def __init__(self,
strategy: Strategy,
tie_breaker: Optional[TieBreaker] = None,
isins: Optional[Iterable[str]] = None,
num_portfolio: Optional[int] = None,
hold_interval=None,
buy_sell_gap=None):
# apply default values
if tie_breaker is None:
from lib.simulate.tiebreaker.max_returns_tie_breaker import MaxReturnsTieBreaker
tie_breaker = MaxReturnsTieBreaker()
num_portfolio = num_portfolio or self.DEFAULT_NUM_PORTFOLIO
hold_interval = hold_interval or self.DEFAULT_HOLD_INTERVAL
buy_sell_gap = buy_sell_gap or self.DEFAULT_BUY_SELL_GAP
funds = fund_cache.get(isins)
self._strategy = strategy
self._tie_breaker = tie_breaker
self._num_portfolio = num_portfolio
self._hold_interval = hold_interval
self._buy_sell_gap = buy_sell_gap
# computed properties
self._prices_df = fund_cache.get_prices(isins)
self._fees_df = calc_fees(funds)
self._last_valid_date = self._prices_df.last_valid_index()
self._broadcast_data(
Simulator.Data(prices_df=self._prices_df,
fees_df=self._fees_df,
num_portfolio=self._num_portfolio,
hold_interval=self._hold_interval))
def update_similar_funds():
funds = fund_cache.get()
corr_df = fund_cache.get_corr()
fees_df = calc_fees(funds)
def to_similar_funds_entry(isin: str,
corr_series: pd.Series) -> SimilarFundsEntry:
similar_isins = corr_series[corr_series > THRESHOLD].index.tolist()
[fund] = fund_cache.get([isin])
fees_per_year = fees_df.at[isin, "total_one_off_fees"] + fees_df.at[
isin, "total_annual_fees"]
returns_per_year = fund.returns["1Y"]
if returns_per_year:
after_fees_return = returns_per_year - fees_per_year
else:
after_fees_return = None
return SimilarFundsEntry(isin=isin,
similar_isins=similar_isins,
after_fees_return=after_fees_return)
similar_funds = [
to_similar_funds_entry(isin, corr_series)
for isin, corr_series in corr_df.iterrows()
]
post_similar_funds(similar_funds)
def predict(self, dt: Optional[date] = None) -> Prediction:
"""
Generates prediction at given date, or the last valid date if not supplied.
:param dt: Date of prediction. Defaults to last valid date.
:return:
"""
if dt is None:
dt = self._last_valid_date
allowed_isins = self._strategy.run(dt, self._prices_df, self._fees_df)
max_isins = self._tie_breaker.run(allowed_isins, self._num_portfolio,
dt, self._prices_df, self._fees_df)
return Simulator.Prediction(date=dt, funds=fund_cache.get(max_isins))
def to_similar_funds_entry(isin: str,
corr_series: pd.Series) -> SimilarFundsEntry:
similar_isins = corr_series[corr_series > THRESHOLD].index.tolist()
[fund] = fund_cache.get([isin])
fees_per_year = fees_df.at[isin, "total_one_off_fees"] + fees_df.at[
isin, "total_annual_fees"]
returns_per_year = fund.returns["1Y"]
if returns_per_year:
after_fees_return = returns_per_year - fees_per_year
else:
after_fees_return = None
return SimilarFundsEntry(isin=isin,
similar_isins=similar_isins,
after_fees_return=after_fees_return)
def _to_events(cls, account: pd.DataFrame) -> List[Event]:
events = []
for i in range(len(account.index) - 1):
next_isins = account.iloc[i + 1, :]["isins"]
if next_isins:
next_funds = fund_cache.get(next_isins)
events.append(Event(
type="fund",
from_date=account.index[i],
to_date=account.index[i + 1],
holdings=[
Holding(
isin=fund.isin,
name=fund.name,
sedol=fund.sedol,
weight=1 / len(next_isins)
) for fund in next_funds
],
pct_change=(account["value"].iat[i + 1] / account["value"].iat[i]) - 1
))
return events
def test_get():
funds = fund_cache.get()
assert len(funds) > 3000