64 coaches online • Server time: 21:42
Forum Chat
Log in
Recent Forum Topics goto Post Chaos Draft League R...goto Post anyone know how to c...goto Post Elf Draft Coach
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
DukeTyrion



Joined: Feb 18, 2004

Post   Posted: Jul 23, 2014 - 12:44 Reply with quote Back to top

harvestmouse wrote:
Zlefin wrote:
Nothing wrong with competing vs ai's. Having some competitive ai's might be fun. Also as an option for academy does sound nice.


There are no AIs competitive enough for a competitive division.


What about Woodstock? Razz
harvestmouse



Joined: May 13, 2007

Post   Posted: Jul 23, 2014 - 12:57 Reply with quote Back to top

Semi retired due to overheating circuitry.
Zlefin



Joined: Apr 14, 2005

Post   Posted: Jul 23, 2014 - 13:01 Reply with quote Back to top

harvestmouse wrote:
Zlefin wrote:
Nothing wrong with competing vs ai's. Having some competitive ai's might be fun. Also as an option for academy does sound nice.


There are no AIs competitive enough for a competitive division.


That's fixable enough; the problem is that there's not enough interest in having one; and not enough willingness to use it.

at least for a basic rather weak, but reasonably playable bot. How good of a CR would you want the bot to be?
Kam



Joined: Nov 06, 2012

Post   Posted: Jul 23, 2014 - 13:12 Reply with quote Back to top

170. At least.

I mean... As good as a coach who got 170 CR through playing a variety of race against a variety of opponents (not your average elfballer who only plays against elfballers or someone who has his CR bloated by a team full of stat freaks).

_________________
GLN 17 is out!
Image
mister__joshua



Joined: Jun 20, 2007

Post   Posted: Jul 23, 2014 - 13:29
FUMBBL Staff
Reply with quote Back to top

Wouldn't you, by necessity, need a variety of Bots at different CRs? Like any game it would need an Easy, Normal, Hard etc. Rookies would get just as little from playing a 170 bot as Legends would from playing a 150 one.

The thing is a low CR bot is easiest to produce, and as said could have a useful role in an Academy like setting.

_________________
"Yeah, well, you know, that's just, like, your opinion, man." - The Dude

Mr. J's LRB7 / Forum
Garion



Joined: Aug 19, 2009

Post   Posted: Jul 23, 2014 - 13:37 Reply with quote Back to top

well grod is the person to speak to about this, but I think he would need access to the client code so he could integrate it. His was built for a different client.

_________________
Image
koadah



Joined: Mar 30, 2005

Post   Posted: Jul 23, 2014 - 13:56 Reply with quote Back to top

Grod made it open so that anyone could create a bot but I don't think anyone did.

Just getting one to the level of a 151CR coach looks like a difficult enough task to me.

Considering that you can already get a poor AI from Cyanide and maybe even a better with BB2 there is not much incentive to put a lot of work into it.

Confusing rookies with a bad AI is probably a negative rather than a positive.

_________________
Image
O[L]C 2016 Swiss! - April ---- All Star Bowl - Teams of Stars - 2 more teams needed
Verminardo



Joined: Sep 27, 2006

Post   Posted: Jul 23, 2014 - 14:02 Reply with quote Back to top

koadah wrote:
Confusing rookies with a bad AI is probably a negative rather than a positive.


I think there is wisdom in this.

It might have a place for learning to know the client but really, the client is not that complicated. We have the Quick Start Guide and the 145 Club, both of which are awesome. Best we can do besides is be welcoming and patient with newbs.
Zlefin



Joined: Apr 14, 2005

Post   Posted: Jul 23, 2014 - 14:18 Reply with quote Back to top

Getting an ai past 140-145 would likely be quite difficult. Not that I have a very strong sense of what exactly the play is like at each number.
Not sure how much past that I could get an ai.
At any rate, it certainly sounds like there isn't much interest in one.
Also, an ai is likely to be good at the fundamentals, and bad at more complicated plays.
harvestmouse



Joined: May 13, 2007

Post   Posted: Jul 23, 2014 - 14:59 Reply with quote Back to top

Zlefin wrote:
harvestmouse wrote:
Zlefin wrote:
Nothing wrong with competing vs ai's. Having some competitive ai's might be fun. Also as an option for academy does sound nice.


There are no AIs competitive enough for a competitive division.


That's fixable enough; the problem is that there's not enough interest in having one; and not enough willingness to use it.

at least for a basic rather weak, but reasonably playable bot. How good of a CR would you want the bot to be?


Really? Well go ahead, Cyanide couldn't do it. Remember this is a much tougher job than making a chess AI. There are far more variables and play styles needed.
Zlefin



Joined: Apr 14, 2005

Post   Posted: Jul 23, 2014 - 15:10 Reply with quote Back to top

Cyanide probably wasn't trying all that hard; game companies often only put moderate effort into the ai's for them.

Last I heard, which was quite awhile ago, cyanide's ai was terrible. Has it gotten any better since then? because beating terrible isn't much of a standard.

The reason I don't make an ai, is because there hasn't been much interest in one here, and cyanide had no interest. And an ai that doesn't get to play versus much of anyone is pointless.
harvestmouse



Joined: May 13, 2007

Post   Posted: Jul 23, 2014 - 15:13 Reply with quote Back to top

Well it sounds like you know more than me about it. The problem I see is....you can make 1 good AI for chess. However for Blood Bowl it's nowhere near that simple.

Different rosters, different skills, different TVs. From where I'm drinking, making a star/super star AI sounds like a tough tough task.
Zlefin



Joined: Apr 14, 2005

Post   Posted: Jul 23, 2014 - 15:16 Reply with quote Back to top

It would certainly be quite a task; which is another reason I'd want a lot of real interest in it to put in the work on it.
I'm not sure what level "star" is, but super star would probably be too hard to make; it'd depend on how well the learning components improve over time. Most companies grossly underuse learning algorithms.
yggi49



Joined: Feb 08, 2011

Post   Posted: Jul 29, 2014 - 06:51 Reply with quote Back to top

Roland wrote:
I'd like to see a PBEM-like division, [P], where you start a game that might take a couple of weeks to complete. You take one turn at a time when prompted by PM for exmple.

that's how many smartphone boardgame-apps work for example

This would be awesome!
Grod



Joined: Sep 30, 2003

Post   Posted: Jul 29, 2014 - 07:25 Reply with quote Back to top

I would like to be able to search blog posts more easily. E.g. get a list of blog posts I have rated or commented on.

_________________
I am so clever that sometimes I don't understand a single word of what I am saying.

Oscar Wilde
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic