def TestCaseWithFit():
from AprFit import AprFit
from winsound import Beep
# Pull the stock data into memory.
stocks = []
for symbol in symbols:
stock = Stock.ShyRetrieve(symbol=symbol, minDate=(datetime.datetime.now() - datetime.timedelta(days=5)))
if len(stock._history) > 0:
stocks.append(stock)
else:
print(f"{stock.symbol} no history")
Beep(300, 500) # Let the user know it's done importing
today = datetime.datetime.now()
# In a loop so data stays in memory if the user wants to change parameters.
while True:
yearsToConsider = AskYears()
if yearsToConsider == "break":
break
scores = []
startDate = today - datetime.timedelta(seconds = 366 * yearsToConsider * 60 * 60 * 24)
for stock in stocks:
if stock.history[0].date > startDate:
continue
# Only plot stocks which have enough history.
try:
apr_fit = stock.get_apr_fit(years=yearsToConsider)
except Exception as ex:
# Looks like I need to do some sanitizing of the data from yFinance. BEP hits a math domain error here and previously gave strange results in other tests.
print(f"Failed to fit curve onto data from {stock.symbol}: {ex}")
continue
# Determine how much this stock is currently overvalued/undervalued
total_dividends_since_t_0 = 0.
index = len(stock.history) - 1
while (stock.history[index].date - today).total_seconds() >= apr_fit.t_0 * 31557600.:
total_dividends_since_t_0 += stock.history[index].dividend
if index == 0:
break
index -= 1
undervalue = apr_fit.y_0 * (1 + apr_fit.rate) ** (0. - apr_fit.t_0) - (stock.history[-1].price + total_dividends_since_t_0)
undervalue /= stock.history[-1].price + total_dividends_since_t_0
N_STDEVS = 2. # Number of standard deviations to use for fitness
if apr_fit.stdev * N_STDEVS > 1.:
score = Score(stock, 0.)
else:
score = Score(stock, (apr_fit.rate + undervalue / 4.) * (1. - N_STDEVS * apr_fit.stdev) + 1. / stock.pe_ratio)
# stdev applies as both uncertainty in the fit and as potential loss
if math.isnan(score.score):
print(f"Failed to calculate score for {stock.symbol}")
continue
# Boost the investments we want to bias towards
if stock.symbol in boost:
score.score *= 1.5
# Nerf the stocks we don't like
if stock.symbol in nerf:
score.score *= 0.5
# Attach metadata to the score so we can print it.
score.rate = 100. * apr_fit.rate
score.undervalue = 100. * undervalue
score.uncertainty = 100. * apr_fit.stdev
scores.append(score)
scores.sort(key=lambda x:x.score)
scores.reverse()
# Recommend how much would have been good to allocated to each
number_of_recommendations = 0
sum = 0
for candidate in scores:
if math.isnan(candidate.score):
continue
if candidate.score <= 0.:
continue
sum += candidate.score
if candidate.score / sum < 0.05:
sum -= candidate.score
break
number_of_recommendations += 1
# Don't recommend >25% in any one asset
LIMIT = 0.25
if scores[0].score > sum * LIMIT:
redistribution = (scores[0].score - sum * LIMIT) / (number_of_recommendations - 1)
for score in scores:
score.score += redistribution
if scores[1].score > sum * LIMIT:
redistribution = (scores[1].score - sum * LIMIT) / (number_of_recommendations - 2)
for score in scores:
score.score += redistribution
scores[1].score = sum * LIMIT
scores[0].score = sum * LIMIT
print("Recommended distribution:")
for i in range(number_of_recommendations):
stock = scores[i].stock
recommended_percent = 100 * scores[i].score / sum
print(f"{recommended_percent:.2f}% in {stock.symbol} Score: {scores[i].score:.2f} P/E: {stock.pe_ratio:.2f} <APR>: {scores[i].rate:.2f} undervalued by {scores[i].undervalue:.2f}% uncertainty: {scores[i].uncertainty:.2f}%")
if plot:
stock.get_apr_fit(years=yearsToConsider, plot=True)
Beep(300, 500)
def OriginalTestCase():
# Import the modules now so it will crash before doing work if a module is missing.
import matplotlib.pyplot as plt
from dateutil.parser import parse
# Pull the stock data into memory.
stocks = []
for symbol in symbols:
stock = Stock.ShyRetrieve(symbol=symbol, minDate=(datetime.datetime.now() - datetime.timedelta(days=5)))
if len(stock._history) > 0:
stocks.append(stock)
today = datetime.datetime.now()
# Let the user know import is complete
import winsound
winsound.Beep(600, 100)
# In a loop so data stays in memory if the user wants to change parameters.
while True:
yearsToConsider = AskYears()
minimumYield = AskMinYield()
if yearsToConsider == "break":
break
if minimumYield == "break":
break
startDate = today - datetime.timedelta(days = 365 * yearsToConsider)
if plot:
print("Plotting...")
fig, subplots = plt.subplots(nrows=2, ncols=1)
dividend_plot = subplots[0]
#growth_plot = subplots[1]
total_yield_plot = subplots[1]
dividend_plot.set_ylabel("Dividend (%/yr)")
dividend_plot.set_xlim(left=startDate, right=today)
max_dividend = 0.
min_dividend = 0.
