On 5/10/07, John O'Laughlin <[EMAIL PROTECTED]> wrote:
>
> Stu,
>
> The question from my point of view is why other systems have had the E so 
> very high. On
> a typical board, the difference if you sim exchanging seven tiles or 
> exchanging six to keep
> an E will almost always be very slight. I don't know all of the details of 
> how Watkins,
> Sheppard, Gordon, and others created their leave values, but I can see how if 
> a program
> has only a very basic concept of synergy (or none at all), "overvaluing" the 
> E might actually
> cause it to play better.

My approach had no concept of synergy except for:
- a V/C factor (similar to the one you give in a different email in
this thread), and
- a duplication penalty for each tile.
I then used successive large series of games played with different
values for all these numbers to tune the final values (based on
winning percentage).  The lack of group synergies in my approach was a
flaw (it would consider QU no better than Q).

I believe Sheppard did have a reasonably effective concept of tile
group synergy.  I believe that Sheppard also tuned his numbers by
playing variants of Maven against itself, but based his valuations on
equity .  (I also have a recollection of Brian saying that at one
point he slightly modified these numbers that were optimized for
computer vs. computer play to work better against human opponents who
tend to play less open than computers do, although I also have a
recollection of Brian denying he ever did this.)

As I understand the Quackle scheme as explained in this thread, AE
would be less highly valued than EE and ERN would be less highly
valued than ERR, whereas both Sheppard's approach and mine would both
prefer AE and ERN, respectively.  So, Quackle must have some
additional synergy factors to account for this - right?

Steven Gordon

Reply via email to