Re: call for duelmasters help

From: ben@abby.chem.ucla.edu

Date: Wed, 28 Oct 1992 13:52:41 +0000

In article <900667c.21.720139890@ace.acadiau.ca>, 900667c@ace.acadiau.ca (GERALD RICHARD CAMERON) writes:
>In article <18537@ucdavis.ucdavis.edu> itlm013@fella.ucdavis.edu (Captain Caveman) writes:
>>In article <00962831.716BD800@abby.chem.ucla.edu> ben@abby.chem.ucla.edu
>writes:
>
>>>IMHO, there are two main pieces of evidence AGAINST the luck stat:
>>
>>Sorry Ben, but here you're dead wrong.  RSI has admitted the existence of
>>a luck factor
>>rating.  Of course, we know skills exist,>> and I beleive in them...>>I
>>think the fact that it was a partial dump may tell you something.  Also,>I'm
>>sure that data inputters wouldn't have access to every single piece of>
>>sensitive information in the code.  Why should skill/luck/learning come up>
>>on a common screen dump?>
>>Certainly I believe that a lot of things attributed to the "luck" factor
>>are really just dice-rlls that can go either way.  However, as I stated abvoe,
>>there is a very good reason to include luck in the program.

Oo! Oo! Someone on my side...
(below, I mean, not JOhn...)
>
>Now, I know we argued about this months ago when I posted that little bit on
>skills (John, do you have that anywhere's?), but I really don't think there'
>s an actual luck stat per se.  This is because there's enough random factors
>in the game already that can make a warrior seem lucky or unlucky.  Also, a
>lot of warriors who it might seem are blessed with a high luck factor do go
>into slumps eventually.  If they were truly 'lucky' this wouldn't happen.
>You remember that slasher I kept talking about?  Well she went to the Isle
>at 19-2, and has gone 1-3 in AD, with an 0-3 in the April tourney.  Also,
>except for last summer's FtF, she never did well in tourneys period.  Now,
>if dhe had been blessed with ahigh luck stat, as her regular arena
>performances would indicate, she'd have gotten at least one TV along the
>line, in spite of her shitty abilities.  She was just like that while in
>regular DM.

Exactly my point.  In general, the sample size (i.e. number of fights) is far
to small to average out the random factors.  Hence the illusion of 'lucky' and
'unlucky' warriors.  One of the problems has to to with the factr that the
probababilities ARE NOT independent.  Assuming winning warriors learn (on avg.)
more skills/fight, the more you win, the more skills, the more you win, the
more skills, ad infinitum.  The opposite is also true, leading to the
conclusion, WITHOUT a luck 'stat', that winning warriors tend to win and learn
more; while loosers keep loosing...  In this way, a 'Dixie' with WT 11, IF they
are good enough to win lots, and 'lucky' enough to get higher thatn average
skills, they can still be a decent warrior, and even suceeed in ADM.  This is
probably not true for WT < 11, however, and the 'minimum' could be 13, or even
15.  So, to get back to the original argument I had with John, a bad 13 WT
warrior may learn 1 skill/turn and take 45 turns to get to ADM, while a good one
might learn >2/fight and become a decent warrior.  From this It would seem
logical to make LU, TP's and WS's with a lower MINIMUM WT, because these styles
win a lot.  All this is without assuming a 'luck' or 'learning' stat, or a
'lenbgth-of-fight' factor (my original propisition), assuming only a small
sample size, and a Win-learn factor.  This is not a proof, just an opionion
based on my (and also hersay) observations, using logical arguments.

John, you also mentioned a 19 WL warrior who missed their first stat train.  I
was 'told' that the 1st stat train was (WL*5%), giving your guy a 95% chance.
Missing this roll once, is unlikely, but haven't you ever fumbled in a %-based
game, (usually 5%..)  If on the other hand, if that SAME warrrior missed his 1st
train 12 times in a row, I would wonder. (odds = 1/20^12 !!!).  NO evidence for
a luck stat here...

To risk flogging a dead horse, the necessary sample size is inversly
proportional to the 'narrowness' of the distribution function around the mean
(presumed Gaussian).  This distribution (discrete, with values from 0-6,
representing the number of skill/fight for that specific fighter.) also varies
from fight to fight, depending on the relevent variables, whatever they may be.
I think, but I'm not a statistician, that these variables, with WT as a
constant, will always 'widen' this distribution, rather than 'narrow' it.  All
I'm trying to say, is that I think that the number of variables increases the
'streak factor' of a given warrior.  In Chemistry language, the 'streak factor'
is a cooperative transition...

Well, enough for now.

Ben
hitz@cuhhca.hhmi.columbia.edu

My idol:(sigh!)
>---------------------------------------------------------------------
>| Gerald Cameron - 900667c@ace.acadiau.ca                           |
>| 201 Eaton House, Acadia Univ., Wolfville, N.S., CANADA, B0P 1X0   |
>---------------------------------------------------------------------
>| Happiness is Love and the Freedom to express it.                  |
>---------------------------------------------------------------------

References

Re: call for duelmasters help
ben@abby.chem.ucla.edu
Tue, 20 Oct 1992 14:42:13 +0000
Re: call for duelmasters help
johnjohn@cco.caltech.edu (John White)
Tue, 20 Oct 1992 22:36:40 +0000
Re: call for duelmasters help
ben@abby.chem.ucla.edu
Fri, 23 Oct 1992 11:58:52 +0000
Re: call for duelmasters help
itlm013@fella.ucdavis.edu (Captain Caveman)
Mon, 26 Oct 1992 18:57:28 +0000
Re: call for duelmasters help
900667c@ace.acadiau.ca (GERALD RICHARD CAMERON)
Mon, 26 Oct 1992 22:51:31 +0000

Referenced By

Re: call for duelmasters help
ben@abby.chem.ucla.edu
Tue, 20 Oct 1992 14:42:13 +0000
Re: call for duelmasters help
johnjohn@cco.caltech.edu (John White)
Tue, 20 Oct 1992 22:36:40 +0000
Re: call for duelmasters help
ben@abby.chem.ucla.edu
Fri, 23 Oct 1992 11:58:52 +0000
Re: call for duelmasters help
itlm013@fella.ucdavis.edu (Captain Caveman)
Mon, 26 Oct 1992 18:57:28 +0000
Re: call for duelmasters help
900667c@ace.acadiau.ca (GERALD RICHARD CAMERON)
Mon, 26 Oct 1992 22:51:31 +0000