Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> question...
        
(Message started by: DannyV on Jun 2nd, 2013, 5:35pm)
Title: question...
Post by DannyV on Jun 2nd, 2013, 5:35pm
Hi all, i just performed an inpromtu experiment vs a bot who i noticed opened with the same couple moves it had used in a previous game (vs me) that i had just been reviewing. i had won the previous game so i thought "well, what if i repeat the same exact responses to all the bots moves as i had done before, will the game play out exactly the same? or is there any 'randomization' to a bot's decisions which will cause a change in decision at some point?" well, i was a little disappointed to see that, sure enough, the bot dutifully repeated all if its exact same moves as before, granting me a "free win" by just copying all the same moves of my own from the other game. despite the free victory, i was actually disappointed, as i was hoping to see the bot be "smarter" than this. so anyways, i have 2 questions... 1) is this standard for all bots? meaning will all bots simply make the same exact calculations and always make the same decision in a given situation, or do the better bots have some way of protecting themselves from heading down this potentially "once beaten, always beaten" path? ... and question #2) if so, is the practice of taking advantage of the bots in the way i did here looked down upon or outlawed?

thanks in advance for any feedback!

Dan
Title: Re: question...
Post by mattj256 on Jun 2nd, 2013, 11:21pm
My off-the-cuff answers.  (1) This is just one of many easily exploitable flaws that weak bots suffer from and stronger bots avoid.  (At least I think the stronger bots avoid it!!)

(2) As far as I know there's no formal rule against it.  There are already plenty of posts about trying to improve the rating system.

Whole History Ratings: http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=talk;action=display;num=1207699394 (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=talk;action=display;num=1207699394)

This is from the 2013 World Championship Rules: (http://arimaa.com/arimaa/mwiki/index.php/2013_World_Championship_Rules#Seeding)
  • Players will be seeded according to the WHRE ratings ... Players with fewer than six WHRE games will be seeded according to their gameroom ratings, adjusted at the discretion of the Tournament Director to account for artificially inflated or deflated ratings, since gameroom ratings can be manipulated much more easily than WHRE ratings. Seeding has a large effect on early-round pairings, but rapidly decreases in importance; after about three rounds
    seeding becomes irrelevant as pairing is then done entirely on the basis of in-tournament results.

Of course I would prefer that you not artificially inflate your gameroom rating!  It's way more fun to leave your rating mostly accurate so that as you improve you can watch your rating improve.
Matthew
Title: Re: question...
Post by rbarreira on Jun 4th, 2013, 3:06am

on 06/02/13 at 17:35:48, DannyV wrote:
1) is this standard for all bots? meaning will all bots simply make the same exact calculations and always make the same decision in a given situation, or do the better bots have some way of protecting themselves from heading down this potentially "once beaten, always beaten" path?


The bots that are most likely to fall onto this trap are the weaker/older bots, as well as bots labeled with P1 / P2. That's because more recently, bot authors became more aware of this problem and took steps to build in some inherent randomness in the bot's decisions. P1/P2 bots are more likely to repeat their moves because CPU speed / server load (another source of randomness) generally does not affect their moves.


on 06/02/13 at 17:35:48, DannyV wrote:
#2) if so, is the practice of taking advantage of the bots in the way i did here looked down upon or outlawed?


There's a general practice of using unrated mode when making experiments or playing games in an abnormal way (the normal way being you playing a game without relying on external help and without handicapping yourself).
Title: Re: question...
Post by chessandgo on Jun 6th, 2013, 4:20pm
If you've memorized the original game it's clearly allowed, however I'm not sure if you can keep a previous game window open during an event game. I'd guess not, although the policy of allowing whatever can't be prevented often applies in the championship rules?
Title: Re: question...
Post by DannyV on Jun 8th, 2013, 12:32pm
excellent feedback. thank you all, i appreciate it. i definitely do NOT plan on using this method to artificially increase my rating. i guess my bigger concern is that the bots are vulnerable to OTHERS doing this. but it sounds like the better bots are probably not so vulnerable to it anyways. also i like the idea of the rule (or potential rule?) to be NOT allowed to have a previous game window open during a current game. it would be nice if this rule could be somehow enforced by the site itself, if that's possible? but either way, thanks again for the thoughtful answers guys!
Title: Re: question...
Post by Marty on Jun 9th, 2013, 3:40am

on 06/08/13 at 12:32:42, DannyV wrote:
... also i like the idea of the rule (or potential rule?) to be NOT allowed to have a previous game window open during a current game. it would be nice if this rule could be somehow enforced by the site itself, if that's possible? ...

you could forbid multiple game windows opened in one way or another, but then a "cheater" could just copy the moves to notepad opened next to his single game window and play according to them. or he could write them on a paper and play according to it. and such "cheating" is undistinguishable from memorizing the game.

more effective would be detecting repeated games and discarding them from rating system. yet even then there remain other ways of messing with rating, so there is not much of a point to lose sleep over that if it is not a common occurence
Title: Re: question...
Post by browni3141 on Jun 9th, 2013, 9:48pm

on 06/09/13 at 03:40:08, Marty wrote:
you could forbid multiple game windows opened in one way or another, but then a "cheater" could just copy the moves to notepad opened next to his single game window and play according to them. or he could write them on a paper and play according to it. and such "cheating" is undistinguishable from memorizing the game.

more effective would be detecting repeated games and discarding them from rating system. yet even then there remain other ways of messing with rating, so there is not much of a point to lose sleep over that if it is not a common occurence


Also, I think blocking multiple windows would cause more harm than good. I personally have windows from older games open a lot when I am playing because I want to analyze them later, and it would be a pain to find and open them again because I had to close them to play a game.
Title: Re: question...
Post by Arimabuff on Jun 23rd, 2013, 7:49am

on 06/09/13 at 21:48:33, browni3141 wrote:
Also, I think blocking multiple windows would cause more harm than good. I personally have windows from older games open a lot when I am playing because I want to analyze them later, and it would be a pain to find and open them again because I had to close them to play a game.

The more we do that kind of increase in complexity the more we augment the chances for the system to crash so we'd better do it for a good reason and not a silly one like the possibility for someone to increase his rating. After all that kind of artificial and undeserved rating can be a double edged sword as it will be a bounty for anybody slightly better than you that can beat you in a human vs human game while you will have only temporarily benefited from it. It'll be like in the old west when the fastest to draw became an automatic target for every young gunslinger who wanted to make a name for himself...

I may be overdramatizing this. ;D
Title: Re: question...
Post by Hippo on Jun 23rd, 2013, 9:12am
You can replay the game using the user script provided by ... risking the deviation... I dont think it is a problem.
Title: Re: question...
Post by DannyV on Jul 8th, 2013, 2:41pm
hippo can you explain what that last statement means?
Title: Re: question...
Post by Hippo on Jul 8th, 2013, 3:29pm

on 07/08/13 at 14:41:52, DannyV wrote:
hippo can you explain what that last statement means?


harvestnow made a script able to reply the game ... and I was lazy to search the forum to be enough specific.

http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=siteIssues;action=display;num=1363906756;start=0#0
Title: Re: question...
Post by harvestsnow on Jul 15th, 2013, 6:27pm
Just to be clear, the reason I made this was to allow experimenting more easily on specific positions. It would typically be used in an unrated game between two human players curious about alternate lines of game xyz, or for EEE-style openings.
It also makes it easier to exploit deterministic bots; for the reasons stated in this thread, I don't think it's a big concern.


Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.