Wow, that is lightning quick. Much faster than what I came up with (video to be posted this week). I think this is the perfect example of optimizing for speed. You left your original code as reference and commentary and managed to come up with an obfuscated but equivalent and faster set of code. In this case, the speed was worth the extra time and effort it takes to maintain the code.

Thanks for a great solution!
Doug

 
function wetprob = puzzler

p = linspace(0.1,0.9,20); % prob. of rain
N = 1e6; % nr. of trips

wetcount = zeros(size(p)); % nr. of times the professor gets wet

for i = 1:numel(p)

    % initial state
    s = 1; % one umbrella available (can be 0, 1 or 2)
    w = 0; % nr. of times the professor gets wet

    q = p(i);
    for n = 1:N
        r = rand < q; % rain?
        w = w + r*(0.5*(s-3)*s + 1);
        s = (1-r)*abs(s-2) + r*(0.5*(1-s)*s + 2);

        %%% equivalent switch ... case structure
        % switch r
        %     case 0
        %         switch s
        %             case 0, s = 2;
        %             %case 1, s = 1;
        %             case 2, s = 0;
        %         end
        %     case 1
        %         switch s
        %             case 0, s = 2; w = w + 1;
        %             case 1, s = 2;
        %             case 2, s = 1;
        %         end
        % end
    end

    wetcount(i) = w;
end

wetprob = wetcount/N; % simulated probability of getting wet
teoprob = (1-p).*p./(3-p); % theoretical probability of getting wet

plot(p, wetprob, 'b--o', p, teoprob, 'r-.')
xlabel('p')
 

These are much better answers than I came up with. I knew there was cleaner math than what I used. Twenty minutes programming will save you two minutes in a library… :)

Later this week, I will post the Monte Carlo code optimization and a graph of the solution.

-Thanks,
Doug

PS, Tim- Pretty sure the professor wore polyester, most of mine did at least! :)

Is the professor upset at getting wet, and if so does he walk faster? Does he wear jeans or polyester; the latter taking longer to dry and feels worse when he gets wet? Is he bald or not? (Since it’s well know that a full head of hair increases the chance of rain, due to an increase static electricity discharged into the atmoshere).

Does the professor meet her true love at a rainy bus stop when a passerby offers her his own umbrella? If so, what is the probability that they will have four children, two boys and two girls, as compared with a different gender mix of four children? (Hint: assuming 4 children, the answer is not 50%).

Have to inked a contract for the book rights, and when will the DVD come out?

Inquiring MATLAB’ers Want to Know…

;-)

The way I did it was write an expression for q(0) as a function of p and q(2), and an expression for q(2) as a function of p and q(0), where q(x) is the probability of having x umbrellas (these expressions are both geometric series). I then substituted out q(2), giving an expression for q(0), which I then just multiply by p to give the answer. I wouldn’t call it elegant :)

 
syms p
A = [0 0 1;  %If the professor has no umbrella, he will for
             %sure end up with 2 at his destination
     0 1-p p; %If has has 1 umbrella, he will bring it
              %with probability p
    1-p p 0]; %If he has 2 umbrellas, he will bring one
              %of them with prob. p

%Calculate eigenvectors of the transpose
[V E] = eig(A.')

%Find eigenvector with eigenval. 1
ind = find( diag(E) == 1 )
Sol =  V(:,ind).'
%Normalize
Sol = Sol / (sum(Sol))

%Solution
Sol(1) * p
 

I get (1-p)/(3-p)*p as the analytical formula.

Nice solution, Petter.
Oliver