Archived from groups: rec.games.chess.computer (More info?)
(Also might apply to other programs.) What types of parameters are modified
in the engine in order to change the "strength of play" setting of the
personalities or when creating new personalities? It also allows setting
values for maximum depth search, randomness, values of pieces, etc., so I am
wondering what other internal parameters would determine strength of play.
Thank you
Archived from groups: rec.games.chess.computer (More info?)
"DMB" <nospamplease@starpower.net> wrote in message
news:-PqdnUkAG78fsejfRVn-tA@rcn.net...
> (Also might apply to other programs.) What types of parameters are
> modified
> in the engine in order to change the "strength of play" setting of the
> personalities or when creating new personalities? It also allows setting
> values for maximum depth search, randomness, values of pieces, etc., so I
> am
> wondering what other internal parameters would determine strength of play.
> Thank you
>
>
I would also be interested in knowing more about how computers work out what
moves to do in order to be a certain rating, good or bad.
Archived from groups: rec.games.chess.computer (More info?)
>From Chessmaster's lower rated personalities, I know that they simply
consist of more or less big changes of the same parameter settings, the
user can edit too. You can see them when you load such a personality
into the "create or change custom personality" dialogue (or what it's
called in the menu). For example, I remember one which evaluated
knights higher than rooks. But they other settings are used too, often
in combination, like limited search depth etc.
The rating numbers are not really reliable, although it has been said
that some of them were tested against USCF rated players to benchmark
them. But AFAIK they did not repeat that for any new CM version. The
computer speed should affect the ratings though, on the individual
system (as a rough estimation of the difference).
I cannot comment on changes other progs make to reduce the strength. I
would guess the ideas are similar, or they apply some randomness...
Archived from groups: rec.games.chess.computer (More info?)
>I would also be interested in knowing more about how computers work
out what
moves to do in order to be a certain rating, good or bad.
I'm no programmer, so if you want a technical answer, read no further
:-)
There is a number of ways of approaching the problem of dumbing-down an
engine, however. I've never owned a copy of any version of ChessMaster,
but as I understand it, that programme sports a number of virtual
"personalities", each with its own distinct playing style.
In a similar vein, Fritz 8 (which I do have) has settings like
"Assassin", "Patzer", "Desperado", "Fraidy-Cat", "Drunk", etc.
What is happening in these cases, in effect, is either that the
programme's look-ahead is reduced, or its evaluation function is skewed
somehow, or, more likely, different proportions of both.
Taking away a programme's opening book can cause it problems, too, but
if that's the only measure adopted, then a programme playing at, say,
2600 strength may be able to evaluate a position which is likely to be
out of *your* book (as well as its own) more accurately than you
can....
As far as programmes' finding good moves is concerned, the longer the
look-ahead, and the better the evaluation (necessarily *both* of these)
the higher the rating.
Of course, having a faster processor (or cluster of processors) and
lots and lots of memory helps, too.
Archived from groups: rec.games.chess.computer (More info?)
DMB wrote:
> (Also might apply to other programs.) What types of parameters are
modified
> in the engine in order to change the "strength of play" setting of
the
> personalities or when creating new personalities? It also allows
setting
> values for maximum depth search, randomness, values of pieces, etc.,
so I am
> wondering what other internal parameters would determine strength of
play.
> Thank you
In general, it should be no surprise that the "Strength of Play"
setting is the one that controls how good the engine is, meaning, how
well it evaluates any given position. And, for the most part, any time
you deviate from the default "Chessmaster" personality settings, you
are going to cause a DECREASE in value.
However, a great deal of testing has been done by computer chess fans
that have resulted in personalities that play as much as 100 points or
more stronger than the default. But many of these personalities are
actually worse than the default when tested in other ways (such as with
different time controls, for example).
But all of the many variables available in Chessmaster will affect how
a personality plays, and therefore will affect the strength of that
personality. The ratings of the personalities in CM8000 were calculated
from over 10,000 games, including almost 3000 games between some of the
personalities and almost 100 USCF rated players rated between about 550
and 2400. The standard ELO calculation was then used to extrapolate the
ratings for all of the personalities in the game.
As far as I know, no testing like this has been done on the CM
personalities since CM8000.
You are about to answer a thread that has been inactive for more than 6 months. If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.