I'll start:

Deck Statistics
Deck created: 117 days ago
Total number of cards: 3169

Cards over a week: 2457 (77.53%)
Cards under a week: 655 (20.67%)
Unseen cards: 57 (1.80%)

Correct over a week: 94.0% (1074 of 1143)
Correct under a week: 70.2% (3910 of 5570)
Correct unseen: 61.2% (752 of 1228)

Average next time: 48 days
Average workload: 65.3 cards/day
Cards added: 27 a day, 813 a month

The 3169 cards in this deck contain:

Kanji Statisctics
1154 total unique kanji.
Jouyou: 1082 of 1945 (55.6%).
Jinmeiyou: 14 of 287 (4.9%).
58 non-jouyou kanji.
Jouyou levels:
Grade 1: 80 of 80 (100.0%).
Grade 2: 154 of 160 (96.3%).
Grade 3: 173 of 200 (86.5%).
Grade 4: 148 of 200 (74.0%).
Grade 5: 137 of 185 (74.1%).
Grade 6: 107 of 181 (59.1%).
JuniorHS: 283 of 939 (30.1%).

And from mousing over the green bar: Total Correct: 5736 out of 7941 (72.2%)

What's in the deck:

Vocab from all the sample decks merged together...a little vocab from random sources...about 100 sentences from various places...vocab not already in the sample decks from minna no nihongo II, RTK2 vocab through about frame 160.

So what? :-)) How does knowing any of the above actually help you to learn?

PS. 67 cards per day looks like a dream... iFlash forces me to review up to 1,000 cards per day on my Chinese deck... :-(

yeah, I'm just wondering how other people have been using it.

and I'm sure resolve wouldn't mind seeing usage statistics.

also, I think the stats are sorta fun and motivating, so I'm curious what others have been up to.

And ya, I think 1000 cards a day seems pretty crazy, but if more people post stats it can give perspective =)

Deck Statistics
Deck created: 182 days ago
Total number of cards: 7590

Cards over a week: 6484 (85.43%)
Cards under a week: 1072 (14.12%)
Unseen cards: 34 (0.45%)

Correct over a week: 88.8% (8364 of 9424)
Correct under a week: 82.6% (8655 of 10476)
Correct unseen: 63.8% (1836 of 2879)

Average next time: 57 days
Average workload: 132.3 cards/day
Cards added: 41 a day, 1251 a month

Kanji statistics
The 7590 cards in this deck contain:

1651 total unique kanji.
Jouyou: 1369 of 1945 (70.4%).
Jinmeiyou: 54 of 287 (18.8%).
228 non-jouyou kanji.
Jouyou levels:
Grade 1: 75 of 80 (93.8%).
Grade 2: 155 of 160 (96.9%).
Grade 3: 191 of 200 (95.5%).
Grade 4: 171 of 200 (85.5%).
Grade 5: 144 of 185 (77.8%).
Grade 6: 120 of 181 (66.3%).
JuniorHS: 513 of 939 (54.6%).

And from mousing over the green bar: Total Correct: 82.8% (18855 of 22779)

Something has to be wrong with that green bar, considering that there are only 7590 cards in the deck.

Serge Wrote:PS. 67 cards per day looks like a dream... iFlash forces me to review up to 1,000 cards per day on my Chinese deck... :-(

Whew, man, that's quite the daily workout! I thought I did a lot, but apparently, I only have 100 or so a day: I'll be thankful for that (although really, should I? Or should I be trying to work up to your level? I'll just say that I got a long ways to go).

Actually the green bar looks ok, it's the total number of reviews

Quote:Correct over a week: 88.8% (8364 of 9424)
Correct under a week: 82.6% (8655 of 10476)
Correct unseen: 63.8% (1836 of 2879)

9424 + 10746 + 2879 = 22779

Ah, okay: that makes a lot of sense.

1000 cards a day sounds nuts to me. How many cards do you have in that deck? I've found that there's a point beyond which any extra studying tends to be counter-productive, and that's somewhere around 150 cards/day.

The average workload stat is a fairly meaningless number. The actual daily workload is probably about 3 times that number.

JimmySeal Wrote:The average workload stat is a fairly meaningless number. The actual daily workload is probably about 3 times that number.

