Hopefully I will be able to fix it tonight, really sorry it's happening to you

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

OK, I fixed the bug causing paulus_germanus's problem. Thanks for the report, let me know if anything else goes wrong.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

hum.. now it's working on Chrome.

Thank you Sir!

Now I got some questions (suggestions?, don't wanna be too impudent here regarding your teaching method:
1. You should name it the "Smith method" (or whatever your real second name is) of learning joseki, I bet it would stick to "dailyjoseki.com" as the site/method grows more popular, as it surely will - making yourself more popular among players. And I think you are fully entitled to do that: even though the methods of learning (joseki, srs) existed before dailyjoseki.com, combining two into one was your brainchild.
2. Are you planning to make similar learning engine for tesuji? (this time based maybe on goproblems.com),
3. Could you tell us how many people subscribed, and how many page-views dailyjoseki.com gets daily/weekly/since the beginning?
4. Will dailyjoseki.com be available off-line?

Lol

1. Honestly, I'm not sure if I thought of it first or not. Some of the good folks here were trying to do this with anki and static images, which sounded like a lot of work to me. I thought of this as soon as I discovered Anki, but I have no idea when that was.

2. In the future, I plan on allowing members to upload any sgf they want for inclusion in the system.

3. I haven't been aggressively marketing (yet), so I don't think I'm quite ready to make all the stats public. I will say the site has administered 4,500 tests as of last night. I'll also say that there have been enough upgraded accounts this year to pay for hosting so far (many thanks to everyone who has, you know who you are!), which is good because the set of data I'm working on will probably require a server upgrade.

4. I'm afraid that's not likely to happen unless it seems like dailyjoseki.com would allow me to quit my day job. I currently don't think the market is big enough to allow that.

Bonus: I have *really* awesome features planned. Unfortunately it's taking a long time to get the next version of my data together, and it is going to require some re-writing. I hope to at least have a demo of the new data available before (the US) go congress.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

I just wanted to give you another example of an error message I got just minutes ago. This one I got on the 4th move of a joseki, on the second joseki of the day. It could be some sort of time out message (I did the first three moves, realized I had real work to do, came back a while later), but I did have "keep me logged in" selected.

It's definitely caused by timing out plus a missing session ID. I'll try and revisit that code this weekend.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

Thanks for all the answears daniel_the_smith It seems that we're gonna get even more goodies in the future

Well, I spent the weekend fighting with ubuntu over a new monitor, and then comcast broke my internet. So, I didn't accomplish much, I'm afraid. But I haven't forgotten.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

If anyone observes the thing jts reported, could you post it here? I haven't yet fixed it, but the server should now be emailing me (with a call stack!) when it happens. Also, I've noticed a couple of you have gotten crashes in the queue page-- next time it happens I should get enough info in my email to find my logic error.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

Once more: (this one was definitely not timing out, although as best I can tell the message is identical)

Fascinating, thank you. Apparently my guess as to what is causing this was not even close. Well, I at least can see the sequence of events in the log. I may or may not be able to try and sneak a fix in before I leave for go congress on Saturday.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

Here a re a couple new suggestions.

First a couple cosmetic suggestions:

1- when some stones are added outside the joseki , numbered them. It is interesting to see the added move stand out, and it is might be useful to sees the actual sequence.
2- at the end of a joseki, you can highlight the whole joseki sequence with number: for weaker player like me the end shape is not enough to retrieve the whole seq and when i leran something i like to consider the whoel seq from start to finish. numbers would help.
You could even highlight in another color the move where an erro was made.


Something more ambitious:
First a bit of offtopic: i used to work on statistical translations engine, and we had similar data mining problems: how to automatically extract decent translations of phrases from a (huge ) corpus of human translated sentences.

It is key to have a control mechanism over the extracted data, otherwise it is really easy to miss something.
This kind of control is even more important if you allow used to upload their own sgf: you risk adding junk into your db (unless you plan that the added sgf are only for use by the one who upload them but then thats a bit of a waste: it would be cool to add then to your db)
A good way to check automatically extracted data is to compare to a human created "golden set". for automatic translations, you would compare to human translations.
Here i would suggest to compare (if possible ie if open source) to josekipedia.
For example some of your extracted sequence might be marked as "good" in josekipedia meaning "good in special cases" so indeed the surrounding stones are very important.
on the other hand "ideal" moves are usually good without a too special surrounding.

it is also possible that some moves would be marked as bad in josekipedia but are present in your data because they are super-special cases.
So adding the josekipedia flags to a joseki would be and interesting cross-check.
You can also have a wrong move due to an error in the sgf (you need 3 occurences so it shouldnt really be a problem but who knows ?)
if josekipedia data is not available, the existence of the line in a joseki dictionary is also of interest.

