Re: PBM design and formulae - a fresh start

From: d8hanma@dtek.chalmers.se (Martin Hansson)

Date: Sat, 03 Jul 1993 09:05:43 +0000

>gym@dcs.ed.ac.uk (Graham Matthews) writes:
>Ah but you have glossed over the key point. How does a player train his
>neural net? By playing! He doesn't start with a book in the physics of
>tennis, he starts by playing. Now if I want to simulate this I should
>allow my players to practice (thats like the players practicing in
>the real world), but to give them the equations would be to say that
>you learn tennis in the real world by sitting and reading physics books.
>
>gym@dcs.ed.ac.uk (Graham Matthews) writes:
>> >c) knowing the equations would not help a player play better because
>> >by the time he/she had solved the equations the ball would have
>> >gone past them.
>saschreu@cs.vu.nl (Schreuder SA) writes:
>> I have to object against this: A monkey will hit the ball in a random
>> direction. From Physics one can conclude that the ball will travel
>> in an almost straight line, and curve downwards. Now supose somebody
>> with this knowledge plays against a monkey. The monkey will loose,
>> because halve of the time, it won't hit the ball in the direction
>> of the net!
>>
>> Now our physist (P) will play against an average tennis player (T).
>> P will have a better chance against T than the monkey has, because
>> P knows how to hit the ball. So knowing the rules gives some compensation
>> for the lack of experience.

>I really don't follow this argument. The player knows to hit the ball in
>the direction of the net because he knows the rules of tennis - NOTE: not
>the equations of motion of a tennis ball, but the rules of the game (eg: what
>is in, out, etc). I could just as well teach the monkey these rules too,
>and then he too would play good tennis. To my mind you continue to mix up
>equations and rules. Rules give you general guidance - the framework - and
>must be known by all players in nearly every game. Equations give you
>details that the entities you are simulating don't know and hence the
>players should not know. In this case a real tennis player does not
>know the equations, so neither should the player in our simulated
>game.

I think that you have just proven the point why it is better to know
the formulas in advance. The point is that when formulas isn't known
an experienced player will whipe out those that start fresh. This is
what all the people that wants formulas wishes to avoid (well me at
least). I think that you being an experienced player of tennis is
equal to having played a game for a long time. I who havn't played
tennis a lot is an unexperienced player. On the tennis court you would
beat me easy. In a game you would also beat me easy if the formulas
weren't known. Now if the formulas were known you would still have
your experience but I would have at least a chance since I would also
know the formulas and wouldn't have to rely on experience wich I
don't have.

Tennis isn't a good example however. It translates poorly to a
computer game. It depends information that the eye and brain can
easily comprehend but it would be extremly difficult to translate to a
_text_ based computer game. Graphics is ok but not text.

You mention in another post that FlatTop was good since you could play
without knowing the formulas. You also say that this is because you
know something of planes and ships. We again have the same situation.
I don't know anything about those so you would beat me in such a game
too.

If there were some good way of making the type of game you seem to
preefer it would probaly be fun to play, but I can't imagine what such
a game would look like.

References

Re: PBM design and formulae - a summary
gym@dcs.ed.ac.uk (Graham Matthews)
Thu, 01 Jul 1993 11:40:27 +0000
PBM design and formulae - a fresh start
gym@dcs.ed.ac.uk (Graham Matthews)
Fri, 02 Jul 1993 12:01:01 +0000
Re: PBM design and formulae - a fresh start
saschreu@cs.vu.nl (Schreuder SA)
Fri, 02 Jul 1993 14:04:09 +0000
Re: PBM design and formulae - a fresh start
gym@dcs.ed.ac.uk (Graham Matthews)
Fri, 02 Jul 1993 16:20:03 +0000