, but I can see how it might seem faster, since the regular practice will generally show you the same words two or three times, even though it's 100% strengthened the first time you see it.
Many thanks for this Greg!
I'm using only timed practice to strengthen skills (I find it more efficient as it requires you to answer straight away - it's more like being in the country and having to speak without too much time to prepare your answers). It does work, my trees are (mostly) golden, but it requires a lot of work. I don't skip any questions though so I don't think your theory is true. Maybe I just give more wrong answers because of the time pressures (I typically get 14-17 / 20 on a timed practice). On one of my trees I'm now trying to do only "normal" practices to see if that slows down the decay.
I am wondering however whether the decay rates vary depending on the tree. I find the upkeep of my Dutch tree a lot easier than that of my Italian tree, the irony being that I would rate my Italian as better than my Dutch.
I don't really think the decay rates vary by tree. Spanish, Italian, and French all seem quite similar.
As for timed practice being a bit like being in the country, I don't really think it is because 80% of what you do on Duolingo is typing answers in English. Maybe if you did timed practice on the reverse tree it might be closer.
I didn't realize I was typing in English 80% of the time. Of course, DL focuses on teaching us to translate as they are able to sell our translations which in turn provides means for them to continuously improve the program. So, I am not surprised that I am better at translating than I am at speaking or understanding fluent French folks.
Surely that mimicks a foreign environment pretty well? when abroad you have to listen and understand a lot more than you have to talk.
I think the decay rate is also influenced by how much activity you do - I thought Italian decayed faster when I was working on it a lot every day, and Dutch was stable even though I only worked on it episodically. Now the situations are reversed and my Dutch is decaying fast while the Italian stays stable.
No on both counts. Timed practice is rushing you to type an English translation in a hurry. It's not hurrying you to read faster--much less listen.
Activity will adversely influence the decay rate if you get lots of words wrong when you do reviews. Or if you are doing lots of new lessons without bothering with reviews. Otherwise, more reviews means less decay. The graphs in the blog post show that beyond reasonable doubt.
Very interesting analysis. I'm a bit surprised about your views on timed practice, and am wondering what you think of my typical approach. You say:
"I think [timed practice] strengthens the tree slowly because time pressure encourages people to skip long sentences. The earlier post from Duolingo said that every word counts, so it would make sense that skipping the long sentences would really hurt you."
Personally, I never use the skip button when using timed practice -- I get 20 questions, and each one I either get right or wrong. Furthermore, I've noticed that correctly answering different questions causes a varying amount of time to be added to the clock, presumably based on the estimated time that should be taken to answer the question -- so I'm not sure that skipping questions would ever make sense.
To me, timed practice is more difficult than standard practice, and I'm not sure why, with this approach, it would be more difficult to gild the tree using timed instead of standard. Have you seen a difference between those who skip questions and those who don't with respect to gilding time?
Incidentally, there's a great script that pauses the countdown in timed practice, allowing perusal of the comments or a dictionary following an incorrect answer.
My sole concern is with people skipping questions because the sentences are too long--or simply "dying" and never finishing long sentences at all. I've talked to lots of folks who told me they did repeated timed practices but never gilded anything. They admitted they skipped long sentences. That's my data--strictly anecdotal.
If you're really managing to finish all 20 (or however many) questions in timed practice, then I agree it should be just as good. And I'm sure it helps you get to where you have an instinctive feel for the right words rather than thinking through rules. But I think few people are able to do it the way you do. Do you really regularly finish the timed practices without the timer running out?
I usually get all 20 sentences perfect, or only miss one, in time practice. I never skip anything. I usually don't even attempt timed practice until I feel confident that I can do it easily. I do it primarily as a way of improving my thinking speed in the language, so that I can think faster in conversations.
If few people are able to do it this way, maybe they haven't gotten good enough. I think DuoLingo is a tool that you can use to take you to varying levels and I've noticed that a lot of people settle for a rather weak command of the language and then feel satisfied, maybe because their tree looks full.
