# Stratestic 📈📊🛠️
[](https://codecov.io/gh/diogomatoschaves/stratestic)
[](https://badge.fury.io/py/stratestic)
The `stratestic` module is a Python package for backtesting, analysing and optimizing trading strategies.
It includes a number of pre-implemented strategies, but it is also possible to create new strategies, as well as
to combine them. It provides a general Machine Learning strategy, which can be further tweaked to your specific needs.
The basic usage is as simple as importing a backtesting class, a strategy and run the backtest. The backtest can
then be refined with optimizations or by incorporating leverage.
Overall it offers a powerful way to explore and experiment with different strategies,
and to evaluate the performance under different conditions.
If you are interested in a trading bot that integrates seamlessly with this library, check out
[MyCryptoBot](https://github.com/diogomatoschaves/MyCryptoBot).
## Installation
$ pip install stratestic
## Usage
1. [ Vectorized Backtesting ](#vectorized-backtesting)
2. [ Iterative Backtesting ](#iterative-backtesting)
3. [ Backtesting with leverage and margin ](#leverage) <br>
3.1. [ Calculating the maximum allowed leverage ](#maximum-leverage)
4. [ Optimization ](#optimization) <br>
4.1 [ Brute Force ](#brute-force) <br>
4.2 [ Genetic Algorithm ](#genetic-algorithm)
5. [ Strategies ](#strategies) <br>
5.1. [ Combined strategies](#combined-strategies) <br>
5.2. [ Create new strategies](#new-strategies) <br>
5.3. [ Machine Learning strategy ](#machine-learning)
<a name="vectorized-backtesting"></a>
### Vectorized Backtesting
The `VectorizedBacktester` is a backtesting class that allows you to test trading strategies
on historical price data. It has the advantage of being faster than the iterative backtesting, but at
a cost of flexibility, as it will be hard or outright not possible to accomplish this for some more
complex strategies. For all the strategies provided by this library, vectorized backtesting is supported.
Below is an example of how to use it for the `MovingAverageCrossover` strategy:
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MovingAverageCrossover
symbol = "BTCUSDT"
trading_costs = 0.1 # This should be in percentage, i.e. 0.1%
mov_avg = MovingAverageCrossover(50, 200)
vect = VectorizedBacktester( # Initializes the VectorizedBacktester class with the strategy.
mov_avg,
symbol,
amount=1000,
trading_costs=trading_costs
)
vect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'
vect.run() # Runs the backtest and shows the results
```
This will output the results in textual and graphical form.
```
************************************************************
BACKTESTING RESULTS
************************************************************
Overview
------------------------------------------------------------
Total Duration 4 years and 38 weeks
Start Date 2018-05-23 13:00:00
End Date 2023-02-13 01:00:00
Trading Costs [%] 0.1
Exposure Time [%] 100.0
Leverage [x] 1
Equity - Initial [USDT] 1000
Equity - Final [USDT] 3215.96
Equity - Peak [USDT] 5356.87
------------------------------------------------------------
Returns
------------------------------------------------------------
Total Return [%] 221.6
Annualized Return [%] 21.49
Annualized Volatility [%] 73.95
Buy & Hold Return [%] 175.98
------------------------------------------------------------
Drawdowns
------------------------------------------------------------
Max Drawdown [%] -61.18
Avg Drawdown [%] -8.2
Max Drawdown Duration 1 year and 38 weeks
Avg Drawdown Duration 3 weeks and 2 days
------------------------------------------------------------
Trades
------------------------------------------------------------
Total Trades 267
Win Rate [%] 32.21
Best Trade [%] 87.77
Worst Trade [%] -21.11
Avg Trade [%] 0.44
Max Trade Duration 5 weeks and 3 days
Avg Trade Duration 6 days and 11 hours
Expectancy [%] 5.9
------------------------------------------------------------
Ratios
------------------------------------------------------------
Sharpe Ratio 0.33
Sortino Ratio 0.28
Calmar Ratio 0.35
Profit Factor 1.0
System Quality Number -0.02
------------------------------------------------------------
************************************************************
```
<p align="left">
<img src="stratestic/utils/drawings/vectorized_results.png" style="width: 100%" />
</p>
<a name="iterative-backtesting"></a>
### Iterative Backtesting
The `IterativeBacktester` is a backtesting class that allows you to test trading strategies
on historical price data. It works by iterating through each historical data point and simulating
trades based on your strategy. This feature allows for a greater degree of flexibility,
allowing you to add more complex logic to the strategies. Below is an example of how you would use this
class to backtest the `MovingAverageConvergenceDivergence` strategy.
