diff options
Diffstat (limited to 'python')
| -rw-r--r-- | python/exploration/dispersion.py | 45 | ||||
| -rw-r--r-- | python/notebooks/Dispersion.ipynb | 181 |
2 files changed, 208 insertions, 18 deletions
diff --git a/python/exploration/dispersion.py b/python/exploration/dispersion.py index fac383c5..b33f2796 100644 --- a/python/exploration/dispersion.py +++ b/python/exploration/dispersion.py @@ -104,9 +104,54 @@ def create_models(conn, df) -> (pd.DataFrame, float): return temp df["predict"] = df.groupby(["index", "series", "date"])["predict"].transform(aux) + df = df.assign( + mispricing=(df.exp_percentage - df.predict) + * df.index_expected_loss + / (df.detach_adj - df.attach_adj) + ) return (df, model) +def create_models_separate(df): + # Takes the output of get_tranche_data + model, calc = {}, {} + df = df.assign( + tranche_loss_per=(df.exp_percentage * df.index_expected_loss) + / (df.detach_adj - df.attach_adj) + ) + df = df.groupby(["date", "index", "series", "tenor", "attach"]).nth(-1) + for attach in df.index.get_level_values("attach").unique(): + calc[attach] = df.loc(axis=0)[:, :, :, "5yr", attach] + model[attach] = smf.ols( + "logit(tranche_loss_per) ~ " + "np.log(moneyness)* logit(gini) + " + "np.log(index_expected_loss)* logit(gini) + " + "np.log(index_duration) + " + "I(np.log(moneyness)**2) + I(np.log(moneyness)**3)", + data=calc[attach], + ).fit() + + calc[attach] = calc[attach].assign( + predict=expit(model[attach].predict(calc[attach])) + * (df.detach_adj - df.attach_adj) + / df.index_expected_loss + ) + + calc = pd.concat(calc, sort=False).reset_index(level=0, drop=True) + normalization = calc.groupby(["date", "index", "series", "tenor"])["predict"].sum() + calc = calc.merge( + normalization, left_index=True, right_index=True, suffixes=["_preN", "_sum"] + ) + calc["predict_N"] = calc["predict_preN"] / calc["predict_sum"] + calc["mispricing"] = ( + (calc["exp_percentage"] - calc["predict_N"]) + * calc["index_expected_loss"] + / (calc["detach_adj"] - calc["attach_adj"]) + * 100 + ) + return model, calc + + if __name__ == "__main__": index_type = "HY" series = 29 diff --git a/python/notebooks/Dispersion.ipynb b/python/notebooks/Dispersion.ipynb index e8e07f4a..0c3b518f 100644 --- a/python/notebooks/Dispersion.ipynb +++ b/python/notebooks/Dispersion.ipynb @@ -19,9 +19,7 @@ "from analytics import on_the_run\n", "from statsmodels.graphics.regressionplots import plot_fit\n", "from pygam import LinearGAM, s, f, GAM\n", - "from utils.db import dbengine\n", - "\n", - "serenitas_engine = dbengine('serenitasdb')" + "from utils.db import dbengine, dbconn" ] }, { @@ -40,9 +38,14 @@ "outputs": [], "source": [ "value_date = (datetime.datetime.today() - pd.offsets.BDay(1)).date()\n", - "start_date = datetime.date(2019,9,27)\n", - "end_date = datetime.date(2020,1,30)\n", - "index_type = 'HY'" + "start = (datetime.datetime.today() - pd.offsets.BDay(1) * 365 *4).date()\n", + "#end = (start + pd.offsets.BDay(1) * 365).date()\n", + "end = datetime.datetime.today()\n", + "index_type = 'IG'\n", + "risk = disp.get_tranche_data(dbconn(\"serenitasdb\"), index_type)\n", + "train_data = risk[start: end]\n", + "gini_calc, gini_model = disp.create_models(dbconn(\"serenitasdb\"), train_data)\n", + "gini_model.fit().summary()" ] }, { @@ -51,11 +54,18 @@ "metadata": {}, "outputs": [], "source": [ - "#Get Gini factor\n", - "date_range = pd.bdate_range(end=value_date, freq='5B',periods=52*4)\n", - "risk = disp.get_tranche_data(index_type, serenitas_engine)\n", - "risk = risk[risk.index.get_level_values(0).isin(date_range)]\n", - "gini_model, gini_calc = disp.create_models(risk, use_gini=True, use_log=True)" + "gini_calc.xs(31, level = 'series')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "#use trained model to fit rest of the data\n", + "f = gini_model.fit()\n", + "risk.loc[risk.index.get_level_values(\"attach\") != attach_max, \"predict\"] = expit(f.predict(bottom_stack))" ] }, { @@ -66,7 +76,7 @@ "source": [ "#Plot Gini if (use gini=True, use_log=False)\n", "to_plot_gini = gini_calc.xs(0, level='attach').groupby(['date', 'series']).nth(-1)\n", - "to_plot_gini['dispersion'].unstack().plot()" + "to_plot_gini['gini'].unstack().plot()" ] }, { @@ -75,9 +85,9 @@ "metadata": {}, "outputs": [], "source": [ - "#Today's mispricing\n", + "#last day: mispricing\n", "today = gini_calc.xs([value_date,33], level=['date','series'])\n", - "today[['exp_percentage', 'predict_N', 'predict_preN', 'mispricing']]" + "today[['exp_percentage', 'predict', 'mispricing']]" ] }, { @@ -88,7 +98,8 @@ "source": [ "#plot mispricing of a tranche through time \n", "attach = 0\n", - "to_plot = gini_calc.xs(attach, level='attach')['mispricing']\n", + "series = 33\n", + "to_plot = gini_calc.xs([attach, series], level=['attach', 'series'])['mispricing']\n", "to_plot.reset_index(['index','tenor'], drop=True).unstack().plot()" ] }, @@ -98,6 +109,15 @@ "metadata": {}, "outputs": [], "source": [ + "gini_calc.xs([attach, series], level=['attach', 'series']).to_clipboard()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ "#plot mispricing of series through time \n", "series = 33\n", "to_plot = gini_calc.xs(series, level='series')['mispricing']\n", @@ -119,6 +139,8 @@ "metadata": {}, "outputs": [], "source": [ + "to_csv('/home/serenitas/edwin/Python/temp1.csv')\n", + "\n", "gini_calc.to_csv('/home/serenitas/edwin/Python/' + index_type+ '_tranche_model.csv')" ] }, @@ -132,12 +154,36 @@ "tranche_returns = tdata.get_tranche_quotes(index=index_type)\n", "tranche_returns = tdata.tranche_returns(df=tranche_returns)\n", "attach = 0\n", - "t = tranche_returns['deladj_return'].reset_index(['index', 'tenor'], drop=True).xs(attach, level='attach')\n", + "t = tranche_returns['delhedged_return'].reset_index(['index', 'tenor'], drop=True).xs(attach, level='attach')\n", "temp={}\n", "for i,g in t.groupby('series'):\n", " temp[i] = (g.dropna()+1).cumprod()\n", "t = pd.concat(temp).reset_index(0, drop=True)\n", - "t.unstack(level='series').plot()" + "t.unstack(level='series').plot()\n", + "tranche_returns.to_csv('/home/serenitas/edwin/Python/temp3.csv')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "attach = 0\n", + "\n", + "returns = tranche_returns.xs(['HY', 29, '5yr', attach], level = ['index', 'series', 'tenor','attach'])['delhedged_return']\n", + "model = gini_calc.xs(['HY', 29, '5yr', attach], level = ['index', 'series', 'tenor','attach'])['mispricing']\n", + "returns = pd.merge(returns, model, left_index=True, right_index=True)\n", + "model_verification = smf.ols(\"delhedged_return ~ mispricing \", data=returns).fit()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "tranche_returns.xs(29, level='series').unstack(level='attach').to_csv('/home/serenitas/edwin/Python/temp1.csv')" ] }, { @@ -257,6 +303,105 @@ "source": [ "today, predict_HY33" ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "deltas = []\n", + "for s in portf.swaptions:\n", + " deltas.append(s.delta)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "#Old Model\n", + "value_date = (datetime.datetime.today() - pd.offsets.BDay(1)).date()\n", + "start = (datetime.datetime.today() - pd.offsets.BDay(1) * 365 *4).date()\n", + "#end = (start + pd.offsets.BDay(1) * 365).date()\n", + "end = datetime.datetime.today()\n", + "gini_model, gini_results = {}, {}\n", + "for index_type in ['HY', 'IG', 'EU', 'XO']:\n", + " risk = disp.get_tranche_data(dbconn(\"serenitasdb\"), index_type)\n", + " gini_model[index_type], gini_results[index_type] = disp.create_models_separate(risk)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "gini_model['HY'][0].summary()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "gini_results['HY']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "gini_results.to_csv('/home/serenitas/edwin/results.csv', header=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "for index_type in ['HY', 'IG', 'EU', 'XO']:\n", + " gini_results[index_type].to_csv('/home/serenitas/edwin/' + index_type + '_results.csv')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "to_plot = gini_results.xs(0, level='attach')['mispricing']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "to_plot.groupby(['date', 'index','tenor']).nth(-1).plot()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "gini_results.xs(31, level='series')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [] } ], "metadata": { @@ -275,7 +420,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.1-final" + "version": "3.8.1" } }, "nbformat": 4, |