I think this is misguided. The game features are intended to motivate us to work harder at learning a language. They're not the end goal. If people are just trying to game the system, they're just hurting themselves. The point of DuoLingo is to learn a language. I find the timed practice a useful tool for getting to that next step of like...thinking on your feet, which is necessary if you're going to function among native speakers.
Excellent point and insights. I have felt that even though both my Italian tree and the reverse are GOLD I don't have a strong enough command to converse in the language. Duo doesn't provide conversational practice. I haven't used the clock much but I will begin to do so. Thanks
An afterthought. Focusing on things like keeping a Gold tree may be fun and challenging but it can distract from the goal of learning a new language.
With very few exceptions, yes; in fact my goal is to have 2 minutes of left over time at the end of every round of 20 questions... I don't always finish with that much time remaining but I often do. I save a couple seconds per question thanks to the script I use that stops the clock once I submit my answer and starts only after I advance to the next question, and I make extensive use of keyboard shortcuts.
I agree, it is quite difficult to get to several questions (because of all the lags when you press the buttons and evaluations).
We have to wait for an update for the "DuoLessonsFix" user script which successfully paused the timer (and I regulary used the "timed practice" to save my streak) :(
It does not work anymore on the rewritten Scala DuoLingo portal website.
"DuoLessonsFix" does NOT work anymore with the new website :-(
I think the best way to make and your tree turn gold and let it stay gold is by doing what you enjoy most. I appreciate your tips but I don't think a golden tree should be the goal. As I can read from many of the comments it can quickly become a chore. Personally, I like the timed practices best, because it trains me to quickly retrieve the information from my brains, a skill that is not trained during the regular lessons. Each his own :)
Oh I agree entirely that making the tree gold should not be an end in itself. However, there is a certain appeal to it, and it does seem that if a person goes about it the right way, it can lead to Duolingo asking them review questions that will be useful for their language learning.
I like the concept of timed practice, and if it works for you, by all means do it. I just think that the way most people use it, it ends up not helping them at all. I think most people only answer the short questions and skip the long ones. Then they just keep restarting whenever they "die" and because they get credit for each question attempted, they're racking up lots of points--but not really learning much. I tried it a time or two myself, but it seemed to give you the same amount of time for short or long questions, making the latter almost impossible to do.
Actually the whole premise of the "gamification" is to help make learning language more fun by making it more like a video game, wherein the player is motivated to accomplish goals in the game because the player is playing the game. So it is in fact intentional, I think, that they have game goals to pursue.
Of course, whatever works for the individual to enjoy & continue to study is likely their best path.
Indeed, I'm a big fan of the gamification! Just not all goals are similarly interesting to me though. I love finishing the lessons for example, but I hated losing 3 hearts. Likewise, I love making my tree gold, but I dislike keeping it gold at all times because then it starts to feel to much like work. Like I said, to each his own!
After I saw this post. I was was doing 7-10 a day. basically everyone that wasn't golden from the top down until it got to about 5 a day then bottom up. I dont see a dif on that but on occasion I would get two at once completed. now the last few weeks I have 3-5 a day. after I do those I usually go back and strengthen the areas longest since last practiced.
I have noticed that since I stopped using timed both my fluency and my ability to actually speak it has improved. Gregs method seems to work well for me. at first it seemed it didnt have an impact but once I had some consistency it did wonders. Now if I could only turn off the timer
Hi, not sure if I qualify to answer this as I have not completed any of my trees yet, but my analytical mind tells me that all you need to do is to practice the lessons where your words are the weakest. To identify these lessons, go to the 'Words' tab, click on 'Strength' column (this should re-order your words by strenght, weakest at the top), highlight the word that is weak and in the right hand side panel there should be info/link which skill corresponds to the weak word, so click on the link and redo the lesson with the weak word, voila.
Thank you very much, Greg.I really appreciate this advice, and it will change how I train -- somewhat.
However, when I choose to strengthen specific skills, then by definition I have narrowed my expectations for what the exercises will cover. For me, that will not strengthen my flexibility with the entire vocabulary, and with unexpected sentences and phrases, and it has always seemed like surprise is the biggest challenge in exchanging ideas with anyone in any language, even your native one.