#growth_plot.set_ylabel("Growth Since (%/yr inflation adjusted)")
#growth_plot.set_xlim(left=startDate, right=today)
max_growth = 0.
min_growth = 0.
total_yield_plot.set_ylabel("Total Yield (%/yr)")
total_yield_plot.set_xlim(left=startDate, right=today)
max_total = 0.
min_total = 0.
annual_yields = []
scores = []
for stock in stocks:
if stock.history[0].date > startDate:
continue
# Only plot stocks which have enough history.
filteredGrowth, growthUncertainty = stock.GrowthAPRWithUncertainty(yearsToConsider)
dividendYield = stock.AverageDividendPercent(yearsToConsider)
dividendUncertainty = stock.DividendPercentUncertainty(yearsToConsider)
if dividendUncertainty > dividendYield:
dividendUncertainty = dividendYield # Never apply a penalty for dividends
annualYield = dividendYield + filteredGrowth
if annualYield < minimumYield:
continue
# Only plot stocks with combined margin better than 10% per year over the past 10 years.
print(f"{stock.symbol} {annualYield:.1f} ± {growthUncertainty + dividendUncertainty:.1f}% average annual yield over the past {yearsToConsider} years.")
print(f"{filteredGrowth:.2f}% from growth and {stock.AverageDividendPercent(yearsToConsider):.2f}% from dividends.")
annual_yields.append((stock.symbol, annualYield))
score = annualYield - 0.431 * (growthUncertainty + dividendUncertainty) # 33rd percentile
scores.append((stock.symbol, score))
if plot:
dates = []
relGrowth = []
divYieldPercent = []
latest = stock.history[-1]
for snapshot in stock.history:
if snapshot.date < startDate:
continue
if snapshot.price == 0.0:
continue
if (latest.date - snapshot.date).days == 0.0:
continue
annual_growth = 100. * (latest.price / snapshot.price - 1.) * 1.02 ** ((snapshot.date - latest.date).days / 365.25) / \
((latest.date - snapshot.date).days / 365.25)
if annual_growth > 1. * (latest.date - snapshot.date).days:
annual_growth = 1. * (latest.date - snapshot.date).days
if annual_growth < -1. * (latest.date - snapshot.date).days:
annual_growth = -1. * (latest.date - snapshot.date).days
dates.append(snapshot.date)
relGrowth.append(annual_growth)
divYieldPercent.append(100. * snapshot.annualDividend / snapshot.price)
if relGrowth[-1] > max_growth:
max_growth = relGrowth[-1]
elif relGrowth[-1] < min_growth:
min_growth = relGrowth[-1]
if divYieldPercent[-1] > 100.:
print(f"{dates[-1]}: {stock.symbol} has {divYieldPercent[-1]}% dividend?")
print("You may not want to trust this data...")
if divYieldPercent[-1] > max_dividend:
max_dividend = divYieldPercent[-1]
elif divYieldPercent[-1] < min_dividend:
min_dividend = divYieldPercent[-1]
dividend_plot.plot(dates, divYieldPercent, label=stock.symbol)
#growth_plot.plot(dates, relGrowth, label=stock.symbol)
annualYield = []
yDates = []
for index in range(200, len(stock.history)):
snapshot = stock.history[index]
if snapshot.date < startDate:
continue
priceLastYear = stock.history[index - 200].price
if priceLastYear == 0.0:
continue
yDates.append(snapshot.date)
annualYield.append(100. * ((snapshot.price - priceLastYear) + snapshot.annualDividend) / priceLastYear)
if annualYield[-1] > max_total:
max_total = annualYield[-1]
elif annualYield[-1] < min_total:
min_total = annualYield[-1]
if plot:
total_yield_plot.plot(yDates, annualYield, label=stock.symbol)
annual_yields.sort(key=lambda x:x[1])
annual_yields.reverse()
scores.sort(key=lambda x:x[1])
scores.reverse()
# Recommend how much would have been good to allocated to each
number_of_recommendations = 0
sum = 0
for candidate in scores:
if not math.isnan(candidate[1]):
sum += candidate[1]
if sum > 0.:
if candidate[1] / sum < 0.05:
sum -= candidate[1]
break
if sum < 0.:
sum -= candidate[1]
break
number_of_recommendations += 1
print("Recommended distribution:")
recommendations = []
for i in range(number_of_recommendations):
symbol = scores[i][0]
recommended_percent = 100 * scores[i][1] / sum
recommendations.append((symbol, recommended_percent))
print(f"{recommended_percent:.2f}% in {symbol} Score: {scores[i][1]:.2f}")
winsound.Beep(300, 500)
if plot:
fig.tight_layout()
fig.autofmt_xdate()
#growth_plot.plot([parse("1/1/2000"), today], [0, 0], label='Zero')
total_yield_plot.plot([parse("1/1/2000"), today], [0, 0], label='Zero')
dividend_plot.set_ylim(bottom=min_dividend, top=max_dividend)
#growth_plot.set_ylim(bottom=min_growth, top=max_growth)
total_yield_plot.set_ylim(bottom=min_total, top=max_total)
dividend_plot.tick_params(labelleft=True, left=True, right=True, bottom=True)
#growth_plot.tick_params(labelleft=True, left=True, right=True, bottom=True)
total_yield_plot.tick_params(labelleft=True, left=True, right=True, bottom=True, labelbottom=True)
dividend_plot.legend()
#growth_plot.legend()
#total_yield_plot.legend()
plt.show()