As suspected... the area around and above 20k on OGS is a surprisingly tough crowd

Interesting to note from that table, that your predicted OGS rank actually gets *worse* as you progress between Oota 11.5k and 10.5k

_________________
Confucius in the Analects says "even playing go is better than eating chips in front of tv all day." -- kivi

There seems to be a big penalty for a wrong answer. I answered the first 16 problems correctly and got a rating upper limit of 4.2 dan. I then answered the next problem wrong and the rating upper limit dropped to 1.3 kyu

It depends on the problem. If you answer wrong on a problem almost every dan player answers correctly then it reasons you are probably not a dan player. The fact that it tries to give you problems that it estimates you have 80% chance of answering correctly, means if you fail to answer correctly it is often easy to draw conclusion you are weaker as the problem was easy. Yet if you answer it correctly, it doesn't prove much about your strength, as the problem was easy. There is variance involved, some problems are more important to get right than others, but on average you need to answer 4/5 questions correctly to maintain rank.

Also early on, when it has fewer answers from you to draw its conclusions about your rank, every answer will affect your rank that much more. With more answers it will fluctuate less.

There has been a bug for several days that messed up all rank calculations. Ranks effectively got stuck, showing some fluctuation but almost no real change. Should be fixed now.

This illustrates the random walk nature of interactive testing before you have a good enough estimate of the skill of the test taker. As I said before, for fuseki skill I think you need at least 25 problems before you start matching the level of the problem to the level of the user.

Now, it is possible that the problem hit a blind spot for Mitsun. Maybe he did give a kyu level response. But in that case I would expect Mitsun to say, OIC, I goofed, but he did not say that. More likely, he gave a dan level response that was slightly suboptimal, according to the latest thinking. Even more likely, he gave an acceptable response that is not currently popular. I have more faith in Mitsun than in the test.

But let's say that he did get that one problem badly wrong. And, assuming that it was a high level problem, given that he had a 16-0 record at that point, then it is absurd to drop his estimated level so much. After all, he is now batting 0.941. (This also illustrates the problem of presenting problems that you expect the user will get right 80% of the time. Getting them right 94% of the time is not all that good a result.) If now, because of the estimated upper limit, Mitsun gets problems that an SDK is expected to get right 80% of the time, how long does it take before he gets an appropriate rating?

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

First of, the whole issue was most likely caused by the bug I fixed a few hours ago. That aside, there is a normal distribution prior for the rank. The lower the sample size the more the prior ends up weighing in the outcome. Hence first few answers regardless if they are right or wrong wont budge the rank too much.

Hi,

I register on this forum just to contact you "the Ootakamoku's owner" =D
I'm using your site a lot since i discovered it around a month ago and i think it's great to learn fuseki =D

But if i register today this is also because of the life and deaths problems you've add

I hope that in your data you said for each problem if it was a fuseki problem or a life and death one because it would make my proposition a lot easier to apply ^^.

I think it would be great to be able to choose what you want to play between fuski problems or life and death, i believe that you can just ask a question at the beginning to ask to your visitors what do they wants to do : only fuseki problems, only life and death problems or 50/50 of each.

I'm doing this proposition for 2 main reasons :
- the first one is that today i played only life and death problems when i came to your site to play a mix ( i really dont like life and death but i need to improve and in the other hand, i enjoy a lot the fuseki problems ) so to allow your visitor to choose what he wants to do would alow him to have a batter expectation to what he is going to find on your website.
- the second one is that, everyone is stronger in a part of the game and weaker in another, to split the category can allow you to give more adapted problems. For exemple, i play around 3-4k in the fuseki according to your site when i think i play above 10k on life and death problems, and i believe this is the same for lots of other people.

SO this was my proposition, i'm sry if i made some huge mistakes in that message, i'm french ( i think i don't have to say more, everyone knows that french people are bad at languages =D )

Thanks a lot once more for your site =d

This is an interesting point, as it could quite arguably be better to tenuki than kill something smallish in gote.

How do I know if I'm logged in or not? Where's the login/logout link?
There's nothing to indicate if I'm still me, of if I'm now a guest.

I had a significant count for a while, then the count suddenly started over. I didn't realize I had been logged out until I followed a link to a problem I thought was wrong and was suddenly presented with another sign in link. Now I'm back to my original count/rank.

In the sidebar-(or under, depending if on desktop, tablet, portrait or landscape) there is a "Hello, username." If it is you, ok, if not, there's a sign in link

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

no, there's not.
just an infinite scrolling text of "you have done X problems", where X, at one point just started over.

Well, when you start the first line is this, then it starts being an infinite scrolling text :/

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

you're assuming I've closed and reopened my browser window
I never close my browser tabs/windows (nor restart my computer).

Haha I see! Since I only use oota's page in my iPad (after checking it works fine here) I open and close the tab every once in a while.

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

When you go to a website, it sets up a session that will eventually timeout without further action. So if you leave it running, say, overnight, your session may expire in that time. If that happens, it will create a new session, but it won't have login info, etc. from the previous session unless it prompts you for it again.

As they say, this is a feature, not a bug, since you don't want a server to keep all the old sessions alive and eating up memory when people have gone on to a different website or turned off their computers or somesuch.

Google's own sign-ins are usually 15-30 days long: I thought sign-in-with-Google was similar, but it seems to be shorter, is it?

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

I'm not sure about sign-in-with-google, but in general it's customizable on the server end.

Yup, I'm unsure also. I guess you can also customise the session length likewise, and Google sign in is just a handy way to register it.

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

ORLY? Please fire up your Terminal, enter “uptime”, and share the result with us

But I also never seldom close browser windows and there are LOTS of tabs I never close. Like the “unread posts” here on L19 and Oooooootakomoku


Grtz, Tom

_________________
“The only difference between me and a madman is that I’m not mad.” — Salvador Dali ★ Play a slooooow correspondence game with me on OGS?

That's my point.. it doesn't prompt to login again, nor does it indicate that I've been logged out. And there's no link.