News:

NEW! Get all Tourneybot Results from the new Board here: http://www.fibsboard.com/tourneybot-results/

Main Menu

Is GnuBG cheating with the dice?

Started by RaughMachine, June 02, 2009, 03:14:31 PM

Previous topic - Next topic

Tom

Quote from: ayran on December 26, 2012, 02:56:26 PM
Wanna know what the 'SEED' number is to roll double sixes ? I know what it is - do you ?  :)
I haven't a clue why that's the case, but it works every time.
Which makes me - a little rank amateur of course - very suspicious about what this 'random number generator' is actually doing  :wacko:

Of course, the SEED is an internal number that the random number generator uses to generate the next number.
So setting that to a specific value could return the same dice all the time.

but like I said EDIT is much easier

tom

ayran

Quote from: Tom on December 26, 2012, 04:19:14 PM
Of course, the SEED is an internal number that the random number generator uses to generate the next number.
So setting that to a specific value could return the same dice all the time.

but like I said EDIT is much easier

tom

:cool:  Ok, thanks for all that info, Tom. Much appreciated !  B)

I think I'll play chess. No possible 'cheating', 'lucky' or 'unlucky' moves in that  :hypocrite:

Happy New Year !  :)

Subusel

I am sending a picture. How can we explain? %100 cheat.
http://imgur.com/1Y5OudC

Zorba

That's a pretty freaky sequence indeed. Rolling 12 doubles or more in only 24 rolls has a chance of 1 in 6,000 roughly. Very rare, yet bound to happen to someone, somewhere, sometime. But you can always use manual dice with gnubg, it will still beat you very often.

More interesting is: Baba's cube to 2 and the recube to 8 were both blunders, costing around 0.18 EMG points each. The pointing on the 2point right after the initial cube, was a massive triple Whopper-with-cheese play costing 0.38 EMG. The loose hit 11/3* a royal double blunder costing 0.27.

So regardless of the dice, there's a lot of room for improvement there!


Gnubg must have been playing on very low settings here, as its last cube to 16 is terrible (5 checkers on ace, versus one checker each on 1,2,3 and 4).

The fascist's feelings of insecurity run so deep that he desperately needs a classification of some things as successful or superior and other things as failed or inferior. This also underlies the fascist's embracement of concepts like mental illness and IQ tests.  - R.J.V.

Luck is my main skill

pck

#24
Quote from: Zorba on March 12, 2013, 01:56:10 AM
Rolling 12 doubles or more in only 24 rolls has a chance of 1 in 6,000 roughly.

Rolling at least 12 doubles in 24 rolls has a probability of

(24 12) * 612 * 3612 / 3624 = 0,001242272 =~ 1/805

where (24 12) is "24 choose 12", that is, 24! / (12! * 12!)

Zorba

#25
I think there's an error in your formula there somewhere, but I don't know what it is. [edit: I think the 3612 should be 2512 and then you have the p for exactly 12 doubles.]

I just use the cumulative binomial distribution, with p=1/6, n=24 and X=12, then look at P {X>=12} which I'm pretty sure is the correct method.

http://stattrek.com/online-calculator/binomial.aspx

which gives you 0.0001697
The fascist's feelings of insecurity run so deep that he desperately needs a classification of some things as successful or superior and other things as failed or inferior. This also underlies the fascist's embracement of concepts like mental illness and IQ tests.  - R.J.V.

Luck is my main skill

pck

#26
Quote from: Zorba on March 12, 2013, 06:58:11 PM
I think there's an error in your formula there somewhere, but I don't know what it is. [edit: I think the 3612 should be 2512 and then you have the p for exactly 12 doubles.]

It needs to be 3012 for exactly 12 doubles. (30 = number of non-double rolls)

And yes, my result for at least 12 doubles was wrong. Selecting 12 places out of 24 and putting doubles into them and arbitrary rolls (including doubles) into the others doesn't work. That method counts certain sequences of the 24 rolls twice or more.

What I need to do for at least 12 doubles is to calculate the

sum for n=12 to 24   of   (24 n) * 6n * 3024-n

(exactly n doubles for n from 12 to 24 -- prevents double counting of sequences)

and divide by 3624

which is what the stattrek calculator does.

boomslang

Quote from: pck on March 12, 2013, 08:21:32 PM

What I need to do for at least 12 doubles is to calculate the

sum for n=12 to 24   of   (24 n) * 6n * 30n          (exactly n doubles for n from 12 to 24)

and divide by 3624

which is what the stattrek calculator does.



It should be: sum for n=12 to 24   of   (24 n) * 6n * 3024-n

(and divide by 3624)

pck

Quote from: boomslang on March 12, 2013, 08:32:30 PM

It should be: sum for n=12 to 24   of   (24 n) * 6n * 3024-n

(and divide by 3624)

Yes, thx. (Corrected in posting above.)


ushimata

Quote from: ah_clem on June 02, 2009, 04:12:42 PM
gnubg does not cheat with the dice.  You can verify this by simply setting it to "manual dice", and playing a few matches by rolling a real pair of dice and entering the rolls yourself.  You'll still get clobbered.


This is a very old tired topic, as every newbie has the impression that gnubg cheats because it wins almost all the time, when the real reason it wins almost all the time is that it's just a much better player than the newbie.

BTW, if you're frustrated with gnubg winning too much, you can dial down it's playing level to "beginner" or "casual player".  This definitely makes it more fun for newbies.  Once you start beating it on a regular basis you can increase its level accordingly.
wrong ! i set it to manual dice and i clubbered the SOB. before that i hardly won a game since netgammon stopped.