General and timed practice at least have the advantage of a wider range of surprise, thus better challenging my flexibility. I don't do timed practice often, because frequently one phrase or sentence will send me off on a 1-to-20-minute long deep dive into dictionaries, phrasebooks, common usage, thesauruses (thesauri?), conjugations, message boards, Spain vs Latin American comparisons, English usage -- even recently figuring out that I've been using the English word "compare" incorrectly all these years (apparently not alone in that). Yikes.
With timed practice, I know that opportunities to do these deeper investigations escape me. However, spoken Spanish will require faster reflexes than mine are, so I try to do them regularly, too, in addition to the generic practices.
All to say, thank you, and I'll put more focus on strengthening specific skills as you suggest, but "general strengthen" and timed practice will still be how I start my day.
I just want to say, though: We all learn differently. And I have been historically extremely poor at learning languages (attempts so far: 8; successes............ next to zilch, unless you count English!). So my methodology is not necessarily recommended for anyone else any more than yours (coming from a professed, and clearly talented, linguist) is for people like me.
This is so helpful. I've been waiting for the outcome of your experiments. It all makes sense now. Well kind of... I only do specific strengthening skills. I do about six a day but I can't get my tree golden. Seventy days in so far. You're not saying I should do less?Maybe you are. I'll do three a day for a while and see what happens. I feel like three a day would only let me gold the newly 'ungold' skills. But I'll try it. Great work! Thanks!
Happy New Year to you too. Yes, when you get words wrong, Duo strengthens them anyway, but it also makes them decay faster. If you're regularly needing 25 or more answers to finish a lesson, I suspect you'll find that your skills don't stay gold. Think of it as Duo's way of saying you should focus on learning the material really well. :-)
Thanks for your question. I use app still with the hearts. So I tried the web. I just did three skills specific strengthen exercises. I had to answer 30, 19 and 29 questions. I guess it would be worth making more of an effort to be more accurate and try to bring the number closer to 17 when I can? Also, I like your blog and what you are saying about reading. Best of luck to you for the new year. Thanks again.
I use hints only occasionally if I really don't know the word. I don't use timed practise. My tree seems to have settled to gold from before the third checkpoint. But after that most skills have only one bar. So it will take a while to get them back up to full strength. Your advise supports what pink doug recommends. You have added not to delete your tree and start again. I did do that. Maybe that is my problem. Thanks again. I seem to be on the right track. I just need a big dose of patience.
I'm interested in that "never do timed practice" thing. I saw your reasoning--that it causes people to skip long sentences. But I never skip anything, and can almost always (unless something interrupts me when I'm doing the exercise) get through all 20 sentences before I run out of time, so are they actually going to be hurting my chance to get and keep my tree gold?
I haven't actually had a lot of problem with having massive numbers of skills un-gild--they're usually things I either had problems with, or that I originally completed via shortcut (those seem to un-gild themselves pretty quickly until I've practiced them a few times). So is doing the full timed practice (20 sentences, usually 17-20 correct, no skipping) a truly bad strategy for me?
If you really do the entire timed practice--no skipping and not making lots and lots of errors, then, no; it's not a bad strategy for you. For most people,though, I think it ups their error rates (which makes the trees weaken faster) and it tempts them to skip long sentences.
I don't skip, and rarely get below 18 correct (usually 19 or 20). I don't do the timed practices unless/until I'm comfortable with the material; if I'm having trouble, I do the untimed so that I can ponder/look things up/etc. I'm glad it isn't a terrible strategy, because I really enjoy the "beat the clock" aspect. (I can deal with "may be somewhat less than optimal," but not "terrible.")
useful advice for learning a new language, but will regard to simply getting back to full fluency with a language one learned years ago (as a child even), maybe it isn't so important to re-gild? Generally once i am reminded what a word in German means, it sticks with me, whereas with Turkish I need continual review.
Do you suggest trying to keep the tree gold as you go, or completing the tree first, then going back and ensuring the tree is gold? I suspect its much easier to complete the entire tree first, but that I'll have a better (if not nearly impossible) experience trying to keep the entire tree gold while moving downwards.
