Well, after studying last night I was a bit bored, so decided to work out the percentages of my daily kanji tests. I always right down the fraction from the summary so this wasn't too difficult. Of course, being a bit of a saddo/geek I put the data into Excel. I also date the page when I do a test, so I put that info in as well.
I was pleased to discover that my correct percentage was pretty consistant at around 85%. Especially as my tests date back when I was just starting to really ramp up my daily kanji learning. This is shown in the total number of reviews per day which is all expired cards, new cards and failed cards from the previous day. I always do them all.
Anyways, with the influx of new cards daily, obviously my expired kanji started to increase, but, my correct percentage seems to have remained stable.
Here's a lovely graph I made to show this

See, the blue line is going up while the red one just wanders about 85%. Proof for me that Heisig plus Leitner is a winning combination.
I kinda wish I had the time everyday to study more kanji and see if the percentage starts to suffer at all, but I've only about 500 left and a full-time job!!

Very nice graph there, interesting job!
It seems like the days of lower percentage are often the ones with a higher number of reviews. This may mean that after a while the brain gets tired and starts failing more and more.
I like this kind of things

It doesn't look quite so nice now. I've had a few times where I haven't had a chance to review for a day or two, so there's sudden spikes where the number goes really high for one day when I catch up. The percentage stays around 80 on the whole though. Now just to finish those last three lessons..

Very nice. These are the types of graphs I'd love to see implemented in this site so we don't have to record and plot our own data. How cool would it be if we could see a plot of total number of cards versus time and watch the curve grow from 0 to 2042 or 3007. We'd see when we got lazy and what kind of overall progress we're making. Unfortunately I'm sure that data doesn't exist but would be cool to implement for the future.

Very nice indeed: great job on the percentage. Someone else had rigged up a spreadsheet with a donut graph that was both quite sexy and hunger-inducing.

EDIT: My "very nice" just happened to correspond with radical_tyro's by pure coincidence: I like.

A part of me would like to see what my own graph would look like, but a part of me knows that I really, really, really don't want to see it.

I'm sure that Fabrice has all the statistics from the site buried somewhere in a vault: whether or not those statistics go back far enough to count is the real question. I recall a thread where he showed which kanji were recalled the most and which ones the least.

[ファブリス]

If it shows something over time then the database would have to store a log or "history", which isn't the case at present.

Well, then, I stand painfully corrected.

Most days were 60+ kanji. Many were 100+! One day was 180! Ack! That's a lot of kanji to review! But congrats on your 80-90% recall rate, Mighty_Matt!

One question: Of the failed kanji, are they wrong because of:

(a) entire story or primitive completely forgotten
(b) recalled a certain primitive just fine but forgot to write it down
© story and primitives remembered perfectly but clumsily missed a stroke or two

If they're (b) or ©, I don't think they should count towards your failure rate.

My failures are almost always due to

(d) seeing the keyword doesn't yield recall of the story, but the story is vivid and clear once I get the right hook.

This seems to happen most with vague keywords or keywords that have similar meanings. Any advice?

radical_tyro Wrote:My failures are almost always due to

(d) seeing the keyword doesn't yield recall of the story, but the story is vivid and clear once I get the right hook.

This seems to happen most with vague keywords or keywords that have similar meanings. Any advice?

Don't know how long you spend on each review. But over the last couple of days I've stopped rushing. Previously I would fail cards that I couldn't get within say 5 seconds. But since Monday, I've been sitting there trying harder to recall my stories for sometimes up to a couple of minutes.

Often this is coz the keyword -> story link isn't that strong. But rather than failing these cards, taking longer over review (then recalling them) might help solidify the linkage. At least that's what I hope.

p.s. @ radical_tyro - don't know where you are up to now but I found your stories for the first 1000 or so characters to be really helpful

akrodha Wrote:Most days were 60+ kanji. Many were 100+! One day was 180! Ack! That's a lot of kanji to review! But congrats on your 80-90% recall rate, Mighty_Matt!

One question: Of the failed kanji, are they wrong because of:

(a) entire story or primitive completely forgotten
(b) recalled a certain primitive just fine but forgot to write it down
© story and primitives remembered perfectly but clumsily missed a stroke or two

If they're (b) or ©, I don't think they should count towards your failure rate.

When I review I use a book of graph paper and try and write them down in the squares (right>left, top>bottom - for practice). I fail myself for practically anything. If it is a fail, I scribble it out and re-write it correctly in red ink in the next square. This gives me an immediate visual idea of how I'm doing.
So, I guess I'm pretty harsh on myself, but most of my mistakes are of the bigger variety like forgetting a primative, or putting something in the wrong place etc. I don't see why I should aim for anything less than perfect in each kanji. Although I accept I'll never know them all 100%.

And, what else would I do in my lunch break at school? Work...?