Proposed changes to the points system and leaderboards

The point system is currently dependent on a sentence’s % Mastered, or in other words when it comes up for review. You get more points for each time you answer a sentence correctly up to a maximum, and then you get the maximum points each time you answer that sentences correctly from then on. This works great and makes sense with the default review intervals (1 day, 10 days, 30 days 180 days), since the idea is you should be rewarded for getting and keeping that cloze word / sentence in your long-term memory.

The ability to customize the review intervals, however, means some users are essentially playing a different game and they can score more points faster. :rocket: We like the ability to customize the review intervals, and agree it’s useful - everyone learns at different speeds and has different goals. But we do want to keep the game fun and competitive. :video_game:

So here are some changes we’re considering:

  1. Once you get a sentence to 100% Mastered and score the maximum number of points (16 for multiple choice, 32 for text input), each time you answer that sentence correctly after that to keep it at 100% Mastered, you score fewer points until you hit a minimum number of points. So in other words, assuming you answer the sentences correctly each time and you’re playing text input, you’d receive 8 points, then 16, then 24, then 32 (100% Mastered :tada:), 24 (still 100% Mastered), 16, 8, 8, 8, etc. (all still 100% Mastered). The rationale here is that players with shorter review intervals who see a 100% Mastered sentence more often won’t be rewarded as heavily, but the game is still the same in getting to 100% Mastered.
  2. More leaderboards - we plan to add leaderboards for number of sentences played, number of sentences mastered, number of new sentences played, number of multiple choice, number of text input, etc. We’d also like to add a way to choose a default leaderboard you’d like to see when you open the leaderboard from your dashboard. The idea here is we’ll be able to compete across more metrics, and having more stats is more fun anyway. :chart_with_upwards_trend:
  3. No more All-Time leaderboards. I expect this one might the most controversial. :slight_smile: Here’s the rationale - given the points change described above, it would be more difficult for new and existing users to rank on the All-Time leaderboard. Clozemaster has also changed a lot since we add the All-Time leaderboard - users who were initially at the top with the default review intervals have been surpassed by users who opt for shorter customized review intervals which doesn’t seem fair, and as we continue to put out more features/changes/updates this will be true for users currently ranking as well. The Weekly and Monthly leaderboards offer a fairer and more accurate ranking. We could also consider replacing the All-Time leaderboard with a 6-Month or Yearly leaderboard.

Phew! That was a lot. :books: We’re always working to improve Clozemaster and we’re always open to feedback, so please let us know if you have thoughts or questions on the changes, or if you think something else might more useful. Thanks!

7 Likes

I think those changes are great, EXCEPT the all-time leaderboard. Please don’t do away with those.

3 Likes

Thank you for addressing the point-scoring issue. :slightly_smiling_face:

4 Likes

These seem like thoughtful solutions, and you appear to grasp the central crux - stats should be fun :grin:
Yes, probably a lot of pushback re: All-timers, and that’s understandable because in a way that’s the most fun of all. I tend not to bother checking in with the weekly board, but to see myself slowly beetling onwards in a big-picture sense - this I find quite heartening, and would be sorry to lose it.
Is there maybe some way to retroactively downgrade the 100% repeats? Then we are all playing the same game. I know, some will probably think it’s like asking billionaires to actually help homeless people… but shouldn’t they?
Or - just change the game and see where things go from here

2 Likes

Leader boards! Starting to sound too much like Duo where the l/boards have taken over from learning. As long as we have the choice to compete or not I’m happy.

7 Likes

I second this. Without the all time leaderboard I will lose motivation. And I hope for point 1 and 2 to be realised asap, so the 2 “1k phrases 20m points cheaters” stop farming points.

4 Likes

I like all the ideas and they’re really thoughtful. I see your point with the all-time leaderboards. For the new user who cares about leaderboards for ranking purposes, which could be a fair number of them, I could see how they would get discouraged pretty quickly when they realize they’re at a disadvantage. Realizing that users cheat can sour the mood on its own, so I guess you’re tackling the problem from both ends this way.

One thing I like about the leaderboards, especially with the smaller languages, is that you get to see who the dedicated users are. And since there’s not a lot of interaction with users on the site (not necessarily a bad thing), you get some kind of sense of community when you recognize the same users on different boards. It’s just a nice creature comfort. So I suppose getting to cut the leaderboards different ways like you say would only enhance that and could be a nice trade-off for the retirement of the all-time leaderboards. (And ftr I would be sad to see them go!)

5 Likes

Great choices! Thank you. My question on #1 : will that take care of collections? Is typing the same collection word is as meaningful and weighs the same as a new word? It could be a gaming feature that takes away from the real learning. On #2 : more stats is interesting, you can see where your learning pattern fits with the others. # 3 : Leaderboards could be toggleable . One could focus on his or her own goals , and if they are curious they can check it out or forget it if they feel like it. My own motivation does not come from the leaderboard but only from the love of the language and I hope I am not the only one who thinks that way.

