Page 5 of 5

Re: Anti-botting Turing test

Posted: 09 Sep 2010 09:30
by Budwise
About the bot-test. Told this to Petros but not sure how seriously he took me.

The number is way too long. Its already hard to read and in combat
it will scroll off really fast (at least on my screen).
If its not readable by sneaky automatic means why not stick to 4 digits
which are much easier to see, remember and type?

I tried it as a test in a reasonably spamfree environment and it usually reset
before I had a chance to read and type all of the loooong number.

It also feels very intrusive and out of place. Yes, I know we have a bot-problem,
but at least drop the pretense of it being a "game". And the silly gem prizes.

And Gorboth as the judge and jury? Gor-Bot H. Need I say more? ;)
Also can I get a a "No Test"-card? Its a bit hard to bot using only a Quicktyper,
so if I don't reply to this game, its probably because I'm making a point
of ignoring it or am almost asleep. (Usually happens after 40 trolls/cows)

Re: Anti-botting Turing test

Posted: 09 Sep 2010 17:06
by gorboth
Budwise wrote: Gor-Bot H. Need I say more? ;)
Hey! Wizard bots are totally necessary! How do you think I check my mail and catch up on boards in the game. Do people really expect me to read boring mail after mail after mail after mail even though it is the same thing I've read before a million times? That is what triggers are FOR people. You want me to actually read my mail without triggers? How about making my mails more FUN.

G-B-H.

Re: Anti-botting Turing test

Posted: 09 Sep 2010 17:14
by Tapakah
A book I like very much translates G B H as Grievous Bodily Harm.

T.

Re: Anti-botting Turing test

Posted: 19 Jun 2011 20:21
by Bendis
I can't create a new topic, so I'll have to post a reply to a hijacked :) topic.

For anti-botting there can be a different approach: instead of a classic (reverse) Turing test, may be applying "intelligence" would help. We can build a (bayesian) model of what a normal player means. Bayesian means that it will keep learning from both hits and misses as the players' behavior changes.

One way (based on my approximative knowledge of the mudlib) is to put something in the player soul that would track the last e.g. 100 actions (we can measure what is the right number of actions). It is not snooping as the info is kept in the memory, with no central place and everything over 100 actions is discarded.

We can work with a couple of willing players to capture actual data and do some initial classification. We will test then the model and improve it based on hits and misses. Once in use, the model will improve based on the every success/failure, we will keep metrics etc...

I will let you decide the punishment once a bot is discovered :D on a different topic hopefully :D

Just thinking out loud,

/Bendis

PS probably this problem is already solved, but as the last posts were still bantering :D I thought to continue the discussion.

Re: Anti-botting Turing test

Posted: 19 Jun 2011 21:41
by Booger
I don't think snooping is "illegal", and no matter what, I don't mind if all you wizzes snoop me anytime you like as much as you like - it's a good way to find out what's good and what's bad, what players don't use and what we (ab)use a lot. Way better than asking us what we want, because you'll get truthful answers.
But back on to the (hijacked?) topic of last post - I'd be happy to be used as test object for this purpose... if you think it'll be of any use. Be my guest and... good luck! (you'll need it :twisted: )