```python
from stratestic.backtesting import IterativeBacktester
from stratestic.strategies import MovingAverageConvergenceDivergence
symbol = "BTCUSDT"
macd = MovingAverageConvergenceDivergence(26, 12, 9)
ite = IterativeBacktester(macd, symbol=symbol) # Initializes the IterativeBacktester class with the strategy
ite.load_data() # Load the default sample data. You can pass your own DataFrame to load_data
ite.run() # Runs the backtest and shows the results
```
This will output the results in textual and graphical form.
```
************************************************************
BACKTESTING RESULTS
************************************************************
Overview
------------------------------------------------------------
Total Duration 4 years and 39 weeks
Start Date 2018-05-16 15:00:00
End Date 2023-02-13 01:00:00
Trading Costs [%] 0.0
Exposure Time [%] 100.0
Leverage [x] 1
Equity - Initial [USDT] 1000
Equity - Final [USDT] 17145.68
Equity - Peak [USDT] 29566.42
------------------------------------------------------------
Returns
------------------------------------------------------------
Total Return [%] 1614.57
Annualized Return [%] 60.58
Annualized Volatility [%] 70.99
Buy & Hold Return [%] 163.16
------------------------------------------------------------
Drawdowns
------------------------------------------------------------
Max Drawdown [%] -56.09
Avg Drawdown [%] -5.46
Max Drawdown Duration 1 year and 22 weeks
Avg Drawdown Duration 1 week and 1 day
------------------------------------------------------------
Trades
------------------------------------------------------------
Total Trades 3136
Win Rate [%] 34.92
Best Trade [%] 45.61
Worst Trade [%] -12.84
Avg Trade [%] 0.09
Max Trade Duration 2 days and 14 hours
Avg Trade Duration 13 hours and 15 minutes
Expectancy [%] 1.72
------------------------------------------------------------
Ratios
------------------------------------------------------------
Sharpe Ratio 0.84
Sortino Ratio 0.8
Calmar Ratio 1.08
Profit Factor 1.01
System Quality Number 0.16
------------------------------------------------------------
************************************************************
```
<p align="left">
<img src="stratestic/utils/drawings/iterative_results.png" style="width: 100%" />
</p>
<a name="leverage"></a>
### Backtesting with leverage and margin
Both the Vectorized and Iterative backtesting classes provide users with the ability to incorporate leverage into a
backtest and visualize the margin ratio evolution during the backtest. This feature enables users to identify
instances where a margin call would occur, leading to a potential loss of all funds. The calculations follow the
rules outlined by Binance, as detailed [here](https://www.binance.com/en/support/faq/how-to-calculate-liquidation-price-of-usd%E2%93%A2-m-futures-contracts-b3c689c1f50a44cabb3a84e663b81d93) and [here](https://www.binance.com/en/support/faq/leverage-and-margin-of-usd%E2%93%A2-m-futures-360033162192).
