TQQQ For The Long Term Minimal v1.4 backtest using synthetic LETFs

This a backtest of the aforementioned strategy using synthetically generated LETFs data that allows us to have valid and accurate simulated close prices, dating all the way back to the inception date of the corresponding unleveraged ETF they are based on. Real and simulated data are combined together to form a longer time-series.

import pandas as pd
import yfinance as yf
import pandas_ta as ta
import plotly.express as px
from IPython.display import IFrame

# Import the LETFs data
data = pd.read_csv("./data/extended-leveraged-etfs.csv", index_col=0, parse_dates=True)
data

	SSO	UPRO	SPXL	SH	SDS	SPXS	QLD	TQQQ	PSQ	QID	...	TYO	UBT	TMF	TBF	TBT	TMV	UGL	GLL	UVXY	TLT
Date
1986-05-19	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	0.69170	NaN	NaN	NaN	NaN	NaN	NaN	10.60539
1986-05-20	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	0.68715	NaN	NaN	NaN	NaN	NaN	NaN	10.58417
1986-05-21	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	0.70741	NaN	NaN	NaN	NaN	NaN	NaN	10.69021
1986-05-22	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	0.70279	NaN	NaN	NaN	NaN	NaN	NaN	10.66902
1986-05-23	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	0.70658	NaN	NaN	NaN	NaN	NaN	NaN	10.69021
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2022-10-12	39.42	27.97	52.77	17.30	54.48	29.31	35.26	18.10	15.14	28.14	...	13.84	24.81	8.06000	22.69	33.01	141.25999	47.6000	36.62	13.25000	100.35000
2022-10-13	41.49	30.14	56.85	16.85	51.63	27.04	36.86	19.31	14.79	26.89	...	14.03	24.32	7.85000	22.93	33.63	145.06000	46.9100	37.12	12.69000	99.39000
2022-10-14	39.60	28.08	52.98	17.23	53.93	28.87	34.62	17.57	15.25	28.51	...	14.20	23.97	7.67000	23.12	34.19	148.89000	45.7400	38.08	13.14000	98.57000
2022-10-17	41.61	30.21	57.00	16.79	51.22	26.66	36.97	19.30	14.72	26.56	...	14.18	23.68	7.53000	23.26	34.61	151.49001	45.9700	37.84	12.61000	98.09000
2022-10-18	42.57	31.11	58.99	16.59	49.96	25.73	37.55	19.78	14.59	26.17	...	14.13	23.80	7.59000	23.21	34.40	150.53999	46.1852	37.72	12.21000	98.32000

9180 rows × 43 columns

We now download the additional unleveraged ETFs required for this specific strategy backtest, and add them to the dataframe. It's possible to generate synthetic versions of these unleveraged ETFs too (there's already TLT in the data above, for example), but the results might be less accurate depending on what index you choose to generate the returns.

# Download the additional unleveraged ETFs required for this specific strategy backtest
unleveraged_etfs = yf.download(["SPY", "QQQ"])["Adj Close"]
unleveraged_etfs.index = unleveraged_etfs.index.tz_localize(None) # remove time info while keeping index type as datetime
# Add the unleveraged ETFs to the main dataframe
merged_df = pd.concat([data, unleveraged_etfs], axis=1)
merged_df

[*********************100%***********************]  2 of 2 completed

	SSO	UPRO	SPXL	SH	SDS	SPXS	QLD	TQQQ	PSQ	QID	...	TMF	TBF	TBT	TMV	UGL	GLL	UVXY	TLT	QQQ	SPY
Date
1986-05-19	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	0.69170	NaN	NaN	NaN	NaN	NaN	NaN	10.60539	NaN	NaN
1986-05-20	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	0.68715	NaN	NaN	NaN	NaN	NaN	NaN	10.58417	NaN	NaN
1986-05-21	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	0.70741	NaN	NaN	NaN	NaN	NaN	NaN	10.69021	NaN	NaN
1986-05-22	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	0.70279	NaN	NaN	NaN	NaN	NaN	NaN	10.66902	NaN	NaN
1986-05-23	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	0.70658	NaN	NaN	NaN	NaN	NaN	NaN	10.69021	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2022-10-13	41.49	30.14	56.85	16.85	51.63	27.04	36.86	19.31	14.79	26.89	...	7.85000	22.93	33.63	145.06000	46.9100	37.12	12.69000	99.39000	268.820007	365.970001
2022-10-14	39.60	28.08	52.98	17.23	53.93	28.87	34.62	17.57	15.25	28.51	...	7.67000	23.12	34.19	148.89000	45.7400	38.08	13.14000	98.57000	260.739990	357.630005
2022-10-17	41.61	30.21	57.00	16.79	51.22	26.66	36.97	19.30	14.72	26.56	...	7.53000	23.26	34.61	151.49001	45.9700	37.84	12.61000	98.09000	269.350006	366.820007
2022-10-18	42.57	31.11	58.99	16.59	49.96	25.73	37.55	19.78	14.59	26.17	...	7.59000	23.21	34.40	150.53999	46.1852	37.72	12.21000	98.32000	271.480011	371.130005
2022-10-19	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	270.489990	368.500000

9181 rows × 45 columns

We now get a subset of this dataframe with only the tickers that are relevant for this particular strategy, and we do a sanity check to make sure there are no issues.

strategy_symbols = ["SPY", "TQQQ", "SPXL", "UVXY", "SQQQ", "TLT", "QQQ"]
strategy_data = merged_df[strategy_symbols].copy()
strategy_data