Now that you say that, I'm very much inclined to agree. While it would be great if the language could be more clear, it's not exactly important: if the card's due, it's due.

The number is a guess based on the number of cards in the deck and the average interval. Mathematics is not my specialty - suggestions on a simple and time-efficient algorithm which is more accurate are welcome.

Here are my stats. It seems like a good number of Anki users have already considerably surpassed my own deck.

Deck created: 1.0 year ago
Total number of cards: 3816

Mature cards: 3776 (98.95%)
Young cards: 26 (0.68%)
Unseen cards: 14 (0.37%)

Correct mature cards: 92.1% (5209 of 5656)
Correct young cards: 78.3% (3464 of 4423)
Correct first-seen cards: 58.0% (343 of 591)

Average next time: 186 days
Average workload: 20.5 cards/day
Cards added: 10 a day, 303 a month

The 3816 cards in this deck contain:

1250 total unique kanji.
Jouyou: 1164 of 1945 (59.8%).
Jinmeiyou: 22 of 287 (7.7%).
64 non-jouyou kanji.
Jouyou levels:
Grade 1: 80 of 80 (100.0%).
Grade 2: 160 of 160 (100.0%).
Grade 3: 182 of 200 (91.0%).
Grade 4: 156 of 200 (78.0%).
Grade 5: 147 of 185 (79.5%).
Grade 6: 119 of 181 (65.7%).
JuniorHS: 320 of 939 (34.1%).

This is my Heisig deck.

Deck created: 117 days ago
Total number of cards: 2042

Cards over a week: 1203 (58.91%)
Cards under a week: 259 (12.68%)
Unseen cards: 580 (28.40%)

Correct over a week: 95.5% (763 of 799)
Correct under a week: 88.6% (4115 of 4647)
Correct unseen: 88.0% (786 of 893)

Average next time: 35 days
Average workload: 41.3 cards/day
Cards added: 17 a day, 524 a month

Period Cards Deck%
1 day 1 0.07%
3-5 days 23 1.57%
6-9 days 110 7.52%
9-21 days 357 24.42%
21-45 days 290 19.84%
1.5-3 months 681 46.58%

When you post your stats, please also tell us what kind of cards you are adding. Are you doing sentences? Just vocab? Kanji?
I'm doing sentences, and those huge numbers of added cards each day that you guys have seem too much! I usually add only around 14-20 sentences a day, and it's still a very time consuming and VERY tiring job!
Anyway, here are my stats:

Deck created: 110日 ago
Total number of cards: 890

Cards over a week: 676 (75.96%)
Cards under a week: 199 (22.36%)
Unseen cards: 15 (1.69%)

Correct over a week: 98.1% (570 of 581)
Correct under a week: 88.3% (2619 of 2967)
Correct unseen: 70.9% (473 of 667)

Average next time: 40 days
Average workload: 22.1 cards/day
Cards added: 8 a day, 243 a month

Period Cards Deck%
1 day 25 2.86%
3-5 days 26 2.97%
6-9 days 87 9.94%
9-21 days 125 14.29%
21-45 days 225 25.71%
1.5-3 months 360 41.14%
3-6 months 27 3.09%

Kanji:

The 890 cards in this deck contain:

783 total unique kanji.
Jouyou: 752 of 1945 (38.7%).
Jinmeiyou: 7 of 287 (2.4%).
24 non-jouyou kanji.
Jouyou levels:
Grade 1: 68 of 80 (85.0%).
Grade 2: 131 of 160 (81.9%).
Grade 3: 144 of 200 (72.0%).
Grade 4: 113 of 200 (56.5%).
Grade 5: 91 of 185 (49.2%).
Grade 6: 68 of 181 (37.6%).
JuniorHS: 137 of 939 (14.6%).

Oh, and I agree that the average workload number, as it is, has little meaning.
It's a mean on all the cards, but it would be useful only if the distribution over time was constant. But that's not the case, because (at least in my case) the number of cards due in the next 4-5 days is much higher than that of longer intervals (because buttons 0,1,2 and 3 are all close in terms of time intervals).

At the moment I can only think of 2 other methods to implement this feature differently:

1) Restrict the mean only on the next 4-5 days.
This has similar problems to the present algorithm, but it is perhaps a little more accurate.

