Modified cumprod using bsxfun?

I am trying to generate Monte Carlo simulations whereby my sample paths are cumulative products. However, I want each iteration step to be allowed to be incremented/decremented only by multiples of a specified minimum value.

The simplest example I can think of to illustrate this is to simulate roulette betting from the casino's perspective. This is the fastest vectorized example I can think of if there was no restriction on the minimum increments:

    % Inputs
    initialCapital = 1e6;
    betFraction = 0.0015;
    numBets = 1000;
    numSamplePaths = 1000;
    winProbability = 37/38;
    winAmount = 1;
    loseAmount = 30;

    % Simulation
    spins = sign(rand(numBets, numSamplePaths) - 1 + winProbability);
    profitFactor = spins;
    profitFactor(spins>=0) = 1 + winAmount*betFraction*spins(spins>=0);
    profitFactor(spins<0) = 1 + loseAmount*betFraction*spins(spins<0);
    accumulatedCapital =  cumprod(profitFactor) * initialCapital;

However, in a realistic situation, the bet sizes must be multiples of a fixed value, let's say it's $1 - so all of my elements in the accumulatedCapital array should be integers. Note that this is different from saying that:

   accumulatedCapital = floor(cumprod(profitFactor) * initialCapital);

Of course, the easiest way is simply to use a loop twice over:

    if spin(n) >= 0
      accumulatedCapital(n) = ...
      accumulatedCapital(n-1) + ...
      winAmount*floor(betFraction(n)*accumulatedCapital(n-1))

    elseif spin(n) < 0
      accumulatedCapital(n) = ...
      accumulatedCapital(n-1)  - ...
      loseAmount*floor(betFraction(n)*accumulatedCapital(n-1))

but this is much slower. Any of you can suggest some vectorized wizardry to solve this? Thanks!