	SPY	TQQQ	SPXL	UVXY	SQQQ	TLT	QQQ
Date
1986-05-19	NaN	NaN	NaN	NaN	NaN	10.60539	NaN
1986-05-20	NaN	NaN	NaN	NaN	NaN	10.58417	NaN
1986-05-21	NaN	NaN	NaN	NaN	NaN	10.69021	NaN
1986-05-22	NaN	NaN	NaN	NaN	NaN	10.66902	NaN
1986-05-23	NaN	NaN	NaN	NaN	NaN	10.69021	NaN
...	...	...	...	...	...	...	...
2022-10-13	365.970001	19.31	56.85	12.69000	59.11	99.39000	268.820007
2022-10-14	357.630005	17.57	52.98	13.14000	64.38	98.57000	260.739990
2022-10-17	366.820007	19.30	57.00	12.61000	58.00	98.09000	269.350006
2022-10-18	371.130005	19.78	58.99	12.21000	56.59	98.32000	271.480011
2022-10-19	368.500000	NaN	NaN	NaN	NaN	NaN	270.489990

9181 rows × 7 columns

strategy_data.info()

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 9181 entries, 1986-05-19 to 2022-10-19
Data columns (total 7 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   SPY     7486 non-null   float64
 1   TQQQ    5943 non-null   float64
 2   SPXL    8769 non-null   float64
 3   UVXY    8264 non-null   object 
 4   SQQQ    5943 non-null   float64
 5   TLT     9180 non-null   float64
 6   QQQ     5944 non-null   float64
dtypes: float64(6), object(1)
memory usage: 573.8+ KB

UVXY is currently of dtype object. Let's try to convert the column to float64, so we can use it in our calculations.

strategy_data["UVXY"] = strategy_data["UVXY"].astype("float64")
strategy_data["UVXY"].info()

<class 'pandas.core.series.Series'>
DatetimeIndex: 9181 entries, 1986-05-19 to 2022-10-19
Series name: UVXY
Non-Null Count  Dtype  
--------------  -----  
8264 non-null   float64
dtypes: float64(1)
memory usage: 143.5 KB

We now create the signals columns we need, using built-in pandas methods or pandas-ta.

# Signals
strategy_data["SPY_MA200"] = ta.sma(strategy_data["SPY"], length=200)
strategy_data["TQQQ_RSI10"] = ta.rsi(strategy_data["TQQQ"], length=10)
strategy_data["SPXL_RSI10"] = ta.rsi(strategy_data["SPXL"], length=10)
strategy_data["SPY_RSI10"] = ta.rsi(strategy_data["SPY"], length=10)
strategy_data["QQQ_RSI10"] = ta.rsi(strategy_data["QQQ"], length=10)
strategy_data["UVXY_RSI10"] = ta.rsi(strategy_data["UVXY"], length=10)
strategy_data["TQQQ_MA20"] = ta.sma(strategy_data["TQQQ"], length=20)
strategy_data["SQQQ_RSI10"] = ta.rsi(strategy_data["SQQQ"], length=10)
strategy_data["TLT_RSI10"] = ta.rsi(strategy_data["TLT"], length=10)
strategy_data["QQQ_5dgain"] = strategy_data["QQQ"].pct_change(5) * 100
strategy_data["TQQQ_1dgain"] = strategy_data["TQQQ"].pct_change(1) * 100
strategy_data["TQQQ_10dSTDreturn"] = strategy_data["TQQQ_1dgain"].rolling(10).std()

# Drop missing values so we start the backtest from the first day all signals and prices are available
strategy_data.dropna(inplace=True)

# Add "BUY" column in first position
strategy_data.insert(0, "BUY", pd.NA) # pd.NA is the new pandas native version of np.nan

strategy_data

	BUY	SPY	TQQQ	SPXL	UVXY	SQQQ	TLT	QQQ	SPY_MA200	TQQQ_RSI10	SPXL_RSI10	SPY_RSI10	QQQ_RSI10	UVXY_RSI10	TQQQ_MA20	SQQQ_RSI10	TLT_RSI10	QQQ_5dgain	TQQQ_1dgain	TQQQ_10dSTDreturn
Date
1999-04-07	NaN	86.938591	81.72095	45.31803	2.879196e+20	2.102258e+08	33.35910	47.432098	75.215919	63.071977	62.293455	63.046299	64.348921	31.463376	71.300601	34.759583	61.585289	3.345116	-3.161540	6.725346
1999-04-08	NaN	88.040405	83.63468	47.05327	2.817470e+20	2.053532e+08	33.66911	47.809162	75.298414	65.188269	67.303069	67.243478	66.252551	30.133641	72.093586	33.332605	68.707198	5.717654	2.341786	6.329253
1999-04-09	NaN	88.060799	85.15391	47.49072	2.644424e+20	2.016712e+08	33.57613	48.105450	75.376765	66.863324	68.476215	67.319813	67.755756	26.628145	72.914247	32.221921	64.709255	4.200681	1.816507	5.661916
1999-04-12	NaN	88.999359	83.40080	48.55757	2.655143e+20	2.058710e+08	33.63814	47.782230	75.454958	62.977932	71.269533	70.799538	64.284968	27.210950	73.899929	34.968226	66.168040	-0.112632	-2.058755	5.549664
1999-04-13	NaN	88.428055	79.98025	47.59537	2.781925e+20	2.143630e+08	33.42109	47.135807	75.530900	55.931812	65.457097	66.043419	57.715211	34.090945	74.441727	40.394174	57.003922	-1.657761	-4.101340	4.904016
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2022-10-12	NaN	356.559998	18.10000	52.77000	1.325000e+01	6.335000e+01	100.35000	262.660004	413.231901	29.380442	32.266999	33.369398	30.256253	65.041294	21.645000	68.684792	34.975614	-6.851552	-0.220507	6.327154
2022-10-13	NaN	365.970001	19.31000	56.85000	1.269000e+01	5.911000e+01	99.39000	268.820007	412.704691	37.938121	43.105232	45.241914	40.328453	57.202852	21.356000	57.165998	31.833812	-3.910495	6.685083	6.389549
2022-10-14	NaN	357.630005	17.57000	52.98000	1.314000e+01	6.438000e+01	98.57000	260.739990	412.132767	31.784061	36.884882	38.488365	33.316123	61.360555	21.004500	65.221002	29.333059	-3.106658	-9.010875	6.806726
2022-10-17	NaN	366.820007	19.30000	57.00000	1.261000e+01	5.800000e+01	98.09000	269.350006	411.613316	42.150708	45.896145	47.993457	44.700758	54.439482	20.710000	52.053713	27.907179	1.103563	9.846329	7.220773
2022-10-18	NaN	371.130005	19.78000	58.99000	1.221000e+01	5.659000e+01	98.32000	271.480011	411.121348	44.739629	49.835568	51.869085	47.179520	49.735237	20.469500	49.595027	29.725895	3.322554	2.487047	6.484238