Only, I'm saying this based only on my deck, but different people may have different shapes to their "due cards" graphs, depending on the number of new cards added each day and their recall rates. So one should study many graphs before making a good algorithm.

2) Make a mean on the PAST workloads.
Every day, record the number of reviewed cards (like you do with the added cards) and make a mean on all the past such numbers.
This is probably a much better method, because it comes from the actual facts, not theoretical predictions, and it's conceptually simpler and more "clean".

The only downsides I can think of are the fact that you have to wait a few days (like at least 30-50) before having a meaningful estimate, and, at the same time, your actual average workload changes over time as you enlarge your deck. One may reach an acceptable balance by restricting the mean on a fixed number of (past) days, like 30 or so, instead of all the deck's history. It can't be too short for statistical reasons, but making it longer would mean a more "sluggish" estimate.

I hope my explanation was understandable enough!

A weighted average will mean reasonable results from day one

JimmySeal Wrote:1000 cards a day sounds nuts to me. How many cards do you have in that deck? I've found that there's a point beyond which any extra studying tends to be counter-productive, and that's somewhere around 150 cards/day.

I totally agree. But this is my Chinese deck with 6,000 characters at various stages of progress.

The way iFlash works is that it presents the same card (if remembered correctly each time) on Day 1, then on Day 2, then in 3 days, in 5, in 7, in 10, 14, 20, 35, 70, 85 and 97. (This is fully programmable, by the way, but I chose to leave the default values.) So even if you launch a relatively small number of cards into the system on a daily basis, the review load builds up due to 'maturing' reviews from the 'upper stacks'. Also, if you neglect it for a couple of days, the 'upper stacks' mature to frightening quantities. I once had around 3,000 cards awaiting the review - took me ages to work this away. The way to clear up any backlog on iFlash is to start reviewing from the topmost stacks - it puts those cards away for a long time...

Anyway, I find that with daily usage the most work goes into the 3 bottom stacks. After that you can pretty much consider the card memorised (although occasionally I make a tone mistake or something...).

Something just popped up on the AJATT blog.

http://www.alljapaneseallthetime.com/blo...e-language

Serge Wrote:The way iFlash works is that it presents the same card (if remembered correctly each time) on Day 1, then on Day 2, then in 3 days, in 5, in 7, in 10, 14, 20, 35, 70, 85 and 97....

Those interval increases are very very slow as far as spaced repetition goes. No wonder you're getting so many reviews.

is the average workload, what you've averaged in the past? Or what your average workload will be in the future?

I found the default iFlash settings to be too demanding when I used them in the past. In fact I bombed out with RTK1 the first time after 5-600 cards because the reviews were getting overwhelming. After a bit of research I settled on the standard Leitner sytem outlined at flashcard exchange.

http://www.flashcardexchange.com/docs/leitner

It's a much more efficient way to study given that I am inputting more cards than before, with less reviews and a recall rate of between 80-100%, averaging at 90%.

Really recommend that you change those default settings.

resolve Wrote:A weighted average will mean reasonable results from day one

Weighted average on what parameter? Since the number of reviews each day varies unpredictably, a statistical approach is best. For any statistical result, one needs a good number of samples. Still, I think this method would be the easiest and most accurate.

khatzumoto Wrote:What about, hold on, so, "Moo" first tone? Where's the accent on the moo?

Whoa, he can read our minds!

nac_est Wrote:Since the number of reviews each day varies unpredictably, a statistical approach is best. For any statistical result, one needs a good number of samples. Still, I think this method would be the easiest and most accurate.

The average workload will depend on the user's ability to remember facts and on that user's card adding profile. I'm not sure about everyone else, but I'm very erratic in how I add cards. I'd like to be more consistent and add the same amount at regular intervals but the reality is that sometimes I add a lot, sometimes I add a little and frequently I go for long periods without adding anything.

Any algorithm you use to give you an average workload number will have to make assumptions on these two variables. Assuming that a user's performance will remain relatively constant seems reasonable but assuming that the user adds cards in a regular fashion might not be.

You're right, wrightak. In cases like yours, where the numbers vary greatly from day to day, the average daily workload is of little use. It can't predict the future!
It is only cool when the number of reviews remains more or less stable (doesn't have to be really constant though).

My deck is the Anki-included Heisig+primitives.
http://s228.photobucket.com/albums/ee244.../Japanese/