Announcing plans for a new weekly-updated global leaderboard
Hello. The aim of this account is to replicate the function of the now inactive score_board and -Leaderboard- accounts. This means that I'll follow a large number of users and make a post with the ranking of the top X number of users in the weekly, monthly, and all-time categories.
Having noted the difficulties the operators of these accounts faced with regard to the amount of work required to maintain them, I plan to circumvent the majority of that work by considerably restricting the number of users I follow. A user is eligible to be followed if they either:
- Have a streak of at least 7 days. or
- Have an all-time total of 10,000XP.
(These exact criteria may be tweaked in the future.)
Most crucially, users who rank below 200th place at the end of the month will be unfollowed.
The vast majority of users interested in a global leaderboard are likely most interested in competing with each other within timeframes of weeks and months, and I may choose to omit the inclusion of all-time record keeping altogether.
If you meet these criteria and would like me to follow you, follow me first. If I have followed you and you do not wish to be included, simply block this account.
If you have any feedback, I would love to hear it. Thank you.
EDIT: By the way, I don't know if this was obvious, but you don't need to continue following this account once I follow you if you want to be followed. The instruction to follow this account doesn't imply that mutual followership is a requirement to be included.
I think it's too bad that DuoLingo doesn't have its own statistics section: I for one would love to know, for example, how common or rare a level 25 in any given language is, how many users have finished the skill tree, or how many users have more than 100,000 XP in total.
I assume that there is a good reason why Duolingo doesn't want us to know the exact figures: It's actually quite rare that somebody catches level 25 or climbed up his tree successfully and that would be pretty bad publicity. Let's assume that, say, only 0,3 % of all users ever complete one language tree - would that be a good statistic to be published? Obviously not and that's where my assumption comes from. There are maybe tens of thousands of really committed learners, but in comparison to the total amount of more than 100,000,000 registered users, the real figures would probably surprise us negatively.
not everyone would have to be in the statistics. It could be possible to do things like: out of all users who reached level 10, X% completed the the tree or got to level 25. So people who just try out a couple of lessons would be excluded. Or you could exclude everyone who has been inactive for more than a given period of time.
Indeed, I kind of suspect the same thing. I do think that a lot of people just do a few lessons for fun or out of curiosity, but I doubt that even 1% of those learning any giving language actually make it far enough to even understand basic conversation.
DuoLingo has in the past boasted about having a higher than average percentage of users who visit the site regularly, so it's not like they're not aware of the figures.
You can find threads detailing these exact figures if you search terms like "hall of fame" in the Discussion section. The problem is that none of these topics remain active, because the projects are ultimately too ambitious and the people who run them burn out.
Once I've finished scripting the procedure for generating a scoreboard, the bulk of the work for this project will be finished. The time needed to update the rankings using this method should only be a matter of a few minutes per week.
Yes, I've seen some "Unofficial Hall of ..." threads, but they of course only count in those who responded to them first. I'd like to see DuoLingo have an automated hall of fame where you can see where you stand. Then again, this might stimulate people to 'cheat', for example with nonsense translation in order to gain quick XP.
Anyway, I'm looking forward to your scoreboard! :)
(That has nothing to do with open source.)
Right. My plan from the beginning has always been to write a small script to extract and enumerate the usernames and scores from the page source on my home page.
My point was just that you can't simply access Duolingo's source code directly to obtain that information. You need to go about it indirectly, such as by operating an account like this, to generate an already compiled ordered list.
As of right now, there are 199 users I follow who have nonzero monthly totals. It should be stressed that my data isn't capable of representing Duolingo's user base objectively by any means at this point. Currently, it only represents a subset of power-users I've found through archived halls of fame, and those who've shown interest in this particular thread.
Regardless, a cursory glance at the rankings indicates that your XP score this week places you somewhere in the 60th percentile, or in other words, slightly above average.
Based on what I can see already, these users are a statistical anomaly. The top six users I follow constitute a plateau of +10000 XP weekly scores. Below them, the XP totals drop dramatically and even out around 2500 or 3000 XP. These scores are an entirely manageable achievement if you dedicate an hour a day to timed practice.
