Username: Leebo
Total correct: 18,407
Total incorrect: 544
Total reviews:18,951
Correct percentage: 97.13%
Hmm, I’m satisfied with that.
Total correct: 21,299
Total incorrect: 7,242
Total reviews:28,541
Correct percentage: 74.63%
oh…
Username: michaeltran9393
Total correct: 44,442
Total incorrect: 3,627
Total reviews:48,069
Correct percentage: 92.45%
Holy shit I’ve done a lot of reviews.
Total Reviews: 160061
Correct: 153954
Incorrect: 6107
Accuracy: 96.18%
Total Reviews: 173,049
Correct: 156,933
Incorrect: 16,116
Accuracy: 90.69%
dogboydog said... My thinking was that the minimum between meaning_correct and reading_correct is how many times you got both correct, whereas the maximum of the two incorrect counts was the number of times you got either one wrong, thereby downgrading the item.Thinking about this more... I think the best you can do is calculate a range that the true "both-correct" percentage falls into.
For example, suppose you have:
meaning correct: 8
meaning incorrect: 4
reading correct: 8
reading incorrect: 4
You could have either of these combinations:
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: oooooooo - - - - (8 correct - 67.78%)
Both: oooooooo - - - - (8 correct - 67.78%)
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: - - - - oooooooo (8 correct - 67.78%)
Both: - - - - oooo - - - - (4 correct - 33.33%)
If you want a fright, try working out how long you’ve actually spent on the site to do those reviews. Using an average of 10 secs per review (I have no idea if that’s accurate but it’ll have to do), I get over…650 hours!!!
rfindley said...Oh, hmm. That's true. I didn't think of that. I guess that makes all of these percentages potentially pretty off. I guess the total reviews should still be accurate though. I'm trying to think of if we have any way to determine where in the range is most likely based on how the item has progressed...guess it doesn't matter.dogboydog said... My thinking was that the minimum between meaning_correct and reading_correct is how many times you got both correct, whereas the maximum of the two incorrect counts was the number of times you got either one wrong, thereby downgrading the item.Thinking about this more... I think the best you can do is calculate a range that the true "both-correct" percentage falls into.
For example, suppose you have:
meaning correct: 8
meaning incorrect: 4
reading correct: 8
reading incorrect: 4
You could have either of these combinations:
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: oooooooo - - - - (8 correct - 67.78%)
Both: oooooooo - - - - (8 correct - 67.78%)
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: - - - - oooooooo (8 correct - 67.78%)
Both: - - - - oooo - - - - (4 correct - 33.33%)
I get this error every time.
Error: Server Error
The server encountered an error and could not complete your request.
Please try again in 30 seconds.
dogboydog said...I bet you could infer some things based on level info, e.g., if an item is level 6 with eight reviews total, that's probably 7 correct (to level 8), and one wrong (down to level 6). That's complicated with the pre-guru status wrong answers only going down one level, so maybe that doesn't work after all, but it's a possibility.rfindley said...Oh, hmm. That's true. I didn't think of that. I guess that makes all of these percentages potentially pretty off. I guess the total reviews should still be accurate though.dogboydog said... My thinking was that the minimum between meaning_correct and reading_correct is how many times you got both correct, whereas the maximum of the two incorrect counts was the number of times you got either one wrong, thereby downgrading the item.Thinking about this more... I think the best you can do is calculate a range that the true "both-correct" percentage falls into.
For example, suppose you have:
meaning correct: 8
meaning incorrect: 4
reading correct: 8
reading incorrect: 4
You could have either of these combinations:
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: oooooooo - - - - (8 correct - 67.78%)
Both: oooooooo - - - - (8 correct - 67.78%)
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: - - - - oooooooo (8 correct - 67.78%)
Both: - - - - oooo - - - - (4 correct - 33.33%)
oruam said...Make sure you are pasting your API key in the URL (e.g. api_key=123456efef...) . That's how it knows what account to look at.I get this error every time.
