Book randomization

[Jack Spencer Jr]

Heya.

The Fighting Fantasy books worked pretty much like Choose-Your-Own-Adventure or Lone Wolf. You do have to hold your place if you use the randomizer on the pages, but that isn't a big deal. I know that in practice, most people probably just used six-sided dice the way I always just used a ten-sided die for Lone Wolf, but it was nice to have the option if you didn't have the dice handy.

[Jason Lee]

To solve the page flippy issue you can just use each result to seed the next one.  

Choose a random page.
The numeric result equals number of letters in the first word.
For the next result flip forward a number of pages equal to the last result.

It's atleast as random as a computer generated number seeded by the time.

[M. J. Young]

Just a random thought (pun not intended)--if you're using this idea for a sort of choose-your-own-adventure book, could you set it up such that at the beginning of play the player selects a number, and this is his word? Thus if the player takes seven, when he needs a die roll, it's always the length of the seventh word on the page.

You'd still have some trouble with the proliferation of two-letter words, and also the aspect that shorter words are more common generally than longer words--but as you were looking at an odd-or-even system, this might not matter.

--M. J. Young

[Jack Spencer Jr]

This is just my opinion, but I really hate the idea of counting the letters in a word to get a random number. How is this better in any way than just printing numbers on the corner of the page and rifling the pages to get a randome number? Not to mention then the bell curve gets weirdly skewed because of the average word length found in the English laguage -and- you would have to be careful not to use overlong words like antidisentarianism (18 letters) and lousing things up. That and it's not always so easy for everybody to count letters. I counted the letters in antidisentarianism four times and got a different answer all four times.

[Stuart DJ Purdie]

Ok, just ran a large set of text files through some custom code. (Total
284961 words)

Interesting stuff.  Firstly, the test corpus consists of English prose, dating over the past 10 years.  Shakespere gives different answears.  All facts below could be artifacts of selective loading in the corpus.  I've not properly eliminated that yet.  

First point, the word "can't" is 4 letters long, and "role-playing" is 11 letters long.

Interestingly,  3 letter words are more common than 2.  Surprised me too.  

Other than that, and possible 1 letter words (I ignored them off the bat), there is a decreasing exponential curve - approximatly frequency(WordLen) = exp(-0.38 * WordLen).

If you take just Odd/Even as your yardstick, then if the shortest word you tak is 2 lettered, then you get a 47/53 split, in favour of even. If you start at 3 letters words, you end up with a 42/58 split, in favour of odd.  On the basis of this, I'd reccomend against anything using odd/even word lengths.  Shame, it's otherwise a very elegant system.

If you wanted to to a target number type mechanism, that's doable.  The length of the word you end up picking is your 'roll'.  The cumulative frequency is quite similarly behaved to an exploding die.  I've stuck a graph up http://wired.st-and.ac.uk/~spurdie/wfreq.svg, or http://wired.st-and.ac.uk/~spurdie/wfreq.png in PNG format.  They show the chance of a random word having at least a minimum number of letters.  Blend this with cruciels plan of using one result to seed the next, and that's a workable solution.

Note that the choice of book can skew this a lot.  A text book would tend to have longer words, with biochemistry really showing something quite different.  However, that's easily balanced off by using one single book for all randomisation, PC and NPC.  Use Sweet Valley High for gritty hardcore,  average SF book for middle of the road, and a biochem textbook for over the top cinematic (Assuming a higher is better mechanic).