If I find the time, I might be able to graph this data to demonstrate some of the abnormalities I'm witnessing casually surveying these data. They're kind of interesting.
I think you might be overestimating the difficulty of immersion practice. With your language levels, you should be decently prepared to translate basic documents. In my experience, immersion was the most effective method of learning, full-stop, so pushing yourself past the learning curve is a huge tipping point that "ordinary" users are entirely capable of.
Thanks, but I think you misunderstood what I meant. My choice of words clearly wasn't the best. By "ordinary" I meant "normal", not "commonplace". Supposedly, I am C2 in German, Spanish and English and would thus have no problem amassing great amounts of XP through immersion; in fact, I already did so as I started to learn Italian and French. I stopped because I didn't like it. I guess everyone is different, but I found immersion to be very ineffective when you're learning a language from scratch, giving you an illusion of knowing things you really don't know. And once you've achieved a level where you can read, I find it more fruitful to read difficult texts with the help of a dictionary. Luckily, I am in a position where I actually need to read such texts for my studies.
This is not to say that immersion is necessarily a bad thing, I totally understand that it can be useful. The point I was after is that it is ridiculously easy to get XP while using it. If you are translating stuff to your mother tongue from a language that you already know fairly well, you can obtain hundreds of XP points without putting in a fifth of the effort of someone who is climbing a tree that is entirely new to them. Ideally, at least in my view, a leaderboard shouldn't measure who is the fastest in leveling up from translation level 99 to 100, but who is putting in the most blood, sweat and tears. And this is why I think immersion XP skews the ranking.
I agree, it's an unfair metric if we consider it to be an evaluation of the total effort spent by any individual in a finite amount of time. However, I don't think the work done by translators is necessarily overvalued, either. The ability to translate well is a consequence of blood, sweat, and tears shed in the past, so I don't know why that shouldn't count for something. You may as well argue that timed practice is unfair as well, since it's a much quicker method of earning XP that is also reliant on past knowledge of your language.
And it's true, most rookie translators do write bad translations. The Immersion program has a corrective system in place for this, however, and from my own experience, having my translations edited by other users made me a better translator.and Spanish-speaker overall.
Like I said before: the number of users who bankroll Immersion is less than ten, by my count. I'll be sure to post rankings which are inclusive of the petit bourgeois of ordinary power-users as well.
I agree that it would be extremely hard to find a perfectly balanced XP system. Yet I would still argue that the immersion XP system is completely disproportional, at least in its current form. Taking your example: through timed practice you can amass a maximum of 20 XP points in about two minutes, whereas immersion can give you 50 XP for "translating" a list of people's names. A lot of people just copy what the DuoBot suggests. This is obviously not the case as regards the people I am following, who are all excellent translators. But even if you are doing legit, good translations of easy, short sentences, you will probably be able to get a lot more than 20 XP points in two minutes, at least if you're decent at typing. My Italian translations were often really bad, yet I still managed to get weekly averages in the thousands. Used the right way, immersion is potentially a great feature, but the XP points you get from it are inflated.
Anyhow, thank you for doing this. It will be interesting to see what the ranking looks like.
Trust me, as someone who doesn't have access to Immersion (unless I do it on the English for Dutch speakers course), I'd like Immersion xp not to be counted too.
But it'd be unfair to the people who are able to do Immersion and do it legitimately. If Immersion xp didn't count, then they'd have to choose between actually learning with Immersion (and not getting any leaderboard xp) and just doing timed practice. While I'd like to have benefits over the people with Immersion for once, it'd still be unfair. :P
The leaderboards appeared on my (desktop) account about a week ago. I discovered yesterday that by going to the privacy page on my account settings, I could make my performance private and exempt myself from the leaderboards. I have done so. Perhaps it is the old fogey in me (I took my first computer course in 1970), but I don't understand this obsession with piling up points and language flags. I do not want to go down that rabbit hole. At some point in every language I've finished here, I just bought a textbook and started learning from that. Duolingo always gave me a great start, but at some point in each language, the website ceases to be the most efficient way to spend learning time.