[Ar-list] Thoughts on various matters

[Carter Butts]

Howdy.  I've been out of town for the most part since the discussion 
started, but have read the comments with interest.  (I was almost afraid 
to say anything, since comments by the initial authors seem to have a 
Kiss of Death effect on the list....I'll brave it, though.  :_))  I 
don't want to interrupt the flow, but let me make a few comments on 
several of the issues which have been raised....

1. Freudian attributes

	While I understand the aversion to the the Freudian system on scientific 
grounds (I'm a social scientist myself, after all), I personally tend to 
like it for the following reasons: it's easy to role-play; most people 
(that I've encountered, anyway) seem familiar with the basic concepts; 
it's reasonably incentive compatible; and it generates behaviors which 
have a certain level of face validity.  It could be changed, however 
(and you can always change it yourself, either informally or via a 
Modular Rule Supplement).
	In terms of what alternatives might exist, the most obvious candidates 
would be based either on the "big five" (B5) personality factors, or on 
something related to Affect Control Theory (ACT).  Both have a 
reasonably large body of research behind them, though I'm less familiar 
with the former than the latter.  (I'm also underwhelmed by what I 
_have_ seen vis a vis the predictive power of the B5, but that's another 
matter.  ;-))  While ACT is pretty strong as a theory, it doesn't really 
lend itself to use in an AR context...the B5 should be better in this 
regard, though whether they'd be playable is less than obvious.
	In any event, I'd be interested in seeing some MRSs on this topic, to see 
what people can come up with.  In terms of breaking existing rules, IIRC 
Chameleon, A New Religion (still an eternal draft), and the Mindrunning 
MRS (an even more eternal draft) use the psycho-social attributes to at 
least some degree.  Mindrunning (again, IIRC) mostly focuses on Identity 
(IDT), though, which will default back up to EGO or WIL in most cases. 
As for the other two, I don't think many special rules are invoked by 
the reality guides, so you could easily replace the existing 
psycho-social attributes with something else.  (Character creation in A 
New Religion would have to be tweaked, though.)

2. Attribute names

	I suppose that one of the eternal struggles between generic and 
genre-specific roleplaying systems is the problem of nomenclature: the 
very essence of the former lies in the idea of standardization, while 
the latter is (by definition) custom crafted for the "look and feel" of 
a particular genre.  AR is something of a compromise...it can be 
adjusted to meet the needs of differing game environments (because they 
_do_ differ), but a certain amount of rule structure and nomenclature 
remains constant so that players have some prayer of being able to move 
from one game world to another.  (And, for that matter, so that writers 
don't have to reinvent the wheel.  Otherwise, what's the point of a 
generic system?)  Like most compromises, this does run the risk of 
making all sides equally unhappy, but we've tried to create a 
sufficiently flexible structure so that it should be easy for dissenters 
to swap out the parts they don't like while retaining the rest.
	So, anyway, about those names: is there anything wrong with renaming 
them?  In principle not (hell, you can do anything you like), but I 
would personally advise restraint in this matter.  If every reality 
guide has a different name for "intelligence" (for instance), it's going 
to get very hard to tell what's going on.  Since attributes (unlike 
skills) are taken to be universal, renaming them at every pass runs the 
risk of obscuring the underlying similarity between game worlds, therby 
making things harder for players.  Of course, that's just my opinion, 
and (as has been emphasized by others) the basic game system is robust 
to such matters....

3. On the DRF as basic mechanism

	John Pierce mentions (and I agree) that the DRF is really the most 
important aspect of the system.  (And, really, it's not the particular 
form of the DRF itself, but a set of axioms it satisfies, but that's 
another matter.)  What Karim and I essentially wound up doing when we 
came up with this mechanism was reinventing something called the 
generalized linear model, which takes the form

   y = f( b0+b1*x1+b2*x2+... )

where y is some variable of interest, the bs are real-valued weights, 
the xs are real-valued predictors, and f is a function taking the reals 
onto the support of y.  (In a statistical context, one usually takes 
various of these to be random variables, but the basic idea is clear 
without this complication.)  For our purposes, we want y to be the 
probability of sucess -- the rolling target -- and the xs to be 
attributes, skills, modifiers, etc.  That way, we can work with simple, 
linear calculations, ont he one hand, but get clear-cut probabilities 
when it comes time to roll the dice.  To do this we have to find an f() 
which will take the reals into the (0,1) interval.  This is an old, 
solved problem, but we were young and naive, and solved it ourselves by 
cooking up the DRF.  The DRF (which is based on an arctangent function) 
is actually only one of many possible functions which could fill the 
proverbial bill, but we like it just the same.  Because it converges 
very slowly at the extremes (i.e., it has heavy tails), it prevents 
players from getting either a "sure thing" or a guaranteed failure very 
easily....at the same time, however, it's got close to a 1:1 
percentage/attribute payoff around 0, which is intuitively appealing.
	As Pierce notes, however, we could use something else.  Logit or probit 
models would be the obvious alternatives, and have the appealing side 
effect of allowing the zealous to easily "fit" AR attributes from 
empirical data using standard statistical packages.  (Wouldn't that be a 
great homework assignment for a stats class?  ;-))  These have very 
light tails, however, which means that convergence to sure 
thing/guaranteed failure will be sharper for sufficiently large 
attribute values.  Also, it is still possible to fit data using the 
existing DRF via heuristic maximum likelihood (not that I expect this to 
be a major concern for most people), so the statistical advantages may 
not be terribly important in practice.  <shrug>  Here again, experiment 
and see what you like.