Also of interes would be the date of first/last appearance of a joseki in the DB:it will show new or outdated joseki?

_________________
In theory, there is no difference between theory and practice. In practice, there is.

Thanks for the comments, perceval.

Are your cosmetic suggestions intended for the study interface? Or the browse page?

I've considered trying to make a link swap arrangement with josekipedia, but I've been busy with other things and haven't pursued it.

The new data I'm working on should be much more authoritative than the data I have at the moment. One thing I eventually want to have is a rating by difficulty. I'm not sure whether it'll be best to crowdsource that or come up with some simple rules. Or maybe I could get my hands on some kgs games of various levels and scan them with my tool to see what "joseki" are popular at what levels.

What I'm doing right now:



That's my tree building program using more than 6 cores and 3gb of ram. The GoGoD database has only 68k games, but when you consider rotations et cetera, it's something like 350 million positions I have to examine to build the tree. It was a bit of a trick just getting this to run. It will use about 120gb of hard drive space while it's working.

But I'm not building the whole thing right now, I'm building part of it to test with. Ah, ok, it's done...

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

I know you guys haven't seen much new from me lately, but I really am working on the website and making decent progress.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

A few thoughts. One, it got the time-out message again a minutes ago:



Two, I've been noticing recently that I've been getting very few new patterns and I've been getting repeats of identical games pretty frequently. Did you tweak something to make this happen? I'm pretty sure that as of a month or two ago, I never noticed the same game coming up twice in a day, or even on consecutive days. But today, I'm pretty sure I saw two games come up twice (not the same joseki variation, the same game), whereas I only saw one variation that I had never seen before. -- Sorry if this complaint sounds incoherent. I just have a strong and unquantifiable sense that something has changed in the last two weeks or so, and it's not really interesting to do DJ when you associate a specific game with a specific sequence of moves.

Good news and bad news: I did manage to replicate that error! But I haven't sat down and fixed it yet. In fact I haven't done anything recently, I've been bad. Nothing's changed in the last two weeks.

I need to implement a system that "retires" positions as they get old, but I don't think that's what you're talking about. Did you check hint 3 on the game? You shouldn't be able to see the same corner of the same game in the same session unless you mess it up the first time. But there's nothing to keep it from showing you two different corners of the same game. I have one game in my queue where the pros are playing mirror go, that case is not elegantly handled...

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

A few more thoughts. But first, I want to thank you again for making DJ, it's pretty awesome and I don't leave suggestions here when I'm just pleased with it.

1. Towards the end of the browse, there's a huge drop-off where at one point it will say "A was played 246 times" and then, on the next branching of the tree, all of the options - including tenuki - add up to 10 or so. Do you know what causes this? I have a niggling worry that some of the variations I'm learning may be extremely rare.

2. The DJ banner at the top of each page is a link, and the different positions on the browse page are not links, so if I'm clicking on DJ while looking at other windows it's really easy to lose the page by clicking on the "negative space" and have no obvious way of getting back to the position I was looking at. Would it be crazy to either give each position its own URL, or disable the banner link?

#2-- good point, I hate it when sites do that, but usually I'm clicking on the side, I guess.

#1-- Do you have an example? That sounds wrong, and I suspect that what is wrong is the tenuki number. I know it is incorrect in positions reachable by either player.

_________________
That which can be destroyed by the truth should be.
--
My (sadly neglected, but not forgotten) project: http://dailyjoseki.com

I must have made a mistake. I was reviewing a game and we were talking through several different joseki starting from 3-4, high approach, high 2-space pincer, and 3-4, low approach, high tight pincer. But I re-checked, and all of the numbers add up. I think I must have, at various points, been on the "B" branch where I thought I was on the "A" branch. There were also other times where I thought the total was off by a small number (1-5); I didn't notice that sometimes the little dots say "other marked spaces played once" and other times they say "other marked spaces played twice or less".