5924 rows × 20 columns

Generate asset allocation history

We create an empty dataframe that will be populated with our asset allocation history as we iterate over the strategy_data dataframe and calculate the strategy signal for each row (day). This allocation dataframe will later be passed to vectorbt.pro for backtesting.

strategy_allocations = pd.DataFrame(index=strategy_data.index, columns=["TQQQ", "SQQQ", "UVXY", "TLT"])
strategy_allocations

	TQQQ	SQQQ	UVXY	TLT
Date
1999-04-07	NaN	NaN	NaN	NaN
1999-04-08	NaN	NaN	NaN	NaN
1999-04-09	NaN	NaN	NaN	NaN
1999-04-12	NaN	NaN	NaN	NaN
1999-04-13	NaN	NaN	NaN	NaN
...	...	...	...	...
2022-10-12	NaN	NaN	NaN	NaN
2022-10-13	NaN	NaN	NaN	NaN
2022-10-14	NaN	NaN	NaN	NaN
2022-10-17	NaN	NaN	NaN	NaN
2022-10-18	NaN	NaN	NaN	NaN

5924 rows × 4 columns

We now write our strategy logic and iterate over each row of the dataframe, finding the asset allocations for each row and putting them in the strategy_allocations dataframe (1 = 100% allocation, 0.5 = 50% etc.)

# Strategy logic 

for row in strategy_data.index:
    if strategy_data.at[row, "SPY"] > strategy_data.at[row, "SPY_MA200"]: # using .at when workining with single values is more efficient than .loc
        if strategy_data.at[row, "TQQQ_RSI10"] > 79:
            strategy_allocations.at[row, "UVXY"] = 1
        else:
            if strategy_data.at[row, "SPXL_RSI10"] > 80:
                strategy_allocations.at[row, "UVXY"] = 1
            else:
                if strategy_data.at[row, "QQQ_5dgain"] < -6:
                    if strategy_data.at[row, "TQQQ_1dgain"] > 5:
                        strategy_allocations.at[row, "SQQQ"] = 1
                    else:
                        if strategy_data.at[row, "TQQQ_RSI10"] > 31:
                            strategy_allocations.at[row, "SQQQ"] = 1
                        else:
                            strategy_allocations.at[row, "TQQQ"] = 1
                else:
                    if strategy_data.at[row, "QQQ_RSI10"] > 80:
                        strategy_allocations.at[row, "SQQQ"] = 1
                    else:
                        if strategy_data.at[row, "TQQQ_10dSTDreturn"] > 5:
                            strategy_allocations.at[row, "TLT"] = 1
                        else:
                            strategy_allocations.at[row, "TQQQ"] = 1
    else:
        if strategy_data.at[row, "TQQQ_RSI10"] < 31:
            strategy_allocations.at[row, "TQQQ"] = 1
        else:
            if strategy_data.at[row, "SPY_RSI10"] < 30:
                strategy_allocations.at[row, "TQQQ"] = 1
            else:
                if strategy_data.at[row, "UVXY_RSI10"] > 74:
                    if strategy_data.at[row, "UVXY_RSI10"] > 84:
                        if (
                            strategy_data.at[row, "SQQQ_RSI10"]
                            > strategy_data.at[row, "TLT_RSI10"]
                        ):
                            strategy_allocations.at[row, "SQQQ"] = 1
                        else:
                            strategy_allocations.at[row, "TLT"] = 1
                    else:
                        strategy_allocations.at[row, "UVXY"] = 1
                else:
                    if strategy_data.at[row, "TQQQ"] > strategy_data.at[row, "TQQQ_MA20"]:
                        if strategy_data.at[row, "SQQQ_RSI10"] < 31:
                            strategy_allocations.at[row, "SQQQ"] = 1
                        else:
                            strategy_allocations.at[row, "TQQQ"] = 1
                    else:
                        if (
                            strategy_data.at[row, "SQQQ_RSI10"]
                            > strategy_data.at[row, "TLT_RSI10"]
                        ):
                            strategy_allocations.at[row, "SQQQ"] = 1
                        else:
                            strategy_allocations.at[row, "TLT"] = 1

