diff options
Diffstat (limited to 'python/exploration/hybb.py')
| -rw-r--r-- | python/exploration/hybb.py | 67 |
1 files changed, 38 insertions, 29 deletions
diff --git a/python/exploration/hybb.py b/python/exploration/hybb.py index 3f3e7c97..9ea7b115 100644 --- a/python/exploration/hybb.py +++ b/python/exploration/hybb.py @@ -9,14 +9,17 @@ import pandas as pd from typing import Tuple from cds_rebook import default_adjustment + def get_defaulted_info(conn, auction_date: datetime.date) -> Tuple[int, str]: with conn.cursor() as c: - c.execute("SELECT id, seniority FROM defaulted WHERE auction_date=%s", - (auction_date,)) + c.execute( + "SELECT id, seniority FROM defaulted WHERE auction_date=%s", (auction_date,) + ) cid, seniority = c.fetchone() conn.commit() return cid, seniority + dr = pd.date_range("2015-09-27", "2019-11-25", freq=bus_day) prices = {} s = -1 @@ -30,48 +33,54 @@ for d in dr: index_list.append(BasketIndex("HYBB", s, ["5yr"], value_date=d)) for index in index_list: index.value_date = d - prices[(index.series, index.version, d)] = 100 * (1 - index.pv()['5yr']) + prices[(index.series, index.version, d)] = 100 * (1 - index.pv()["5yr"]) if index._lastdate == d.date(): cid, seniority = get_defaulted_info(conn, d.date()) - _, fee = default_adjustment(cid, seniority, index.maturities[0]) + accrual_days, fee = default_adjustment(cid, seniority, index.maturities[0]) weight = next(w for _, w, c in index.items() if c.defaulted) - old_price = 1 - index.pv()['5yr'] - new_price = (index.factor * old_price - weight * (1 - fee))/(index.factor - weight) + old_price = 1 - index.pv()["5yr"] + new_price = (index.factor * old_price - weight * (1 - fee)) / ( + index.factor - weight + ) prices[(index.series, index.version + 1, d)] = new_price * 100 serenitas_pool.putconn(conn) -hybb = pd.DataFrame.from_dict(prices, orient='index') +hybb = pd.DataFrame.from_dict(prices, orient="index") hybb.index = pd.MultiIndex.from_tuples(hybb.index) -hybb.index.names = ['series', 'version', 'date'] -hybb.columns = ['HYBB'] +hybb.index.names = ["series", "version", "date"] +hybb.columns = ["HYBB"] hybb = hybb.sort_index() -hybb = hybb.groupby(['series', 'version']).pct_change() -hybb['day_frac'] = hybb.groupby(["series", "version"]).transform( - lambda s: s.index.get_level_values('date').to_series().diff().astype("timedelta64[D]") / 360 +hybb = hybb.groupby(["series", "version"]).pct_change() +hybb["day_frac"] = hybb.groupby(["series", "version"]).transform( + lambda s: s.index.get_level_values("date") + .to_series() + .diff() + .astype("timedelta64[D]") + / 360 ) hybb["HYBB"] += hybb.day_frac * 0.05 -hybb = hybb.dropna().groupby('date').nth(-1) -df = index_returns(index="HY", tenor="5yr", series=[25, 26, 27, 28, 29, 30, 31, 32, 33], - from_date=datetime.date(2015, 9, 27), - years=None) +hybb = hybb.dropna().groupby("date").nth(-1) +df = index_returns( + index="HY", + tenor="5yr", + series=[25, 26, 27, 28, 29, 30, 31, 32, 33], + from_date=datetime.date(2015, 9, 27), + years=None, +) df = df["price_return"].reset_index(level=["index", "tenor"], drop=True) -hy_returns = df.dropna().groupby('date').nth(-1) -hy_returns.name = 'HY' +hy_returns = df.dropna().groupby("date").nth(-1) +hy_returns.name = "HY" -returns = pd.concat([hybb['HYBB'], hy_returns], axis=1) +returns = pd.concat([hybb["HYBB"], hy_returns], axis=1) # weekly -beta_weekly = (returns - .ewm(span=5) - .cov() - .groupby(level=0) - .apply(lambda df: df / np.diag(df))) +beta_weekly = ( + returns.ewm(span=5).cov().groupby(level=0).apply(lambda df: df / np.diag(df)) +) beta_weekly = beta_weekly.xs("HY", level=1)["HYBB"] ## monthly -beta_monthly = (returns - .ewm(span=20) - .cov() - .groupby(level=0) - .apply(lambda df: df / np.diag(df))) +beta_monthly = ( + returns.ewm(span=20).cov().groupby(level=0).apply(lambda df: df / np.diag(df)) +) beta_monthly = beta_monthly.xs("HY", level=1)["HYBB"] |