The problem is that there's no ONE way a 1500 (or whatever) rated human
plays. Some know a lot of very obscure words,  Some have got most of the
probable stems totally memorized, some may be lacking in vocabulary but very
good at board vision and strategy ETC.. 

 

The problem with playing Quackle at the highest level is that, along with
the demoralization factor, I think it effects strategy in an unrealistic way
when your opponent will always make the best possible play. Like if someone
was trying to maximize their (slight) chances of winning against Quackle
would they play very defensively and get into that habit against less
skilled players where it may be a disadvantage?

 

I think the way a dumbed-down Maven will sometimes come out with a very
obscure word or brilliant play is a good way to learn. But if you wanted to
make "artificial stupidity" more realistic. I think it might be as simple as
breaking the skills in the game down into categories like "Vocabulary in
high probability words", "vocabulary in low probability words", "endgame
strategy" and some others. Then you could have sliders with 100 representing
the "perfect" player in each category and 0 representing a complete novice
player in that particular category.. So you could customize it so, for
example, your computer opponent wouldn't play as many very obscure bingos
but would play a very good endgame. It would be interesting to either match
yourself against a computer opponent with about the same level of strength
in different areas that you have or make it stronger in areas that you are
weaker and improve that way.

 

Mike

 

 

From: [email protected] [mailto:[EMAIL PROTECTED] On Behalf Of
Graham Toal
Sent: Tuesday, September 04, 2007 11:38 AM
To: [email protected]
Cc: [EMAIL PROTECTED]
Subject: Re: [quackle] turning Quackle down

 

On 9/4/07, John Van Pelt <[EMAIL PROTECTED] <mailto:verbalobe%40gmail.com>
> wrote:
> for one thing, the "blunders" that the 1600-level Maven made did not
consistently resemble real 1600 players in any way.

Although it's probably not a discussion for the Quackle list, if you
want to take it next door to wordgame-programmers, I'm interested in
working out the details of what exactly would be needed for an
Artificial Stupidity System to make the computer play as convincing as
a poor human. Over the years I've made a mental list of a few points,
though maybe it's time we started writing some of them down in case
anyone ever gets round to doing this.

For example, active vs passive vocabulary, plausible phonies, use of
front/rear hooks, parallel plays, etc can all be parameterised as
features to use or not within a game.

My idea of a playing like an ASS would include tracking opponents so
that the computer always played a *little* better than them but not
much, so they have a chance of winning a good percentage of the time
(eg 40%); it should track which of the play techniques the human has
mastered and use the same things itself, eg when a human finally plays
a parallel play, then the ASS can do so too, but not *too*
impressively.

If it doesn't have a history of games played, it shouldn't try to play
an even game by using too simple an algorithm of trying to match the
score for example - you don't want it to bingo only when you bingo, or
to gradually fall behind and then place QUETZAL on a TW to catch up in
a single play.

It should also have a teaching role - after a few games, and once its
play is matched to the opponent, it could introduce *one* new
technique (eg putting tiles on both the start and end of a word to
extend it) which the human hasn't yet done, with the intention of the
human learning better ways to play. But not all at once in an
intimidating way.

It should *never* be obvious that the ASS is throwing the game. In
fact it might be fair to start it off with only a basic vocabulary of
English, and have it learn new words from web pages it happens across,
and it might also have a limited memory for word storage so that it
forgets seldom used obscure words that it has learned a few days after
learning them. Google's "did you mean" prompt might be something it
could learn from. It could also ask a dictionary authority for a
check now and then.

That's a few ideas off the top of my head, maybe y'all have some more.

Graham
PS GMail doesn't let me redirect the reply-to so could you please
follow up to wordgame programmers unless there's a lot of interest and
a strong consensus that this is relevant to Quackle.

 

Reply via email to