# Replace NaN values with 0
strategy_allocations.fillna(0, inplace=True)

strategy_allocations

	TQQQ	SQQQ	UVXY	TLT
Date
1999-04-07	0	0	0	1
1999-04-08	0	0	0	1
1999-04-09	0	0	0	1
1999-04-12	0	0	0	1
1999-04-13	1	0	0	0
...	...	...	...	...
2022-10-12	1	0	0	0
2022-10-13	0	1	0	0
2022-10-14	0	1	0	0
2022-10-17	0	1	0	0
2022-10-18	0	1	0	0

5924 rows × 4 columns

Saving the data to .csv

# Save strategy_allocations to CSV for later use
strategy_allocations.to_csv("./data/strategy_allocations.csv")

We also save a dataframe with the close prices of the instruments used in our strategy allocation. This will be needed for backtesting in vectorbt.pro.

strategy_prices = strategy_data[["TQQQ", "SQQQ", "UVXY", "TLT"]]
strategy_prices.to_csv("./data/strategy_prices.csv")

# Save the entire strategy data. This won't be needed later, but it's nice to have, just in case we want to reuse it somewhere.
strategy_data.to_csv("./data/strategy_data.csv")

Backtesting the strategy

To run this backtest you'll need to have a valid subscription to vectorbt.pro.

Loading data into vectorbt.pro and preparing it for the backtest

First, we are loading the tickers close prices from the .csv file we saved earlier.

import vectorbtpro as vbt

vbt_prices = vbt.CSVData.fetch("./data/strategy_prices.csv", index_col=0, parse_dates=True) # same syntax you would use with pd.read_csv

vbt_prices.data

{'strategy_prices':                                TQQQ          SQQQ          UVXY        TLT
 Date                                                                      
 1999-04-07 00:00:00+00:00  81.72095  2.102258e+08  2.879196e+20   33.35910
 1999-04-08 00:00:00+00:00  83.63468  2.053532e+08  2.817470e+20   33.66911
 1999-04-09 00:00:00+00:00  85.15391  2.016712e+08  2.644424e+20   33.57613
 1999-04-12 00:00:00+00:00  83.40080  2.058710e+08  2.655143e+20   33.63814
 1999-04-13 00:00:00+00:00  79.98025  2.143630e+08  2.781925e+20   33.42109
 ...                             ...           ...           ...        ...
 2022-10-12 00:00:00+00:00  18.10000  6.335000e+01  1.325000e+01  100.35000
 2022-10-13 00:00:00+00:00  19.31000  5.911000e+01  1.269000e+01   99.39000
 2022-10-14 00:00:00+00:00  17.57000  6.438000e+01  1.314000e+01   98.57000
 2022-10-17 00:00:00+00:00  19.30000  5.800000e+01  1.261000e+01   98.09000
 2022-10-18 00:00:00+00:00  19.78000  5.659000e+01  1.221000e+01   98.32000
 
 [5924 rows x 4 columns]}

We now generate a dataframe in a format compatible with vectorbt.pro, which will be populated with our allocation weights later.

symbol_wrapper = vbt_prices.get_symbol_wrapper(symbols=["TQQQ", "SQQQ", "UVXY", "TLT"], freq='1d') # the order of symbols= should match the order of columns in strategy_allocations
print(symbol_wrapper.shape)
print(strategy_allocations.shape)

(5924, 4)
(5924, 4)

The wrapper should have the same number of rows and columns as the data we have loaded. Now we can fill symbol_wrapper with data from our allocations dataframe.

tickers_allocations = vbt.PortfolioOptimizer.from_allocations(symbol_wrapper, strategy_allocations)
allocations = tickers_allocations.allocations # can also type tickers_allocations.get_allocations()
allocations

symbol	TQQQ	SQQQ	UVXY	TLT
Date
1999-04-07 00:00:00+00:00	0.0	0.0	0.0	1.0
1999-04-08 00:00:00+00:00	0.0	0.0	0.0	1.0
1999-04-09 00:00:00+00:00	0.0	0.0	0.0	1.0
1999-04-12 00:00:00+00:00	0.0	0.0	0.0	1.0
1999-04-13 00:00:00+00:00	1.0	0.0	0.0	0.0
...	...	...	...	...
2022-10-12 00:00:00+00:00	1.0	0.0	0.0	0.0
2022-10-13 00:00:00+00:00	0.0	1.0	0.0	0.0
2022-10-14 00:00:00+00:00	0.0	1.0	0.0	0.0
2022-10-17 00:00:00+00:00	0.0	1.0	0.0	0.0
2022-10-18 00:00:00+00:00	0.0	1.0	0.0	0.0

5924 rows × 4 columns

Now we get a dataframe with the close prices from the data we have loaded.

close_prices = vbt_prices.get()
close_prices

	TQQQ	SQQQ	UVXY	TLT
Date
1999-04-07 00:00:00+00:00	81.72095	2.102258e+08	2.879196e+20	33.35910
1999-04-08 00:00:00+00:00	83.63468	2.053532e+08	2.817470e+20	33.66911
1999-04-09 00:00:00+00:00	85.15391	2.016712e+08	2.644424e+20	33.57613
1999-04-12 00:00:00+00:00	83.40080	2.058710e+08	2.655143e+20	33.63814
1999-04-13 00:00:00+00:00	79.98025	2.143630e+08	2.781925e+20	33.42109
...	...	...	...	...
2022-10-12 00:00:00+00:00	18.10000	6.335000e+01	1.325000e+01	100.35000
2022-10-13 00:00:00+00:00	19.31000	5.911000e+01	1.269000e+01	99.39000
2022-10-14 00:00:00+00:00	17.57000	6.438000e+01	1.314000e+01	98.57000
2022-10-17 00:00:00+00:00	19.30000	5.800000e+01	1.261000e+01	98.09000
2022-10-18 00:00:00+00:00	19.78000	5.659000e+01	1.221000e+01	98.32000

