It seems to me like that algorithm "unfairly" penalizes groups of
similar words. I am thinking that e.g. signore and eringos would be
penalized relative to ignores and regions, since the s and e hooks
would be available to either and the g and n are in the middle, so
anytime you can play a 7 in that rack, you can likely play either
signore and eringos but if you need to start with an i or an r, you'd
put more weight on the more commonly known words. Maybe it's not the
best example, but you can probably generate a better one.
Put another way: Is knowing ceratin much less valuable than knowing
tacrine if, in the million games you simulate, ceratin can be played
iff certain can be played for equal points (generating zero equity
advantage for both) but sometimes tacrine can be played with a big
equity advantage? Assuming you know certain, I guess ceratin is less
valuable than tacrine, but knowing either ceratin or certain (one or
both as a group) might be more valuable than knowing tacrine in this
case.
I guess I would argue for looking at the equity advantage of the play
relative not to the second-best play, but the next-best play that uses
different letters from your rack.
Have you thought about adding points for words that are less commonly
known (I'm thinking you could flag all 4- to 15-letter OWL2 words that
do not appear in a common spell-checker, say) as being more likely to
draw a challenge?
On 4/11/06, John O'Laughlin <[EMAIL PROTECTED]> wrote:
>
> Each time a word is played, I note the difference between it (its
> equity) and the second best play. This will be high for QI if it is
> your only Q play, and low or zero for something like ERASION when it
> is one of many equally good bingos.
>
> John
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/quackle/
<*> To unsubscribe from this group, send an email to:
[EMAIL PROTECTED]
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
