How to make a market that is complex but learnable

magic_market.py

import numpy as np
def make_stock(length=100, num_stocks=2):
    alpha = 0.9
    k = 2
    cov = np.random.normal(0, 5, [num_stocks, num_stocks])
    cov = cov.dot(cov.T) # This is a positive semidefinite matrix, e.g. a covariance matrix
    A = np.random.multivariate_normal(np.zeros(num_stocks), cov, size=[length]) # sample noise, with covariance 
    B = np.random.multivariate_normal(np.zeros(num_stocks), cov, size=[length]) # sample another noise, with covariance
    bs = [np.zeros(shape=num_stocks)] # 
    ps = [np.zeros(shape=num_stocks)] # The prices

    for a, b in zip(A, B):
        bv = alpha * bs[-1] + b # calculate some trend
        bs.append(bv)
        pv = ps[-1] + bs[-2] + k * a # Previosu price + previous trend factor, plus some noise
        ps.append(pv)

    #     ps = [0]
    #     for a,b,common in zip(A,BB,commonNoise):
    #         ps.append(ps[-1]+b+k*a+2*common)
    #     P = np.array(ps)
    #     P = np.exp(P/(P.max()-P.min()))
    ps = np.array(ps).T # reshape it so that its [length,stocks] 
    R = ps.max(1) - ps.min(1) # Scale factor
    prices = np.exp(ps.T / (R)) *np.random.uniform(10,250,num_stocks) # Normalize, exponantiate then make the prices more varied
    return prices