Are "tricks" to increase Duolingo's "fluency percentage" really worth nothing? [EXPERIMENT]
Hello, Duolingo users!
So, yesterday I was kind of looked down upon for answering a question about fluency percentage, by someone who told me that all these "tricks" to increase fluency percentage on Duolingo were all simply CxxP.
So I decided to try my own tricks and see if they "really" work, while documenting the progress step by step with screenshots
These are the tricks I recommended:
- To review lessons with "dead words" until they appear
- To try not to "peek" on the answers
- To install Fabian Becker's "skill strength viewer" userscript.
I started by noting my fluency score as shown in my user data.
My "home" screen shows this number as just 52%, as seen below.
My WHOLE tree is gold, although not all of it fits the screen.
Now, my first point is that Duolingo shows skills as "gold" even without actually reaching 100% strength internally. This can not only be witnessed by seeing the "Skill strength viewer" userscript, but also by noticing how many words are "overdue" in the "words" tab that's included in Duolingo itself.
After some time practicing, the internal strength of my "family" skill in Portuguese went from 45.3% to 91.5%.
Making my fluency percentage go from 52.43% to 52.49%. Just by practicing. That alone should be proof enough to show that skills are being shown as "gold" even without having actually reached full strength, and that the userscript fulfills its purpose of helping you see their real strength.
I continued by doing the same with "verbs: imperative" and "questions".
Notice how "questions" is also gold with only 58.7% strength. "Family" had 45.3%, and "verbs imperative" 50.2%.
Now the next is a special case, because I know for a fact that my "verbs present 2" and "verbs present 3" skills have a significantly high number of "dead words" that I took the time to count myself many days before. So I'm showing the change before I start practicing them, and the result right after.
We're starting with 108 dead words, 85.6% of strength for the learned contents (so far), followed by percentages taken before and after practice. Time is shown in all screenshots.
Finally, after some time practicing (not starting any new lessons I might add), "dead words" have dropped from 108 to just 95
**FLUENCY PERCENTAGE** went from 52.43% to 52.52%.
GRANTED, the percentage did not go a whole lot up, it DID increase enough for a day. That goes to show that it's not the "fluency score" that is bugged, but the mechanism that shows
A MIRAGE, AN ILLUSION that skills are supposedly
GOLD even though they're actually as weak as 45.3% in some cases. And last but not least, that not everything is CRXP just because you go around crying
without even trying.
Author of the "Skill strength viewer" here.
This is a nicely done analysis! I just want to point out that while the script definitely helps you target weak skills it also has a weakness in that it computes the strength percentage somewhat incorrectly. It's a bit tricky to reason about how Duolingo exactly computes their skill strength but one known factor is that it groups lexemes (variations of words) and treats them as one - my user script tracks each lexeme as a different word.
I deleted my other reply, I think I now see what you mean.
Correct me if I'm wrong: what you're saying is that their script that calculates how much strength is needed for a skill to be shown "gold" is instructed to look for any lexeme of a word with good strength to consider the whole word as "learned", therefore influencing the whole "gold" bars.
If that's so, then they're contradicting themselves. In their "words" tab, any other lexeme is also counted as a different word, and even if a person does not have your userscript installed, Duolingo's percentage is still influenced if you find it when practicing.
The proof can also be seen here, given that I only /practiced skills that were already gold, and the internal percentage still went up after finding every other lexeme.
According to your 2nd screenshot, you have earned 527 XP on Sunday - in one day.
Once the tree is golden, there is no shortcut to XP, so you must've done at least 50 lessons which may have simply improved your overall fluency by the 0.092% (from 52.435% to 52.527%), regardless of the focus on skill-sets.
Yes, that is because I have already made a habit of reviewing, so to show changes I had to dig pretty much deeper than any other day. Don't you think it's logic that I already tried it before?
The point is to show that "dead words" (or "overdue" words as shown in the "words" tab) can be dug to get them to influence the fluency score.