It's important to note that these calculations assume the selected margin is _Isolated_, and the position mode
is _One Way_. Below an example is shown:
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MovingAverageCrossover
symbol = "BTCUSDT"
trading_costs = 0.05
mov_avg = MovingAverageCrossover(20, 150)
vect = VectorizedBacktester(
mov_avg,
symbol,
amount=1000,
trading_costs=trading_costs,
leverage=7 # Pass the desired leverage here
)
vect.load_data()
vect.run()
```
This will output the following results and plot:
```
************************************************************
BACKTESTING RESULTS
************************************************************
Overview
------------------------------------------------------------
Total Duration 4 years and 38 weeks
Start Date 2018-05-21 11:00:00
End Date 2023-02-13 01:00:00
Trading Costs [%] 0.05
Exposure Time [%] 100.0
Leverage [x] 7
Equity - Initial [USDT] 1000
Equity - Final [USDT] 0.0
Equity - Peak [USDT] 17012.48
------------------------------------------------------------
Returns
------------------------------------------------------------
Total Return [%] -100.0
Annualized Return [%] -100.0
Annualized Volatility [%] 144.81
Buy & Hold Return [%] 157.53
------------------------------------------------------------
Drawdowns
------------------------------------------------------------
Max Drawdown [%] -100.0
Avg Drawdown [%] -15.65
Max Drawdown Duration 4 years and 7 weeks
Avg Drawdown Duration 4 weeks and 7 hours
------------------------------------------------------------
Trades
------------------------------------------------------------
Total Trades 82
Win Rate [%] 28.05
Best Trade [%] 350.39
Worst Trade [%] -71.82
Avg Trade [%] 0.3
Max Trade Duration 2 weeks and 5 days
Avg Trade Duration 4 days and 10 hours
Expectancy [%] 26.08
------------------------------------------------------------
Ratios
------------------------------------------------------------
Sharpe Ratio 0.2
Sortino Ratio -0.45
Calmar Ratio -1.0
Profit Factor 0.86
System Quality Number -0.38
------------------------------------------------------------
************************************************************
```
<p align="left">
<img src="stratestic/utils/drawings/backtesting_with_margin.png" style="width: 100%" />
</p>
As evident from the results, employing a leverage of `7` led to a margin call during the backtest,
showing that this particular strategy would have implied a total loss of the funds, unless more margin was
added to the positions in the meantime.
<a name="maximum-leverage"></a>
#### Calculating the maximum allowed leverage
The backtesting class also offers an API to determine the maximum permissible leverage for a backtest,
ensuring that the margin ratio remains below a specified threshold. This can be accomplished by following the
steps outlined in the following example.
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MovingAverageCrossover
symbol = "BTCUSDT"
trading_costs = 0.05
mov_avg = MovingAverageCrossover(20, 50)
vect = VectorizedBacktester(
mov_avg,
symbol,
amount=10000,
trading_costs=trading_costs,
)
vect.load_data()
vect.maximum_leverage(margin_threshold=0.8) # The margin threshold will be the maximum margin_ratio allowed during the
# backtest. If omitted, then the default value of 0.8 is used. Must be
# # between 0 and 1.
```
Which will output the maximum leverage without a margin call. In the example above, the result would be:
```shell
Out[2]: 5
```
<a name="optimization"></a>
### Optimization
<a name="brute-force"></a>
#### Brute Force
Both the iterative and vectorized backtesters offer an optimization API, enabling the discovery
of parameter combinations that yield optimal performance in a backtest, with optimization focused
on a specific metric.
The options for optimization metrics are: `Return`, `Sharpe Ratio`, `Calmar Ratio`, `Sortino Ratio`,
`Win Rate`, `Profit Factor`, `System Quality Number`, `Expectancy`, `Volatility`, `Maximum Drawdown`,
`Average Drawdown`, `Maximum Drawdown Duration`, `Average Drawdown Duration`. The default is `Return`.
The default optimization algorithm is brute force, entailing an analysis of all possible cases.
Below is an example demonstrating how to utilize this API:
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import Momentum
symbol = "BTCUSDT"
trading_costs = 0.1
mom = Momentum(30) # Initialize the strategy object with any values.
vect = VectorizedBacktester(mom, symbol=symbol, trading_costs=trading_costs) # The VectorizedBacktester class could also be used
vect.load_data() # Load the default sample data. You can pass your own DataFrame to load_data
# Pass as an argument a dictionary with the parameters as keywords
# and with a tuple with the limits to test and the desired step.
# In this case we are optimizing the strategy with the parameter 'window'
# between the values of 1000 and 1500 with a step of 10
vect.optimize(dict(window=(1000, 1500, 10)), optimization_metric='Sharpe Ratio')
```
This will output the best parameters and show the corresponding best result. For this example, it would be:
```shell
100% (50 of 50) |########################| Elapsed Time: 0:00:26 ETA: 0:00:00
Out[2]: ({'window': 1400.0}, 0.9786648787774422)
```
<a name="genetic-algorithm"></a>
#### Genetic Algorithm
In the prior illustration, we utilized the default optimizer—a brute force optimizer—which exhaustively
tests all conceivable parameter combinations. However, as the number of parameters grows, the search space
expands exponentially, rendering real-time computation unfeasible for extensive input data and parameter
combinations. In such scenarios, employing a genetic algorithm can significantly reduce the time required
to converge towards an optimum, albeit without guaranteeing the attainment of the global optimum.