3 Likes

On the topic of collections, it seems to me that if you add Clozemaster sentences to a new collection, your sentence played count goes up even if you haven’t played the sentence in the new collection. That probably shouldn’t happen if we’re going to have sentences played leaderboards. Ideally, it shouldn’t count the same sentence twice even after you’ve played it in the new collection.

3 Likes

Personally, I’m happy with the proposed changes. Removing the All-Time leaderboards will be painfully for a few very dedicated learners, but one has to consider what’s best overall for all the learners using Clozemaster. Oh, and yes. Annual leader boards, I want that.

3 Likes

Yes, good point. After the big update in May, I ended up with a LOT of duplicate sentences. Japanese was the worst, where I got duplicate collections as well.

1 Like

I agree with the general principle, but I’m a little concerned that having a question eternally at 8 points may be a bit of a disincentive to review it. Making 32’s RARER would certainly help with the dopamine hit factor, though, which would be a good thing from both the learning perspective and the “gaming the system” perspective.

This is one of my concerns with Favourites where the universal 2 points made me feel “why bother?” (Actually I’m also grateful for that since it prompted me to discover the far more effective custom collections, but more on that later.) Possibly having the scores “regenerate” over time (after x months since the last review it comes back up to 16, after x+n months it comes back to 24, etc) may more accurately reflect the educational value of the question, especially if you can winnow out over time the questions which don’t need long term reviews using the Ignore option as I discussed here.

One other thing to consider too; if you get a question wrong, you drop back to zero. It might therefore be possible for a “player” to milk the system by “blowing up” their questions once they drop back to 8.

I like the general idea but I’m not crazy about these two:

I think that multiple choice has a perception of being “the easy way”, but its usage isn’t always a choice of the player. I’m aware of some members who use multiple choice ONLY because of physical disabilities which make typing difficult. Other members use multiple choice because they are learning a completely different language, often where the alphabet is alien to them (and incompatible with their keyboard). In my own case I’ll sometimes go to multiple choice where I cannot for the life of me understand what Backfeed Betty and her cohort have said. The worst example of this that I can think of in Italian is “Un generale è un ufficiale militare di alto rango”, with the last word being pronounced more like “ramma”. The first time I encountered that I was on the phone app where trying to figure out the word by flip-flopping between apps and translating is tedious so I just hit the multiple choice button. It’s not statistically interesting that I do that on the odd occasion, just as it’s meaningless that someone with advanced arthritis (say) is pretty much forced into doing it regularly. But I do indeed agree in principle that more stats are better.

I’m not wild about complete abolition either. For good or for bad, it’s part of the site’s history. However I recall posting something quite a while back about “de-emphasising” it; shoving it somewhere where players need to drill down a bit to see it AND expanding it way beyond the top 10; top 100 at least, top 1000 even better. As I also mentioned way back when, I do like the idea of a yearly board but I still think that a “rolling” board (updated, say, weekly rather than an annual snapshot) would be a better way to distribute “a moment in the sun” since it would emphasise consistent effort rather than a few big pushes. (Though that may be less of a problem with some form of point 1 being implemented.)

I do have a dog in this fight since I’m moving further and further toward collections (both custom and built in) and away from FFT, for example. In German, where there are fewer “standard” collections than in Italian, I can see custom collections making up the bulk of my learning eventually.

I understand the concern that they can be used for “gaming” the system (though Mike’s first option should slow a lot of that down), but yes, I would hope that collection words carry the same weight both in scores and sentences played (which I’ll come back to).

Most of my custom collections are “topic” based (medical, clothes, food, transport, cars, roads, money, the workplace) or theme based (modal verbs, verbs of doing like fare, agire, etc). I try to emphasise language that I would actually need for the most part. There are two key reasons for this; first, by emphasising practical language that I can see myself using it’s more likely to stick, and second because there will be examples of verbs and nouns that come from a common base in the one collection, and IMHO it’s far easier to remember things that connect to other things. It’s the connections that we remember far more than the individual words. When you play say 100 sentences, and 10 of them have verbs and nouns descended from the same root, and all of them relate to the same subject, I see way better retention than, say, FFT where 100 unconnected words are thrown at me.

The “most common” collections are second best at this to my mind, because they (or at least the earlier ones) tend to emphasise the language that you will come across in practical use of the language.

However I see custom collections as being the single most potent weapon in the Clozemaster arsenal, and would rather that they be valued accordingly.

