Re: [game-lang] Late to the party

On Sat, Aug 4, 2012 at 3:10 AM, Simon McGregor <londonien@xxxxxxxxx> wrote:

In pool tournaments at my University, the rules stated that
intentional fouls led to a game loss. This situation could arise if
you were snookered and likely to foul anyway - in principle, you are
supposed to do your best to play one of your balls, but there could be
an advantage to playing a different shot instead to leave your
opponent in a worse position after your foul.

It seems that the official enforcement rules of the Magic card game
also provide different penalties for intentional and unintentional
infractions:

www.wizards.com/dci/judge/MTG_DCI_Judge_Penalty.asp

It's interesting that intent comes into these in totally different ways. In the pool tournaments, the move taken is within the scope of the game, and one could argue that the judgement based on intent reflects a gap or flaw in the game design; in Magic, the infractions involve events outside the scope of the game that affect the result, and the judgement of intent is about preserving a larger social contract.

Yes - in a computer implementation of a game, rules violations can be
prevented (leaving aside the possibility of hacking). But one can
imagine a variant of Mao in which it is not legal to deliberately and
knowingly play a "violating" move in order to strategically incur the
card penalty without otherwise changing game state. It's not clear how
to formalise such a rule.

I think that a sufficient formalization of that rule would be an instance of an artificial general intelligence. :)

...and getting back to something more concrete, I think that might be a limiting problem in mechanizing Mao: a tutor program would be able to give out penalty cards based on pre-defined rules, but if (as in human play) the tutor isn't necessarily the player informed of a new rule, the tutor will try to give penalties to the enforcer of a new rule. You could try to design a "meek tutor" that can recognize the difference between breaking existing rules and enforcing new ones, but that's likely to be hard, if not AI-hard.