[MUD-Dev] Re: WIRED: Kilers have more fun

[J C Lawrence]

On Tue, 07 Jul 1998 03:23:11 -0000 (GMT) 
Matt Chatterley<matt at mpc.dyn.ml.org> wrote:

> The attitude of a lot of people towards naughtiness (lets be
> specific here, and look at the PKing, PStealing situation), is this:

> "This thing happening is X, thus to [en/dis]courage people [to/from]
> doing it, we will do Y, with Z consequences."

> The thing is, the entire approach is based on negative reinforcement
> - you add mechanisms to punish people who PSteal or PKill, and to
> allow retribution, and so forth - this is all well, and good (well
> thought out methods for this can be *fun* to play with, in a
> reasonable environment). How about positive enforcement to toe the
> line, though? It's often overlooked (particularly when a rapid
> solution is desirable).

Quoting froma message I wrote in July 1998:

URL:http://www.kanga.nu/~petidomo/lists/mud-dev/1998Q2/msg00880.html

--<cut>--

Exactly, and then once you have arrived at an interesting goal,
attempt to see what other goals the injection of this new goal creates 
in the system, and how you might take advantage of that.

Key here, and this is the lesson that M59, UOL, and company seem to
have been fighting, is the reverse of the old catechism of catching
more flies with honey: You can't stop players from trying to do things
you don't want them to do.  They're going to try to do whatever enters
their tiny gourd like craniums.  You can't stop it If you fight it
you'll just end up setting yourself up as the target de jour, and have
created an instant confrontational situation.  "How can I work around
these stoopid admins?".

Fight the players, even under the covers, and by golly, you're a much
more canny and interesting opponent than a game ever could be and its
a LOT easier to get emotive about the big nasty admin who is stopping
you than it is to get angry about the big nasty force of gravity that
is stopping you from flying...

In some manner you need to set yourself up as always working /with/
the players, even while attempting to proscribe their actions.  Its a
funny balance, and the echo of the only real effect of law making
being that of creating more outlaws/law-breakers.  

My visualisation of the correct process (ie more likely successful) is
aking to channeling a torrential river.  You can't just put up dams.
Torrential rivers have a habit of not liking dams, of under-mining
them, over-flowing them, or just busting clear thru them.  You have to 
guide, channel, divert, put big juicy carrots on the ends of
not-too-long sticks, provide reward-paths that just "accidentally"
happen to lead elsewhere than the behaviour you'd like to avoid.

You can't stop the river.  You can dig a nice deep gorge and "persuade"
the river that it would be rather fun going down that way.  The really 
tricky bit is to ensure that its never, "go down that way instead of
doing this", as they you've just set up something to be pushed
against.  You have to just do the, "Hey, this over here is neat!",
part of the distraction while you silently build a hug dam in the
background and work your arse off so nobody notices it.

--<cut>--

> An alternative of course, is to change our perception of normal,
> instead of a binary state (Normal, Criminal, or 0, 1), how about
> trinary? Back to the old Good, Neutral, Evil (no Chaos theory
> accepted at this time). You're evil if you do bad things (PS,
> PK). You're good if you kill a PKer, or return goods which a PS
> took, to their rightful owner. If you do neither, you are neutral
> (criminal status = -1, 0, 1 for Good, Neutral, Evil). If you are
> Good, and then PK or PS, you flick straight to evil (no
> 'adjustments'), but.. does that make sense? Now we only have a
> 'snapshot' of you - you were Good, but now, you're Evil! Decreased
> granularity can perhaps tackle this; the point is that if you ever
> PK, PS, or counter one of these crimes, you will *never* be 'normal'
> again. Punishment is doled out to the Evil (they get bashed by the
> Good). The Good are rewarded (bounties, and reward money). The
> Normal.. well.. anyone care to jump in?

I have intensely disliked reputation systems for a long while now as
being both excessively simplistic and presenting a too easily
trivialised model for players to manipulate for something I see as
very multi-dimensional.  Your text above gives an idea however.

