Friday, February 21, 2014

design factoid: humans do not intuit probabilities or statistics very well at all ; no math version


We see and choose based on perceived and thus not quantified extremes, not the highest probability; in other words, what we want, or what we fear has a BIG influence on how we think about chance.  Thus, State lotteries, thus, too, much superstitious behavior; also, much criticism of game design (see, actual relevance !).

Normally this works well enough, since the universe of avoiding cave bears and such is far from perfectly random and probabilistic, and there is a big bias represented by getting it wrong(eaten) simply based on choosing the average return. So we maximize the best outcomes, and avoid the worst, regardless of likelihood.  But, when it is highly probabilistic, and we don’t crunch the actual numbers, this don't work so good.  


The left side of the probability distribution for most of Human history
 
 Thus, looking at a probability tree (such as  in Traveller chargen), and deciding “Yuck, chances are I’ll get a one skill wonder or die before I play” Is usually waaaaay wrong.  *
Bottom line: it isn’t rocket science, but it isn’t as easy as speaking. It’s more like reading.   Barring accident or mutation, everybody learns to speak eventually, just by virtue of hardwiring; reading takes work.  But almost anyone can get there.  Rocket science, apparently not so easy, given how often Rocket scientists incinerate themselves and the local area code.



So, let’s look at Traveller more (any excuse will do):  original traveller handled Random/allocated chargen on a direct and a meta level.  Mostly random, but with a big dollop of allocation at a leverage point. Your stats were random, and your character history was dice ruled after you pick the career you are going to play.   And that choice was absolutely critical – most stat profiles could find a good fit in a couple of careers, so you have some choice –however, a bad fit made it very unlikely that you would have a successful resolution before you started play. 
Originally, this meant that you either “mustered out” (were fired) early, with few skills and benefits, or Died.  After the first edition, the vast beeping and squeeking about dying before play became so loud that even without the internet opinion amplipolarizer, that got changed to “fired with consequences” .  But still.   
If you’ve actually sat down and generated tons of Classic style characters (LBB1-3 + COTI for those who know) you find out that, yes indeed, choosing a good match for career and stats does give you a fairly good range of options, and boosts your survival (or non-termination with prejudice) significantly.  And, well, Scouts always die.  But that is on the label.
So, while it looks insanely stochastic (see! real technical jargon) it isn’t even close to as random as the armchair Pseudotravellers (those who only read or skim  the rules before posting and/ or making up their mind)  would have you believe.
 
This relates to the next topic: random vs purchase chargen design in rules lite systems (like, say, BAGS)
*For instance, did you know that by fiddling with the reenlistment roll typically to make it easier, one can actually increase your likelihood of dying ? You can, trust me –and there is a sweet spot that optimizes the two positive outcomes (survival and retention) . Interestingly, the tables are often already there.




1 comment:

Ed Allen said...

That's an amusing set of labels. And I like the cavebear eating the low end of the distribution as the long tail shows its heels.