Or another way to think about it, that would be an average of 306 reviews per day since I joined the site. Now, it's gotten up to there more than a couple times, but 300/day is happily an exception rather than a rule.
<br>My review count gives me a daily average of about 146.36 per day, not sure if that seems reasonable or not actually... I currently have 68 coming up in the next 24 hours<br><br>The user has a "creation_date" property (which looks like Unix epoch thing). I'll try adding an "Average reviews per day" in the output see how that looks<br><br><br>
Now that I understand it counts basically every time you’ve ever pushed “enter” during reviews, the number makes a lot more sense. Still, I’ve pushed “enter” more than a quarter million times? Jeez.
“TOTAL REVIEWS: 25816 Accuracy: 96.208%” Ouch, kinda regret that I’m
too lazy to correct typos on reviews at times like this, but that seems
about right (my avg. percentage has to be at least 90% including the
times I totally mess up my sessions).
Btw, I’d call the current state of the script a quite lovely one, if we could just solve that 32 (was it?) review riddle it’d probably be quite nice, but I’m kinda too tired to think now, I even had a hard time realizing that Naryoril’s quick calculation wasn’t incorrect and every fault actually does add “4-6 +[n_fault]” reviews until burn.
Quick Edit: Just did a 25 item review session without radicals of which I got 2 readings wrong once (each… and one was another case of laziness xD), which added 52 to the counter (25868), so I guess that can be considered correct. Idk, just felt like telling, it might help at some point.
When you have an insanely high accuracy like you do, I would say so. I bet I could look up every meaning and every reading when I do my reviews and still not get as high an accuracy.
So from what everyone has said, the only thing seriously wrong with this app is the way that wanikani counts reviews.
For example a better way of getting your actual accuracy, and not an accuracy influenced by the review straight after you got it wrong ( which if have any short term memory would be an instant easy review ). Example:
5 correct and 3 incorrect actually equals: 2 correct and 3 incorrect. Which you can see is a major difference.
So for each review you get wrong, you’ll be adding one extra review that you probably don’t want to count, In reference to the script you can do this:
var actualCorrect = iCorrectTotal-iIncorrectTotal;
var dPercentage = (actualCorrect/(actualCorrect+iIncorrectTotal)) * 100;
Hello, thanks for trying this out and thanks for posting all your results.
I did the same as RayderBlitz and compared the numbers before and after a review session
----
TOTAL REVIEWS: 88435 [+82460 -5975] Accuracy: 93.244% over 603 Days, Avg RPD 146 TOTAL REVIEWS: 88505 [+82521 -5984] Accuracy: 93.239% over 603 Days, Avg RPD 146
32 Reviews
9 Incorrect 23 Correct
Added 70 to review count
----
And looking at the Review Summary Page, it made me realise that an incorrect answer is not the same as an incorrect review. If you get an item wrong 7 times in a single review session it will only treat it a one incorrect review.
But now I am not sure how many ticks will be added to meaning_incorrect or reading_incorrect. Does it get 7 new ticks or only one?
I am going to have to wait for another review session, pick an item, check its meaning_incorrect and reading_incorrect counts, get it wrong more than once and see how those numbers change. (Unless Viet is reading this and could tell me what happens
Also, I have made a trivial update adding some new numbers to the final output. Date, how long you have been using the site, and average Reviews/Answers per day.
rade134 said...
So from what everyone has said, the only thing seriously wrong with this app is the way that wanikani counts reviews.
For example a better way of getting your actual accuracy, and not an accuracy influenced by the review straight after you got it wrong ( which if have any short term memory would be an instant easy review ). Example:
5 correct and 3 incorrect actually equals: 2 correct and 3 incorrect. Which you can see is a major difference.
So for each review you get wrong, you'll be adding one extra review that you probably don't want to count, In reference to the script you can do this:
EDIT: Although it will still be useful to know if multiple incorrect answers in a single review session cause the meaning_incorrect or reading_incorrect numbers to change once or multiple times
rade134 said...
So from what everyone has said, the only thing seriously wrong with this app is the way that wanikani counts reviews.
For example a better way of getting your actual accuracy, and not an accuracy influenced by the review straight after you got it wrong ( which if have any short term memory would be an instant easy review ). Example:
5 correct and 3 incorrect actually equals: 2 correct and 3 incorrect. Which you can see is a major difference.
So for each review you get wrong, you'll be adding one extra review that you probably don't want to count, In reference to the script you can do this:
EDIT: Although it will still be useful to know if multiple incorrect answers in a single review session cause the meaning_incorrect or reading_incorrect numbers to change once or multiple times
hoovard that is true, it would be nice: Btw I got at it and implemented my method myself, might not be 100 percent correct. But test it against your script and see the different types of results you get:
Btw I noticed on the version I just did, if you press the "review count analysis " button a couple of times you get slightly different counts every time. Can you check if this occurs on your version?
Btw I noticed on the version I just did, if you press the "review count analysis " button a couple of times you get slightly different counts every time. Can you check if this occurs on your version?
Ah, yes, I see that too.
I will disable the button after a press until the results are cleared. It was setting off multiple runs of data retrieval with multiple presses.
Also, I just did a review and compared the correct/incorrect counts before and after.
Each incorrect answer adds 1 to the meaning_incorrect or reading_incorrect values. So even though it is only one incorrect review it added 5 to each value (I deliberately got it wrong 5 times).
So I guess to only count by reviews instead of answers we have to only count meaning/reading correct value pairs (as there will only be one correct answer for each per review), but work out the accuracy using correct/incorrect ratio. Is that what your code is doing? I haven't had a chance to look at it properly yet.
Btw I noticed on the version I just did, if you press the "review count analysis " button a couple of times you get slightly different counts every time. Can you check if this occurs on your version?
Ah, yes, I see that too.
I will disable the button after a press until the results are cleared. It was setting off multiple runs of data retrieval with multiple presses.
Also, I just did a review and compared the correct/incorrect counts before and after.
Each incorrect answer adds 1 to the meaning_incorrect or reading_incorrect values. So even though it is only one incorrect review it added 5 to each value (I deliberately got it wrong 5 times).
So I guess to only count by reviews instead of answers we have to only count meaning/reading correct value pairs (as there will only be one correct answer for each per review), but work out the accuracy using correct/incorrect ratio. Is that what your code is doing? I haven't had a chance to look at it properly yet.
The assumptions on which my code is based on, is that you'll only ever get an item wrong once within the same review, so the 5 added incorrect review example you gave up above would not occur. The reason I make this assumption is because you will see the item again within 1-30 seconds so the short term memory of any reasonable person would remember with a minute error rate.
The assumptions on which my code is based on, is that you'll only ever get an item wrong once within the same review, so the 5 added incorrect review example you gave up above would not occur. The reason I make this assumption is because you will see the item again within 1-30 seconds so the short term memory of any reasonable person would remember with a minute error rate.
Not necessarily. If I get something wrong I don't always look at the answer right away. I might try to get it right more than once before looking.
I don't think you can reliably make this assumption that there would only be one wrong answer per review.
Would it be very difficult to use this (or a variation of it) to find out how many reviews you’ve done on any given day rather than as an average?
Alternatively, does anyone know if there’s something out there already that does that? I seem to recall people posting charts showing their daily number of reviews over time (but I may very well be imagining things).
