Backgammon AI Cheats

[StanJr]

Like there is no tomorrow.

 

I have never played a real game of Backgammon where doublets come so naturally. But in Atari Backgammon, the AI gets a doublet one out of every five tosses. Is it programmed to be a low-down dirty cheater or is it just exploiting shoddy programming?

 

Either way, it's making me insane.

 

:mad:

[SoundGammon]

Well, you can use your own dice and "dial" them in! The right difficulty switch in "b" position let's you do that!

Have you ever seen it roll double nines yet?

[+atari2600land]

Oh, I thought it was about cheats for the Backgammon game. I wondered what kind of cheats there would be.

[Nukey Shay]

Simply put, the random generator in the program sucks (for both players). IIRC it was a choice between rolling too many doubles or virtually none at all. As mentioned, you can dial in your own values.

[+Random Terrain]

Simply put, the random generator in the program sucks (for both players). IIRC it was a choice between rolling too many doubles or virtually none at all. As mentioned, you can dial in your own values.

 

Can it be hacked so it doesn't suck?

[Nukey Shay]

Yeah...come up with a different routine for $C5 & $C6

 

One way that should make the results more-unpredictable would be to use one or both free bytes of ram memory to hold a random seed value to be passed to $C2 for the number of roll time frames (the game uses absolute value #$78 at LF347 - and does it again if bits 0-2 end up with a value of zero or seven). Or, you could alter how a "random" value is picked (ram location $E2 is just ROL'ed with whatever happens to be in carry status...which is why the same dice values appear frequently together).

 

It's easier just to use real dice

 

 

The program is not intelligent enough to "cheat". The double-nines issue is a specifically-coded easter egg.

[StanJr]

No double nines, yet, but that would be cool to see.

 

Dialing in your own numbers is the worst kind of cheating. I would roll a six doublet every time and win for sure.

 

I guess I'll keep playing with the crappy dice, although the chance for a gammon is highly unlikely and a backgammon pretty much impossible.

 

Stupid shoddy programming.

[Pioneer4x4]

..., the AI gets a doublet one out of every five tosses....

The odds are 1 every 6, that doesn't sound that far off.

[Nukey Shay]

Revealing the easter egg can be done easily in board setup of Acey-Deucey...start the game with both difficulties set at B and select a higher number for the first die so that you have the first move. Block all 6 of it's opening spots so it has no valid opening moves, and flip Left Difficulty to A to make your move with one of the remaining stones. Give the dice a 1 and 2 roll for the computer's turn to allow it to pick doubles.

 

This is mistakenly called a programming bug on the Digital Press Easter Egg page:

http://www.digitpress.com/eastereggs/26backgammon.htm

 

Being allowed to pick doubles in this case IS a programming error (being unable to move the initial 1 and 2 should end it's turn according to the rules)...but the easter egg does not depend upon that mistake. It can be revealed any time the computer is allowed to pick doubles but has no valid moves for any dice value 1-6 (regardless of where the stones are).

[Nukey Shay]

BTW how would you roll double-6's consistantly using real dice? Don't you have any that are not loaded?

[StanJr]

That's my point, if you can roll yourself a guaranteed doublet-six every time, that is the worst kind of cheating as you could never get that with real dice.

[Nukey Shay]

So don't do it. Dial in the values you've thrown from the real (non-loaded) dice. Problem solved.

[StanJr]

Oh, I see. That is an interesting option, isn't it. I assume that I dial in the AI rolls as well?