5924 rows × 4 columns

Running the backtest

We now have everything that is needed to run our backtest: a dataframe with the close prices and one with the target allocations, both in a valid format recognized by vectorbt.pro . We can proceed with running the backtest and showing the results.

portfolio = vbt.Portfolio.from_orders(
    
    close=close_prices,
    #bm_close=put_your_benchmark_close_prices_here,
    size=allocations, # Here we pass the allocations we calculated earlier
    size_type="targetpercent",
    group_by=True,  
    cash_sharing=True,
    call_seq="auto",
    freq='D', # needed to calculate annualized returns and some metrics like sharpe, sortino etc in portfolio.stats()
    fees=0, # can specify a transaction fee
    )

Read the documentation of Portfolio.from_orders() for more info. It's possible compare the performance to a benchmark if you pass the desired benchmark strategy/ticker time series to #bm_close . Beware that the close_prices dataframe and the benchmark you are providing must have the same index for accurate comparisons.

Analyze backtest performance

In this notebook we will show just the tip of the iceberg of what's possible with vectorbt.pro . Read the documentation for more ideas and metrics.

# This is the most basic overview of the portfolio metrics
portfolio.stats()

Start                         1999-04-07 00:00:00+00:00
End                           2022-10-18 00:00:00+00:00
Period                               5924 days 00:00:00
Start Value                                       100.0
Min Value                                     45.944895
Max Value                            49532199635.866318
End Value                            43538788092.477097
Total Return [%]                     43538787992.477097
Benchmark Return [%]                          -20.26587
Total Time Exposure [%]                       99.561107
Max Gross Exposure [%]                            100.0
Max Drawdown [%]                              73.491664
Max Drawdown Duration                 875 days 00:00:00
Total Orders                                       1910
Total Fees Paid                                     0.0
Total Trades                                        959
Win Rate [%]                                  56.680585
Best Trade [%]                                65.372054
Worst Trade [%]                              -37.213453
Avg Winning Trade [%]                          9.067937
Avg Losing Trade [%]                          -5.788179
Avg Winning Trade Duration    7 days 18:17:54.033149171
Avg Losing Trade Duration     7 days 18:41:46.077348066
Profit Factor                                  3.598681
Expectancy                              47471405.655066
Sharpe Ratio                                   1.813416
Calmar Ratio                                   1.805824
Omega Ratio                                    1.368747
Sortino Ratio                                  3.026992
Name: group, dtype: object

Returns

We can go even further than that and do in-depth analysis of returns and other metrics.

portfolio.returns_stats()

Start                        1999-04-07 00:00:00+00:00
End                          2022-10-18 00:00:00+00:00
Period                              5924 days 00:00:00
Total Return [%]                    43538787992.477104
Benchmark Return [%]                         -20.26587
Annualized Return [%]                       132.712975
Annualized Volatility [%]                    88.617529
Max Drawdown [%]                             73.491664
Max Drawdown Duration                875 days 00:00:00
Sharpe Ratio                                  1.813416
Calmar Ratio                                  1.805824
Omega Ratio                                   1.368747
Sortino Ratio                                 3.026992
Skew                                          1.516343
Kurtosis                                     16.768799
Tail Ratio                                    1.117592
Common Sense Ratio                            2.600781
Value at Risk                                 -0.06385
Alpha                                         3.985386
Beta                                         -0.107782
Name: group, dtype: object

Annual Returns

We can see the annual returns of the strategy, and the advantage of having the output in a simple pandas series is that we can then manipulate and display this data however we want to analyze it in different ways, for example by making a bar plot of the annual returns, in this specific case.

# Annual returns
annual_returns = portfolio.annual_returns # can also type portfolio.get_annual_returns()
annual_returns

Date
1999-04-07 00:00:00+00:00    -0.225791
2000-04-06 00:00:00+00:00     4.087178
2001-04-06 00:00:00+00:00    -0.189695
2002-04-06 00:00:00+00:00     1.131165
2003-04-06 00:00:00+00:00     0.884160
2004-04-05 00:00:00+00:00    -0.289397
2005-04-05 00:00:00+00:00     0.356911
2006-04-05 00:00:00+00:00    -0.166240
2007-04-05 00:00:00+00:00     0.687942
2008-04-04 00:00:00+00:00     2.966231
2009-04-04 00:00:00+00:00     1.305499
2010-04-04 00:00:00+00:00     1.417747
2011-04-04 00:00:00+00:00     0.821203
2012-04-03 00:00:00+00:00     0.130452
2013-04-03 00:00:00+00:00     1.395278
2014-04-03 00:00:00+00:00     0.842624
2015-04-03 00:00:00+00:00     1.350014
2016-04-02 00:00:00+00:00     0.675128
2017-04-02 00:00:00+00:00     1.948730
2018-04-02 00:00:00+00:00     1.696418
2019-04-02 00:00:00+00:00    25.100971
2020-04-01 00:00:00+00:00     9.515804
2021-04-01 00:00:00+00:00     1.459666
2022-04-01 00:00:00+00:00     4.898051
Freq: 365D, Name: group, dtype: float64