My approach was to keep it gold. I think that makes it much more likely that you'll actually learn the language. If you try to rush the tree, you'll reach a point where doing new lessons is extremely difficult because you have forgotten so much of the earlier material. Maybe you'll finish it anyway, but you're more likely to get frustrated and quit.
I can't believe this is a year old and I am just now seeing it. I've been on DL for 10 months. I've got Gold Owls in Italian and the reverse but have let both go mostly colored. (6 weeks out of the country with minimal internet access). I had learned the advantage of working the trees bottom up, however this does not appear to apply to the app. I have never seen the app complete other than the skill being reviewed.
I think there may be advantage to timed review once you are strong in the language. Even though you have to type it all in your native tongue It should encourage more real world processes; that is understanding a phrase/sentence in Italian without having to translate every word to English.
I want to become a reasonable speaker of the language not necessarily pass for a native but able to communicate fluently with a native. Any suggestions along that line?
I'll go read your reflections now.
Thanks and that definitely has some useful tips in it. The one thing it doesn't cover is that I believe the revision/strengthening for some trees has bugs in it and doesn't cover all the words it needs to. I have one tree (Italian from French) that de-gilds much much faster than my other ones and in some cases no amount of skill specific strengthening will get the skill gold again (I've tried doing 10 passes or so on that same skill just to check). The only solution I've found to this is to run through the entire lesson again. But even doing that this particular tree is driving me nuts, it feels absolutely impossible to keep gold (I checked it again this morning and I've had 25 skills de-gild on me in the last two days!!!). None of my other trees do this...
There is some point in strengthening skills that are already gold and it will delay them decaying. What the right mix is to complete the tree is really up to you - if you're still using hints or making mistakes other than typos the revision is probably still useful. If on the other hand you are always revising things you know perfectly I'd say drop it and move on. I've generally not had too much trouble keeping the tree gold as I go along it's just the completed Italian from French tree that is a headache.
Yikes! Yesterday, I reached 59% fluency in my French and although I practice every day and never leave a bulb ungolden, tonight almost my entire tree is un golden! SHOCKER! I have been trying to regild but it is like I will have to completely redo the course. What happened? I did nothing differently.
You can challenge yourself to get as many as you want but I have now set mine low so if I have a busy day, it will tell me I'm done - then I can decide to go back and rack up some more xps if I have the time. I have reached 60% and have noticed lately that the more accurate I am, there are less bulbs to redo the next day. I think if we are making too many mistakes then the system is triggered to slow our progress down. And I was making silly mistakes too - but DL doesn't differentiate between a slipped keystroke with a really bad error. Now I just go slower (-:
I agree. I found my typos or on occasion still SPANLISH are the killers A mispelled English word is more accepted than a mispelled Spanish word. probably because that can change the entire meaning. (example huevo hueva...etc.. One thing that frustrates is I cannot see the timer and the result at a glance so I really often dont know how much time I have left until I hear the BEEP OF DEATH!!
I had read on someone's post about keeping the tree golden - he suggested to avoid the timer. I realize that many people use it to force themselves to think faster, but for me, I became all thumbs with the pressure and was making the silliest mistakes. So I no longer use the timer and I focus more on accuracy which seems to work as far as the golden part goes. For pronunciation - I have invited a French-speaking person from France to stay with us... Hopefully that will help push me to use what I know. Of course, the dream is to go to France next year!
Please clarify: Your blog does not say "never use hints." Your blog says never to use hints (use Duolingo to check what a word means) unless you really aren't sure, on the theory that you do want to let the algorithm know where you really are, so the workload placed on you accurately represents the review needs you are likely to have. There are other online dictionaries besides Duolingo, so the point matters. Does taking a hint result in an over-compensation in the amount of review required to stay gold?
@Brrrb: Duome.eu indicates that Greg hasn't been active on Duolingo for 1736 days now, so I doubt that you'll receive a response. Which is unfortunate, as his blog on the topic was very interesting. Equally sadly, the spaced-repetition method (based on skill/word strength) that Duo used at the time, on which Greg was commenting, is no longer apparently valued or used by Duo, although it's still seems to be calculated. Personally I thought it was great, and far superior to the current Cracked Skills method.