def plot_sharpe_ratios(data, w_vec, gammas, show=False):
for i, gamma in enumerate(gammas):
plt.plot(gamma, compute_scores(data, w_vec[i])[2], 'bo')
plt.xlabel('Gamma')
plt.ylabel('Sharpe Ratio')
if show:
plt.show()
def plot_equally_weighted(data, show=False):
N = len(data)
equal_w_vec = [1 / float(N)] * N
mean, std, sharpe = compute_scores(data, equal_w_vec)
plt.plot(std, mean, 'co')
if show:
plt.show()
def plot_individual_assets(data, show=False):
for i in range(len(data)):
new_w_vec = [0] * len(data)
new_w_vec[i] = 1
mean, std, sharpe = compute_scores(data, new_w_vec)
plt.plot(std, mean, 'ro')
plt.annotate('coin {}'.format(i), xy=(std, mean))
if show:
plt.show()
def test_scores(prices):
"""DEPRECATED"""
N = len(prices)
results = []
for i in range(N):
new_w_vec = [0] * N
new_w_vec[i] = 1
results.append([i] + list(compute_scores(prices, new_w_vec)))
# MPT
gammas = np.logspace(-2, 2, num=100)
w_vec_results, ret_results, risk_results = mpt_opt(prices, gammas)
results.append(['MVO'] + list(compute_scores(prices, w_vec_results[0])))
# Market Cap
mcap_vec = mktcap_opt(load_marketcaps())
results.append(['MCAP'] + list(compute_scores(prices, mcap_vec)))
# Equal Weights
eq_w_vec = [1 / float(N)] * N
results.append(['EQ'] + list(compute_scores(prices, eq_w_vec)))
sorted_results = sorted(results, key=lambda x: -x[-1])
for result in sorted_results:
print(result)
def train(self, start_index=None, end_index=None):
sliced_prices = slice_index_params(self.prices, start_index, end_index)
gammas = np.logspace(-2, 2, num=100)
w_vec_results, ret_results, risk_results = mpt_opt(sliced_prices, gammas)
# This is slow - bisect so it runs in log(N) time
sharpe_ratios = [compute_scores(sliced_prices, w_vec_results[i])[-1]
for i, gamma in enumerate(gammas)]
max_sharpe_ratio = max(sharpe_ratios)
max_sharpe_ratio_index = sharpe_ratios.index(max_sharpe_ratio)
self.weights = w_vec_results[max_sharpe_ratio_index]
self.gamma = gammas[max_sharpe_ratio_index]
self.trained_sharpe_ratio = max_sharpe_ratio
return (self.weights,
self.gamma,
ret_results[max_sharpe_ratio_index],
risk_results[max_sharpe_ratio_index],
self.trained_sharpe_ratio)
def test(self, start_index=None, end_index=None):
sliced_prices = slice_index_params(self.prices, start_index, end_index)
return compute_scores(sliced_prices, self.weights)
def plot_marketcap_weighted(data, show=False):
mktcap_w_vec = mktcap_opt(load_marketcaps())
mean, std, sharpe = compute_scores(data, mktcap_w_vec)
plt.plot(std, mean, 'mo')
if show:
plt.show()