# Make an bar plot of the annual returns
annual_returns_plot = px.bar(annual_returns, x=annual_returns.index, y=annual_returns.values, labels={"x":"Year", "y":"Portfolio Return"}, title="Annual Returns")
annual_returns_plot.show()

# Delete this when running the notebook yourself
annual_returns_plot.write_html("./html/annual_returns.html", default_width=800, default_height=500)
IFrame(src='./html/annual_returns.html', width=800, height=500)

Plots

Vectorbt.pro also has a lot of already built-in interactive plots that can be generated very easily. Below we'll plot all the available plots just for reference

# Let's plot all the possible subplots just for reference. By default there's only the return curve.
all_portfolio_plots = portfolio.plot(settings=dict(bm_returns=True), subplots="all")
all_portfolio_plots

/Users/lorenzominghetti/Projects/vectorbt.pro-1.7.1/vectorbtpro/generic/plots_builder.py:396: UserWarning:

Subplot 'orders' does not support grouped data

/Users/lorenzominghetti/Projects/vectorbt.pro-1.7.1/vectorbtpro/generic/plots_builder.py:396: UserWarning:

Subplot 'trades' does not support grouped data

/Users/lorenzominghetti/Projects/vectorbt.pro-1.7.1/vectorbtpro/generic/plots_builder.py:396: UserWarning:

Subplot 'trade_pnl' does not support grouped data

/Users/lorenzominghetti/Projects/vectorbt.pro-1.7.1/vectorbtpro/generic/plots_builder.py:396: UserWarning:

Subplot 'asset_flow' does not support grouped data

/Users/lorenzominghetti/Projects/vectorbt.pro-1.7.1/vectorbtpro/generic/plots_builder.py:396: UserWarning:

Subplot 'assets' does not support grouped data

FigureWidget({
    'data': [{'legendgroup': '0',
              'line': {'color': '#2ca02c'},
              'mo…

# Delete this this when running the notebook yourself
all_portfolio_plots.write_html("./html/all_portfolio_plots.html", default_width=800, default_height=1000)
IFrame(src='./html/all_portfolio_plots.html', width=800, height=1000)

It's also possible to specify custom subplots. The example below would create a cumulative returns and rolling drawdown plot. You can set the settings across all the subplots by using the global template mapping by adding template_context=dict(window=10) for example.

# subplots = [
#     ('cumulative_returns', dict(
#         title='Cumulative Returns',
#         yaxis_kwargs=dict(title='Cumulative returns'),
#         plot_func='cumulative_returns.vbt.plot',
#         select_col_cumulative_returns=True,
#         pass_add_trace_kwargs=True
#     )),
#     ('rolling_drawdown', dict(
#         title='Rolling Drawdown',
#         yaxis_kwargs=dict(title='Rolling drawdown'),
#         plot_func=[
#             'returns_acc',  # returns accessor
#             (
#                 'rolling_max_drawdown',  # function name
#                 (vbt.Rep('window'),)),  # positional arguments
#             'vbt.plot'  # plotting function
#         ],
#         select_col_returns_acc=True,
#         pass_add_trace_kwargs=True,
#         trace_names=[vbt.Sub('rolling_drawdown(${window})')],  # add window to the trace name
#     ))
# ]

# portfolio.plot(subplots, template_context=dict(window=10))

Drawdowns

Like the returns, drawdowns can also be analyzed extensively.

portfolio.get_drawdowns().stats()

Start                           1999-04-07 00:00:00+00:00
End                             2022-10-18 00:00:00+00:00
Period                                 5924 days 00:00:00
Coverage [%]                                     86.79946
Total Records                                         332
Total Recovered Drawdowns                             331
Total Active Drawdowns                                  1
Active Drawdown [%]                             12.100031
Active Duration                           2 days 00:00:00
Active Recovery [%]                                   0.0
Active Recovery Return [%]                            0.0
Active Recovery Duration                  0 days 00:00:00
Max Drawdown [%]                                73.491664
Avg Drawdown [%]                                 8.102684
Max Drawdown Duration                   875 days 00:00:00
Avg Drawdown Duration          15 days 12:41:19.758308157
Max Recovery Return [%]                        294.319213
Avg Recovery Return [%]                         14.915101
Max Recovery Duration                   249 days 00:00:00
Avg Recovery Duration           7 days 12:06:31.540785498
Avg Recovery Duration Ratio                      1.367338
Name: group, dtype: object

drawdown_plot = portfolio.get_drawdowns().plot()
drawdown_plot

FigureWidget({
    'data': [{'line': {'color': '#1f77b4'},
              'mode': 'lines',
              'name'…

# Delete this this when running the notebook yourself
drawdown_plot.write_html("./html/drawdown_plot.html", default_width=800, default_height=500)
IFrame(src='./html/all_portfolio_plots.html', width=800, height=500)

underwater_plot = portfolio.plot_underwater()
underwater_plot

FigureWidget({
    'data': [{'fill': 'tozeroy',
              'fillcolor': 'rgba(220,57,18,0.3000)',
         …