Instead of having a single alignment or reputation value, or even a
pair of such values, presumed perpendicular as UOL is doing, instead
just keep simple counts.  "Good" actions would add to the character's
"good" statics, evil to his "evil" stat, etc, and the discrete
counting stats are kept seperately.  Thus using only the simple
good/neutral/evil trio, a stat call might look as so:

    > score bubba 
    ...
    Good: 934
    Neutral: 1,650
    Evil: 3,961

    > score boffo
    ...
    Good: 2890
    Neutral: 3098
    Evil: 735

    > score bernie
    ...
    Good: 6289
    Neutral: 837
    Evil: 5934

Where Bubba would be a fairly wicked chap, Boffo a kinda
middle-of-the-road semi-decent fellow, and Bubba being more than a
little schitzoid.  Expand this for your other behavioural scales and
you get a character being defined as the balance of the ratios of his
counts aross multiple scales.

Taking the old number cruncher of:

  "Urk, killed a baby so I got a -1,000 evil, so I'd better go script
kill Orcs for a while to get my reputation back above zero..."

Which leaves said player with a sum zero reputation (ha!), you would
instead have a score much like Bernie above whcih shows significant
(seeming) attempt to counterbalance ultra-good or ultra-evil acts.  A
possibly really nice touch here would be to expand the count to also
enclude a maximum single delta:

    > score blobbo
    ...
    Good: 6289    (1,012)
    Neutral: 837  (16)
    Evil: 5934    (112)

WHich of a sudden reveals that Bubba has committed at least one very
serious good act (per the 1,000 max delta), but generally idles along
commiting minorly neutral acts almost constantly with a slightly lower 
number of fairly evil acts and (probably) only a few massively good.

The next problem is with players who play a lot and so get very large
count stats, sufficiently large tha the max delta values are dwarfed
and so bear little relevance:

    > score bernie
    ...
    Good: 1,872,1894  (864)
    Neutral: 923,895  (1,050)
    Evil: 2,907,834   (900)

The relative ratios of the max deltas to the count scores is
meaningless.  About the most you can say is to gauge how long the
character has been playing (which is an extremely useful and thereby
questionable datum in itself) and do a similar character assessment as 
with the original Bernie above (schitzoid).  

So you need some sort of aging.  You can't afford to keep full stat
histories as the comuptational load as well as the storage
requirements becomes a bit extreme in the above case.  A couple
addresses:

  1) Keep sampled historical values for the max deltas (say once a
week going back for 8 weeks, and then a 9'th cumulative running
average of all prior averages), and then display the average of those
averages in addition.  This has the benefits of being computationally
cheap and light on storage (a total of 33 stored values, of which only 
27 are computed with on a simple arithmetic mean):

    > score bernie
    ...
    Good: 1,872,1894  Max: 864    Average: 532
    Neutral: 923,895  Max: 1,050  Average: 1,000
    Evil: 2,907,834   Max: 900    Average: 64

  The resultant communication is now quite a bit different as you can
spot trends in his balance of actions.  Bernie does an awful lot of
minorly evil actions, a minor number of extremely neutral actions, and
a large number of very good.

  2) As an extention or variation of #1, keep sampled historical
values for the individual counts (say, the same 8 weeks and a 9'th
runner) and then display only the current averages:

    > score bernie
    ...
    Good average: 6289    Max: 864
    Neutral average: 837  Max: 1050
    Evil average: 5934    Max: 900

  This loses the extra data of indicating how long Bernie has been
played by keeping the numbers reasonable while preserving the value of 
the basic ratios.

--
J C Lawrence                               Internet: claw at null.net
(Contractor)                               Internet: coder at ibm.net
---------(*)                     Internet: claw at under.engr.sgi.com
...Honourary Member of Clan McFud -- Teamer's Avenging Monolith...