4. Adjusting the precision of the DRF

	Altrichter Ferenc mentions the idea of implementing the game using d20s 
instead of d100s.  This is eminently doable: since the rules are 
continuous, even the percentile system is actually just an 
approximation, and you can use d1000s, d10s, d4s, or whatever, depending 
on the desired level of precision.  Here's what you need to do: treating 
everything else as-is, just create a new DRF table based on the scores 
for probabilities 0.05, 0.1, 0.15, ... 0.95, 1.0.  These correspond to 
rolls of 1 through 20, respectively, and you choose your rolling target 
by rounding up.  (If you like, you can go from 0 to 0.95 and round 
down...this might make more sense, actually, but it's up to you.)  The 
basic idea is that you are using the dice to generate uniform random 
numbers in the [0,1] interval and comparing them to the appropriate 
rolling target (as generated by the DRF)...the level of precision here 
is entirely up to you.  In test games, Karim usually uses a calculator 
or PDA with the DRF pre-programmed, and thus we get nearly exact values; 
this could be fun or overkill, depending on your perspective.  I think 
that percentile dice are a nice compromise, but this is really a matter 
of personal preference.

5. Fast action

	Perhaps there's a misstatement (it wouldn't be the first one) in the PRG, 
but IIRC fast action doesn't require a roll for each action.  You make 
one initiative roll per round, which gives you your "budget" of action 
counts to spend for until the next round.  Everything else is simply a 
matter of deducting the appropriate number of action counts from your 
budget when your number comes up...when everyone gets to 0 or below, the 
new round is declared.  (Of course, any debts from one round are 
deducted from your new ACs for the next one.)  One can debate whether we 
need these per-round initiative rolls (we could just let time flow 
continuously, mapping ACs to some set fraction of a second, and leave it 
at that), but you're certainly not supposed to roll with each _action_. 
  Hopefully this is a misunderstanding rather than a typo.  :-)

6. Too much calculation

	To get to the more general point of whether there is too much calculation 
as-is, this is always a tricky issue.  Of course, things should be kept 
as simple as possible, but OTOH there have been a lot of calls for fancy 
complex things (like bleeding and organ failure rules :-)) which are 
just not possible to implement in a realistic way without a little extra 
effort.  Still, I think there are things that can be readily 
improved/simplified.  Training and aging are a mess as-is, and need to 
be revised.  For this, I need to go try to dig up some good references 
on learning curves (does anyone have any?), and think a bit more about a 
viable random process model for attribute evolution.  The trick, of 
course, is to find a reasonable model which reduces to some sort of 
simple solution...the underlying theory can be very complex, so long as 
the answer which comes out is easy to implement.  After many discussions 
re: armor, we had a hard time coming up with something much simpler than 
the existing rules which was also very realistic; there's progress which 
could be made here, as well.  One frustration here is a lack of any real 
data on the subject.  I've got some stopping power statistics on modern 
weapons which can be used to calibrate the damage system (and I can 
fudge it for ancient weapons), but precious little on armor.  My 
engineering friends haven't been very helpful here, either, since their 
concensus seems to be that armor/weapon interactions are very 
complicated.  :-)  Barring further insight, then, I suppose we must go 
with what we have....

-----

OK, so there are some rambling reactions to some of the issues which 
have been raised.  I'm glad to see a new outbreak of discussion, and 
thanks, BTW, for the positive comments.  (Joe Mason's comment that he 
thought that the game was an elaborate joke at first doesn't count. 
;-))  I'll be interested in hearing what you folx come up with, though 
my participation in the discussion may be intermittent...I've got to fly 
out to give a talk on Tuesday, and I haven't finished preparing it yet. 
  :-(  But hey, the first draft of my thesis is done, and I have a job 
offer, so life isn't all bad.  If only I could think of a way to work AR 
development into an undergraduate course....

     -Carter