Accessing a genetic algorithm solver through this optimization API is exemplified below. This API leverages
[geneal](https://github.com/diogomatoschaves/geneal) under the hood. For a comprehensive understanding of
the available parameters that can be passed to the solver, please refer to the documentation.
All the parameters are optional, but it is recommended to play with them to achieve the best results.
```python
from stratestic.backtesting import IterativeBacktester
from stratestic.strategies import MovingAverageCrossover
symbol = "BTCUSDT"
trading_costs = 0.1
mov_avg = MovingAverageCrossover(30, 200)
ite = IterativeBacktester(mov_avg, symbol, amount=1000, trading_costs=trading_costs)
ite.load_data()
opt_params = {
"sma_s": (800, 1200),
"sma_l": (1200, 1600)
}
ite.optimize(
opt_params,
optimizer='gen_alg',
pop_size=10, # population size (number of individuals)
max_gen=20, # maximum number of generations
mutation_rate=0.1, # mutation rate to apply to the population
selection_rate=0.6, # percentage of the population to select for mating
selection_strategy="roulette_wheel", # strategy to use for selection. see below for more details
fitness_tolerance=(1E-5, 10), # Loop will be exited if the best fitness value does not change more than
# 1E-5 for 10 generations
verbose=False, # Whether to print best fitness at every iteration
plot_results=True, # Whether to plot the results at the end
)
```
```shell
\ | # | 191 Elapsed Time: 0:06:38
Out[2]: ({'sma_s': 1030, 'sma_l': 1206}, 1206.09)
```
In the above example, we allowed a population of 10 individuals to evolve over 20 generations,
resulting in approximately 200 calls to the backtester. Had we opted for the brute force algorithm,
the number of calls would have skyrocketed to 160,000 (400 * 400), translating to a significantly
longer processing time on a standard machine.
<a name="strategies"></a>
### Strategies
<a name="combined-strategies"></a>
#### Combined strategies
It is possible to combine 2 or more strategies into one, by means of the `StrategyCombiner` class. The options
for combining the strategies are `Unanimous` or `Majority`. The `Unaninmous` option signals a buy or a sell
if all the individual strategy signals agree (unanimous), whereas the `Majority` method provides a buy a
or sell signal if the majority of the individual strategy signals points in one direction.
Here's an example of how that could be achieved:
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MovingAverageCrossover, Momentum, BollingerBands
from stratestic.backtesting.combining import StrategyCombiner
symbol = "BTCUSDT"
trading_costs = 0.1
mov_avg = MovingAverageCrossover(30, 200)
momentum = Momentum(70)
boll_bands = BollingerBands(20, 2)
# The strategies are passed on to StrategyCombiner as list.
combined = StrategyCombiner([mov_avg, momentum, boll_bands], method='Unanimous')
vect = VectorizedBacktester(combined, symbol, amount=1000, trading_costs=trading_costs)
vect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'
vect.run()
```
This strategy combiner class can also be optimized using the same API, with the difference that the
optimization parameters have to be passed in an array. See the next example:
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MovingAverageCrossover, Momentum
from stratestic.backtesting.combining import StrategyCombiner
symbol = "BTCUSDT"
trading_costs = 0.1
mov_avg = MovingAverageCrossover(30, 200)
momentum = Momentum(70)
# The strategies are passed on to StrategyCombiner as list.
combined = StrategyCombiner([mov_avg, momentum], method='Majority')
vect = VectorizedBacktester(combined, symbol, amount=1000, trading_costs=trading_costs)
vect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'
# The optimization parameters are passed as an array of dictionaries containing the
# parameter intervals and steps for each individual strategy.
vect.optimize([dict(sma_s=(20, 40, 2), sma_l=(100, 200, 1)), dict(window=(60, 80, 1))])
```
<a name="new-strategies"></a>
#### Create new strategies
This module comes with some default strategies ready to be used, but chances are you will want
to expand this and create your own strategies. This can be easily achieved by using the template class below,
which inherits the `StrategyMixin` class:
```python
from collections import OrderedDict
from stratestic.strategies._mixin import StrategyMixin
class MyStrategy(StrategyMixin):
"""
Description of my strategy
Parameters
----------
parameter1 : type
Description of parameter1.
parameter2 : type, optional
Description of parameter2, by default default_value.