Yes and no. That will happen if you have the “Ready to Review” box checked in the sentence addition dialog (which it is by default). I’d guess that the assumption is that if it’s on your review list then it’s “in play”. If you uncheck that box then your sentences played number doesn’t change until you actually play that sentence in your collection. (I’ve tested this to confirm.) The design makes sense from the point of view of anyone who is using the program to learn, but I agree that if a sentences played leaderboard is to become a thing then this is a potential exploit that needs to be addressed.

Yyyyeah… BUT it depends on how you create your collections. If you are just using the “Add to Collection” button that pops up when you select the cloze, then I agree that you’re just playing the same sentence whether it be in… wherever you got it from, FFT, Most Common, or wherever, or whether it be in your custom collection.

HOWEVER, that’s not the way I do it, and it’s probably not the way that most “power users” (for want of a better term) of custom collections do it. If I come across an interesting word that I want to drill into my brain, regardless of whether it’s on CM or wherever, I’ll typically go to the Add Sentences option of the relevant collection and do wildcard searches for any verb forms of the word, any noun forms, any adjective forms whatever, as appropriate. I may end up adding half a dozen to a dozen sentences from those searches… and the sentences concerned may be in collections that I have yet to go anywhere near. They could be ones that I’ve yet to see in FFT, they could be in the 10000 Most Common words that could be a year away, etc.

The key point is that the presence of sentences in a custom collection doesn’t automatically imply that the sentence is also being played in another collection. It could well be that the only place that you’re playing it is in the Custom collection. (And that doesn’t even factor in custom created sentences.)

I would therefore rather that any sentences that are being played in a custom collection be counted in the Sentences Played count… with at least a proviso that if sentences are deleted from a custom collection, or the collection itself is deleted, the number drops accordingly.


Edit: Unless the system can tell that it’s the same sentence, regardless of where it’s being played, in which case counting it only once would be ideal. I’ve always suspected that when you add a question to the custom collection it adds a new row to the relevant table, but if I’m wrong about that it may make life easier.

(Yes, I know that this doesn’t get around the possibility people cheating by creating custom collections that they play once and never touch again, but that problem already exists even without custom collections. Someone who is after a huge sentences played count could set their “Max reviews per round when playing new sentences” value to 0 then play through FFT and every other standard collection to get every sentence to 25% mastered. I can’t think of a way around that, but if anyone else can…)

2 Likes

I think this is a really good way to use the site, after you’re past a certain level. It’s what I did in French after the big revamp in May that took away 20,000 of my played and mastered sentences. Once you have a good grasp of most of the vocabulary, I don’t think there’s a lot of value in reviewing most sentences three times to get them to 100% mastered. I did still find there to be value in hearing new sentences once to make sure I understood them, and moving difficult new words to a “difficult” collection that I did review.

I get far more value from the first time I hear a new sentence, than subsequent times when I can often simply remember the translation.

2 Likes

It’s funny you should say that, because after I posted that I started to ask myself whether that would really be cheating at all; it may well be a legitimate way of learning, just as you describe.

1 Like

I definitely don’t think it’s cheating - it means you’re playing for minimum points!

2 Likes

Except… we need to start thinking in terms of the proposed “new world” paradigm of “sentences played” as well, not just points! :thinking:

1 Like

I still can’t see how playing new sentences - assuming they’re legitimate sentences from the Clozemaster collection or other meaningful sentences you’ve uploaded - can be considered cheating, whether or not someone elects to review them. I only see a problem if someone creates multiple collections of the same sentences for the purpose of boosting their sentence count, or uploads very simple sentences for the same reason.

I felt I learned a lot more once I turned to a “breadth” approach for tackling the available resources here at Clozemaster rather than a “depth” approach (in a language I already had basic competency in, and with the proviso that I’m here primarily for aural comprehension and not to be able to write perfectly).

2 Likes

If their objective is to brute force their way up a “sentences played” leader board, it would be pretty easy to do if they just play every sentence in every collection using multiple choice. No learning required. I’d rate that as being cheating every bit as much as any other form that already exists for the points based leaderboards.

1 Like

First, it’s not that easy to do. It’s certainly far less easy than playing the same sentences day after day for 32 points. There are over 110,000 sentences in French from English and I assume a similar amount in the other large language pairings.

Second, being able to get the right multi-choice answer does still take some learning!

2 Likes

I think the proposed changes are good, and great comments have been made above.

I second Floria7’s words though: points, scores, leaderboards are mostly vanity; this isn’t learning languages. For a bunch of users it’s helpful by motivating them because it leverages an addicting effect, but for me and some other users I believe, it’s mostly a distraction.

As such, I would appreciate it if the layout of the main page was kept as sober as it is currently, with all the additional leaderboards/stats features you might add available in a separate view, not directly from the homepage.

6 Likes