@duome
Hello Maxime,
I know as a Java developer myself that development time is valuable and if code changes are too big or would require quite some effort for a re-engineering that requests probably can't be easily fulfilled....that's dev business.
Often you can't realize more simply what some end users or business analysts are dreaming about.
So may I ask you again if some new sort order can more easily achieved in ranking tables like www.duome.eu/en/pt without you having to put into endless of hours of your spare time?
I still would like to see that the Vocab column gets dropped from the sort order.
We're still having the bigger issue that the L1-L5 + L6 level (incl. golden tree flag for 100% strength) is not counted in with more priority.
Some people are higher in the list even with L1 or L2 levels.
A user with L5 is currently placed second, then a few L2+L3 users follow and then once again L4 and L5 trees are somewhere in the middle.
The user Danikalifornia is placed on the first spot but only has the L3 crowns level but others with L4 and L5 (or golden trees) follow later.
Also the ordering by XP (directly after levels and crowns) is not working in the intended way.
First spot only has 69005 XP while the user Redwinger on the second spot has 298102 XP.
If you simply drop the Vocab column from the DB SQL statement then this will get much better, hopefully.
- Users with e.g. a golden L3 tree (100%) shall be higher ranked than users on a normal L3 crown level but without having achieved 100% strength (if this is even possible on a path).
- L6 Legendary comes first, then L5 golden tree, L5, only then L4-L1...
Your algorithm "Sorted by skills » lexemes » vocabulary » crowns » xp" which is highlighted in the text description does NOT 100% count that in (as you probably have no ordering by the level column).
The 100% strengh golden thing (for ranking these completed trees a bit higher) would be missing.
This concept of "Sorted by skills » lexemes » vocabulary » level (incl. 100% golden strength) » crowns » xp" should work better for our ranking tables, especially in long established volunteer courses (like French from German).
For the last column "skills" the +2 or +3 bonus skills would still need to be dropped so everyone with a 91 Portuguese skills tree or 129 French skills (from German) is treated the same way.
See my previous comments and suggestions for a detail description.
..(...)..
Personally I would put people with a COMPLETED tree/path first.
In the https://duome.eu/en/fr or https://duome.eu/en/es ranking table (CEFR tree, ultra long, too many constant updates, path changes back and forth) or the two English from PT/ES courses you can clearly see that several users are included in the first page(s) that have no deep coloured language flag which means that their course (new tree, with higher skills) is NOT completed, not yet.
Some of them seem to have achieved a higher progress completion (>90%).
The skill/lesson tree data is a bit strange tracked as so many different path versions now seem to all exist with only minor differences (306, 303, 301, 300, 299, 282, 281, 280, 278, 277, 275, 274, 272,...).
Not sure what to do if the course completion is >90% progress (finishing "soon").
My main idea in the above comments was to move incomplete trees/paths out from the main table and to flag it with some /incomplete Url or similar very different language table.
Where is the benefit in adding all of the different trees and incomplete courses at the top of the www.duome.eu/en/fr ranking table?
In that case there is not any encouragment for users with L6 Legendary, L5 golden trees, L4+L5 trees (completed) much further down in the ranking table if newer + incomplete trees (even with very low progression or low XPs) get ranked much higher in the main competition table.
The big issue for those CEFR courses seems to be that the skills/lesson data is not different enough, only has small changes.
But with several rows below each other so your default "Sorted by skills » lexemes » vocabulary » level (incl. 100% golden strength) » crowns » xp" sort algorithm does not work that great anymore (only having 2-3 to 4 A/B trees in parallel like it was in 2018/19 - 2021/22 before all this path mess started).
Even if you move all not-finished courses to a new /en/fr/incomplete table it will create some mess because the level and crowns columns don't come first with the Sql order statement.
Thinking more about it I would say that the main ranking table /en/fr needs to drop the first skills and lexemes columns from the SQL DB sort statement, to honour some serious work much better for competitions between Duome / Duolingo users.
Someone with more crowns shall come first and the level (+ golden tree for 100%...which hardly will be achievable nowdays with those ultra-long CEFR courses) is the leader column.
If someone has a newer tree but only completed it on L1-L3 crowns and other learners have maybe a few less skills (on a slightly older tree or parallel tree/path with less skills/units) but have achieved a higher crown level - which is still duoable on some old Android app versions or with the Fiddler website workaround - then we should honour their long-term work more.
Again, is this for FINISHED courses (trees or paths) to make the competition between users more accurate.
For the other /incomplete ranking table you can make the level / progress column the leader column if you want and only then to use the skills/lexemes column.
No issue to additionally show the VOCAB column, only for display.
But I definitely would remove it from the SQL DB "order by" statement.
..(...)..
If somebody wants to use any of those Duome target language ranking tables with a very strict focus on skills + lexemes
(+vocab) maybe you need to put it in a new /en/fr/new(incomplete)trees table.
There you don't have to care so much anymore about any achieved level, progress, XPs and for how long a user has been learning or how much work was put into it....only the newest course data shall be placed at the top.
Only showing the "most cutting-edge A/B" tree versions (order by skills/lexemes), even if a learner has freshly joined Duolingo as a NEW user.
Big issue right now is that /new_trees and /incomplete trees can either mean low XP vs higher XP, low progress vs >90% progress:
Some OLD users may have finally received a new tree (with high skills), not everything was RESET to L0 by the DL system, they may have been given the chance to maintain several of their L3+ crown levels.
Only difference is that their language flag is now NOT fully coloured but transparent (they lost their golden owl!), as after their tree/path update not ALL changed/restructured/added or moved skills/lessons (path units) are 100% finished on L1-L3 crowns...but leaving enough L1+ crowns for them on so many L1+ skill crown levels that they still have quite some impressive tree progress (e.g >90%, even 70-85% progress is pretty good with a ultra-long CEFR course where many users will drop out much earlier compared to much shorter courses with 69-129 max skills in a tree).
Q: What do you think about my idea to use maybe 3 tables per language?
If we only have two (complete + incomplete) ranking tables - per language - with the suggested code modifications that info about finding the "latest cutting-edge" tree/path versions on Duolingo (to know if it makes sense to create new user test accounts for those who get easily bored and have finished 1-2 or 3 years ago) doesn't work that great.
At the moment the focus shall be on a separate third table (order by skills/lexemes/vocab like you do it right now) which surely is still some interesting data to see if we need to actively search for it. But the current implemention definitely doesn't honour the crowns, (long-time) progress, work put into, how many years they have been learning their target language (XPs)...to make it to the TOP of your ranking tables - as a competition.
Just writing down my ideas and impressions for some generic brainstorming.
Hope the produced English in this longer comment is good enough to be able to follow it. Sounds a bit rounder to me..
I think I have edited my 3rd or 4th review of the same text with some typo fixes and improved text passages