# Delete this this when running the notebook yourself
underwater_plot.write_html("./html/drawdown_plot.html", default_width=800, default_height=500)
IFrame(src='./html/drawdown_plot.html', width=800, height=500)

drawdown_df = portfolio.get_drawdowns().records_readable
drawdown_df

	Drawdown Id	Column	Peak Index	Start Index	Valley Index	End Index	Peak Value	Valley Value	End Value	Status
0	0	group	1999-04-08 00:00:00+00:00	1999-04-09 00:00:00+00:00	1999-04-16 00:00:00+00:00	1999-04-19 00:00:00+00:00	1.009293e+02	8.917106e+01	1.017766e+02	Recovered
1	1	group	1999-04-20 00:00:00+00:00	1999-04-21 00:00:00+00:00	1999-10-28 00:00:00+00:00	2000-01-06 00:00:00+00:00	1.086694e+02	4.954582e+01	1.268284e+02	Recovered
2	2	group	2000-01-06 00:00:00+00:00	2000-01-07 00:00:00+00:00	2000-04-14 00:00:00+00:00	2000-08-02 00:00:00+00:00	1.268284e+02	4.594489e+01	1.304440e+02	Recovered
3	3	group	2000-08-02 00:00:00+00:00	2000-08-03 00:00:00+00:00	2000-08-07 00:00:00+00:00	2000-08-31 00:00:00+00:00	1.304440e+02	1.135298e+02	1.365279e+02	Recovered
4	4	group	2000-09-01 00:00:00+00:00	2000-09-05 00:00:00+00:00	2000-09-07 00:00:00+00:00	2000-09-26 00:00:00+00:00	1.404799e+02	1.139203e+02	1.433934e+02	Recovered
...	...	...	...	...	...	...	...	...	...	...
327	327	group	2022-09-21 00:00:00+00:00	2022-09-22 00:00:00+00:00	2022-09-26 00:00:00+00:00	2022-09-29 00:00:00+00:00	3.661402e+10	3.311646e+10	3.800744e+10	Recovered
328	328	group	2022-09-29 00:00:00+00:00	2022-09-30 00:00:00+00:00	2022-09-30 00:00:00+00:00	2022-10-03 00:00:00+00:00	3.800744e+10	3.601294e+10	3.851073e+10	Recovered
329	329	group	2022-10-03 00:00:00+00:00	2022-10-04 00:00:00+00:00	2022-10-05 00:00:00+00:00	2022-10-07 00:00:00+00:00	3.851073e+10	3.488746e+10	3.993029e+10	Recovered
330	330	group	2022-10-11 00:00:00+00:00	2022-10-12 00:00:00+00:00	2022-10-12 00:00:00+00:00	2022-10-13 00:00:00+00:00	4.272210e+10	4.262790e+10	4.547761e+10	Recovered
331	331	group	2022-10-14 00:00:00+00:00	2022-10-17 00:00:00+00:00	2022-10-18 00:00:00+00:00	2022-10-18 00:00:00+00:00	4.953220e+10	4.353879e+10	4.353879e+10	Active

332 rows × 10 columns

As another example of how it's possible to play with the data, let's create a dataframe that sorts the percentage drawdowns from biggest to smallest and shows the drawdown start and recovery date.

drawdown_df["Drawdown PCT"] = ((drawdown_df["Peak Value"]-drawdown_df["Valley Value"])/drawdown_df["Peak Value"])*100
biggest_drawdowns = drawdown_df[["Peak Index", "Valley Index", "Drawdown PCT"]].sort_values(by=["Drawdown PCT"], ascending=False)
biggest_drawdowns["Peak Index"] = biggest_drawdowns["Peak Index"].dt.tz_localize(None)
biggest_drawdowns["Valley Index"] = biggest_drawdowns["Valley Index"].dt.tz_localize(None)
biggest_drawdowns

	Peak Index	Valley Index	Drawdown PCT
17	2001-04-06	2001-09-21	73.491664
42	2004-01-16	2006-07-14	65.079365
2	2000-01-06	2000-04-14	63.773964
22	2001-12-07	2002-06-11	58.781410
1	1999-04-20	1999-10-28	54.406845
...	...	...	...
165	2016-03-14	2016-03-15	0.063342
300	2021-07-12	2021-07-13	0.053529
29	2003-07-09	2003-07-10	0.033248
207	2017-11-09	2017-11-10	0.029929
284	2020-12-29	2020-12-30	0.016628

332 rows × 3 columns

Analysis per single asset inside the portfolio

With group_by=False it's possible analize the contribution to returns, drawdowns and all the other metrics for each single ticker in the portfolio separately from the others, for really in-depth analysis of what's contributing to the performance of the portfolio.

portfolio_not_grouped = vbt.Portfolio.from_orders(
    
    close=close_prices,
    #bm_close=put_your_benchmark_close_prices_here,
    size=allocations, # Here we pass the allocations we calculated earlier
    size_type="targetpercent",
    group_by=False,  
    cash_sharing=True,
    call_seq="auto",
    freq='D', # needed to calculate annualized returns and some metrics like sharpe, sortino etc in portfolio.stats()
    fees=0, # can specify a transaction fee here
    )

We can get a dataframe with all the stats for each single asset

single_asset_stats = portfolio_not_grouped.stats(agg_func=None)
single_asset_stats