Error: Server Error
The server encountered an error and could not complete your request.
Please try again in 30 seconds.
Combined
Total Reviews: 80549
Total Reviews: 80549
Correct: 77840
Incorrect: 2709
Accuracy: 96.64%
Can't believe I've done 80549 reviews
teeth7 said...I was just thinking of similar stuff, what extra information we have and how we could use it. I guess you could rule out certain combinations based on the current level though, if not absolutely determine the correct one.dogboydog said...I bet you could infer some things based on level info, e.g., if an item is level 6 with eight reviews total, that's probably 7 correct (to level 8), and one wrong (down to level 6). That's complicated with the pre-guru status wrong answers only going down one level, so maybe that doesn't work after all, but it's a possibility.rfindley said...Oh, hmm. That's true. I didn't think of that. I guess that makes all of these percentages potentially pretty off. I guess the total reviews should still be accurate though.dogboydog said... My thinking was that the minimum between meaning_correct and reading_correct is how many times you got both correct, whereas the maximum of the two incorrect counts was the number of times you got either one wrong, thereby downgrading the item.Thinking about this more... I think the best you can do is calculate a range that the true "both-correct" percentage falls into.
For example, suppose you have:
meaning correct: 8
meaning incorrect: 4
reading correct: 8
reading incorrect: 4
You could have either of these combinations:
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: oooooooo - - - - (8 correct - 67.78%)
Both: oooooooo - - - - (8 correct - 67.78%)
Meaning: oooooooo - - - - (8 correct - 67.78%)
Reading: - - - - oooooooo (8 correct - 67.78%)
Both: - - - - oooo - - - - (4 correct - 33.33%)
This is awesome. RFindleys numbers on mobile get kind of squashed. It first looked like I had done 160 billion reviews.
Reviews:161991828134480
Correct:153391749232831
Incorrect:8607891649
Accuracy:94.69%95.68%95.22%
This is really great, thanks to all who put this together. It’s interesting to see that I’m a good bit more likely to get the meaning wrong than the reading. rfindley, I’m assuming you’re counting radicals as a meaning? Is it easy to add an option to filter those out?
Stagrid said... This is awesome. RFindleys numbers on mobile get kind of squashed. It first looked like I had done 160 billion reviews.Let me know if it’s still squashed on Mobile. I added some dynamic size adjustment that hopefully works for small screens.Reviews:161991828134480
Correct:153391749232831
Incorrect:8607891649
Accuracy:94.69%95.68%95.22%
ffUQilNfBXcvoFRe said…This is really great, thanks to all who put this together. It’s interesting to see that I’m a good bit more likely to get the meaning wrong than the reading. rfindley, I’m assuming you’re counting radicals as a meaning? Is it easy to add an option to filter those out?[Added:] There’s now a breakdown into Radicals, Kanji, and Vocab.
It doesn’t include individual sums by item type, but I didn’t want to turn it into a data overload. Maybe I’ll add that to the Items tab eventually.
Reviews: 20616
Correct: 19051
Incorrect: 1565
Percentage: 92.41%
rfindley said...Oh, that’s slick. I like how you color code fields like they would be on WK.Stagrid said... This is awesome. RFindleys numbers on mobile get kind of squashed. It first looked like I had done 160 billion reviews.Let me know if it’s still squashed on Mobile. I added some dynamic size adjustment that hopefully works for small screens.Reviews:161991828134480
Correct:153391749232831
Incorrect:8607891649
Accuracy:94.69%95.68%95.22%
ffUQilNfBXcvoFRe said…This is really great, thanks to all who put this together. It’s interesting to see that I’m a good bit more likely to get the meaning wrong than the reading. rfindley, I’m assuming you’re counting radicals as a meaning? Is it easy to add an option to filter those out?[Added:] There’s now a breakdown into Radicals, Kanji, and Vocab.
It doesn’t include individual sums by item type, but I didn’t want to turn it into a data overload. Maybe I’ll add that to the Items tab eventually.