		Start	End	Period	Start Value	Min Value	Max Value	End Value	Total Return [%]	Benchmark Return [%]	Total Time Exposure [%]	...	Avg Winning Trade [%]	Avg Losing Trade [%]	Avg Winning Trade Duration	Avg Losing Trade Duration	Profit Factor	Expectancy	Sharpe Ratio	Calmar Ratio	Omega Ratio	Sortino Ratio
symbol
TQQQ	TQQQ	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	100.0	74.270866	1.045371e+07	9.041877e+06	9.041777e+06	-75.795680	73.227549	...	12.175537	-8.342310	12 days 00:54:54.915254237	13 days 09:48:10.140845070	2.303589	23184.044678	1.336266	0.822345	1.281332	2.092036
SQQQ	SQQQ	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	100.0	35.563083	1.683875e+03	1.480125e+03	1.380125e+03	-99.999973	7.241729	...	10.014137	-7.289226	2 days 08:39:35.257731958	2 days 10:04:26.666666666	2.682986	8.123802	0.617504	0.176093	1.382720	0.984780
UVXY	UVXY	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	100.0	95.815675	1.891803e+04	1.891803e+04	1.881803e+04	-100.000000	3.190412	...	13.570371	-4.542983	2 days 12:40:45.283018867	2 days 10:54:32.727272727	62.436518	247.605630	1.086683	0.718410	3.314147	3.587407
TLT	TLT	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	100.0	91.964316	1.718486e+02	1.427350e+02	4.273496e+01	194.732172	15.749494	...	1.293369	-1.165640	4 days 10:15:29.032258064	3 days 12:41:32.307692307	1.274152	0.183412	0.384537	0.089135	1.146790	0.568820

4 rows × 29 columns

# Alternative way to grab the stats for a specific asset, in this case SQQQ
portfolio_not_grouped.stats(group_by=False, column=2)

Start                         1999-04-07 00:00:00+00:00
End                           2022-10-18 00:00:00+00:00
Period                               5924 days 00:00:00
Start Value                                       100.0
Min Value                                     95.815675
Max Value                                  18918.027881
End Value                                  18918.027881
Total Return [%]                           18818.027881
Benchmark Return [%]                             -100.0
Total Time Exposure [%]                        3.190412
Max Gross Exposure [%]                            100.0
Max Drawdown [%]                              34.706887
Max Drawdown Duration                 323 days 00:00:00
Total Orders                                        152
Total Fees Paid                                     0.0
Total Trades                                         76
Win Rate [%]                                  69.736842
Best Trade [%]                                65.372054
Worst Trade [%]                              -20.203265
Avg Winning Trade [%]                         13.570371
Avg Losing Trade [%]                          -4.542983
Avg Winning Trade Duration    2 days 12:40:45.283018867
Avg Losing Trade Duration     2 days 10:54:32.727272727
Profit Factor                                 62.436518
Expectancy                                    247.60563
Sharpe Ratio                                   1.086683
Calmar Ratio                                    0.71841
Omega Ratio                                    3.314147
Sortino Ratio                                  3.587407
Name: 2, dtype: object

We can get even more detailed returns stats for each single asset, and then potentially analyzing or plotting them in many different ways.

# Even more detailed returns stats for each single asset
single_asset_returns = portfolio_not_grouped.returns_stats(agg_func=None)
single_asset_returns

		Start	End	Period	Total Return [%]	Benchmark Return [%]	Annualized Return [%]	Annualized Volatility [%]	Max Drawdown [%]	Max Drawdown Duration	Sharpe Ratio	Calmar Ratio	Omega Ratio	Sortino Ratio	Skew	Kurtosis	Tail Ratio	Common Sense Ratio	Value at Risk	Alpha	Beta
symbol
TQQQ	TQQQ	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	9.041777e+06	-75.795680	62.349412	71.354278	75.819009	875 days	1.336266	0.822345	1.281332	2.092036	1.141272	17.109108	1.024298	1.662942	-0.054903	1.098805	0.513265
SQQQ	SQQQ	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	1.380125e+03	-99.999973	12.122497	39.486874	68.841582	3217 days	0.617504	0.176093	1.382720	0.984780	1.195679	61.734717	inf	inf	0.000000	0.363114	0.156942
UVXY	UVXY	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	1.881803e+04	-100.000000	24.933761	34.343159	34.706887	323 days	1.086683	0.718410	3.314147	3.587407	16.572080	401.969725	inf	inf	0.000000	0.470850	0.013698
TLT	TLT	1999-04-07 00:00:00+00:00	2022-10-18 00:00:00+00:00	5924 days	4.273496e+01	194.732172	1.522337	6.200627	17.079081	1869 days	0.384537	0.089135	1.146790	0.568820	0.613140	27.380370	1.359131	1.379821	-0.002743	0.012374	0.144512

single_asset_annual_returns = portfolio_not_grouped.get_annual_returns()
single_asset_annual_returns.plot()
# It's also possible to do a yearly bar chart with this data

<AxesSubplot: xlabel='Date'>

Remember the huge number of plots we generated earlier? There's the ability to do the same thing for each of the assets in the portfolioto plot a single asset of the portfolio with orders, trade pnl and returns. It's possible to specify additional subplots. Let's try see all the possibilities.

# We plot all the available plots for column 3 aka TLT
portfolio_not_grouped_plot = portfolio_not_grouped.plot(column=3, subplots="all") # we're plotting all the possible subplots just for reference
portfolio_not_grouped_plot

FigureWidget({
    'data': [{'legendgroup': '0',
              'line': {'color': '#1f77b4'},
              'mo…

# Delete this this when running the notebook yourself
portfolio_not_grouped_plot.write_html("./html/portfolio_not_grouped.html", default_width=800, default_height=500)
IFrame(src='./html/portfolio_not_grouped.html', width=800, height=500)

There is really so much more that can be done, once you get familiar with how things work. Feel free to contact me on discord ( Raekon#7854 ) if you need any help running this demo