Attributes
----------
params : OrderedDict
Parameters for the strategy, by default {"parameter1": lambda x: x}
Methods
-------
__init__(self, parameter1, parameter2=default_value, **kwargs)
Initializes the strategy object.
update_data(self)
Retrieves and prepares the data.
calculate_positions(self, data)
Calculates positions based on strategy rules.
get_signal(self, row=None)
Returns signal based on current data.
"""
def __init__(
self,
parameter1: <type>,
parameter2: <type> = <some_default_value>,
data=None,
**kwargs
):
"""
Initializes the strategy object.
Parameters
----------
parameter1 : type
Description of parameter1.
parameter2 : type, optional
Description of parameter2, by default default_value.
data : pd.DataFrame, optional
Dataframe of OHLCV data, by default None.
**kwargs : dict, optional
Additional keyword arguments to be passed to parent class, by default None.
"""
self._parameter1 = parameter1 # Each specific parameter that you want to add to the strategy
# must be initalized in this manner, with a _ followed by the name
# of the parameter
self._parameter2 = parameter2
self.params = OrderedDict(
parameter1=lambda x: <type>(x),
parameter2=lambda x: <type>(x)
)
StrategyMixin.__init__(self, data, **kwargs)
def update_data(self, data):
"""
Updates the input data with additional columns required for the strategy.
Parameters
----------
data : pd.DataFrame
OHLCV data to be updated.
Returns
-------
pd.DataFrame
Updated OHLCV data containing additional columns.
"""
super().update_data(data)
# Code to update data goes here. Check the given strategies for an example.
return data
def calculate_positions(self, data):
"""
Calculates positions based on strategy rules.
Parameters
----------
data : pd.DataFrame
OHLCV data.
Returns
-------
pd.DataFrame
OHLCV data with additional 'position' column containing -1 for short, 1 for long.
"""
data["side"] = # Code to calculate side goes here
return data
def get_signal(self, row=None):
"""
Returns signal based on current data.
Parameters
----------
row : pd.Series, optional
Row of OHLCV data to generate signal for, by default None.
Returns
-------
int
Signal (-1 for short, 1 for long, 0 for neutral).
"""
# Code to generate signal goes here
return signal
```
You would replace `MyStrategy` with the name of your strategy, and replace "Description of my strategy"
with a brief explanation of what your strategy does.
`__init__()` is where you initialize your strategy parameters. In the case of our example strategy outlined
above, `parameter1` and `parameter2` would be replaced with the actual names of your strategy's parameter(s),
and `<type>` would be replaced with the appropriate data types of your parameters.
This is very important for appropriate type checking on the frontend.
The `params` attribute is an `OrderedDict` that specifies the default parameters for your strategy.
The key is the parameter name, and the value is a lambda function that converts the user's input
into the appropriate data type.
Finally, we need to call StrategyMixin.__init__(self, data, **kwargs) in order to initialize the parent class.
`update_data()` should contain code to retrieve and prepare the data for your strategy. This is where you can
add indicators or manipulate the data and create new columns that will then be used to calculate a signal.
And example if you were developing a momentum strategy would be to calculate the moving average for the selected window.
`calculate_positions()` should contain code to calculate the positions for your strategy based
on the current data. This is where you input the logic of your strategy in a vectorized way. For the same example
of the momentum strategy, here you'd add the logic for getting the signal of when it was a BUY or a SELL.
Note that this may not be possible if your strategy is very complex. In that this method can
be ignored, and only the IterativeBacktester can be used.
`get_signal()` should contain code to generate the signal for a given row of data. The signal
should be an integer, where -1 represents a short position, 1 represents a long position,
and 0 represents a neutral position.
**In any case it is highly recommended to check the existing [strategies](https://github.com/diogomatoschaves/stratestic/tree/main/stratestic/strategies) to get a better
idea of how to implement these methods.**
<a name="machine-learning"></a>
#### Machine Learning Strategy
This library implements a machine learning strategy, which can be used for backtesting
in the same way as for the other strategies, but it has the difference that first a model is trained on
the training part of the data, and subsequently the backtest is performed on the test set.
This ensures that we're performing out-of-sample backtesting, but it also means that more data is required for
a meaningful backtest and good model fit.
What follows is a simple example to demonstrate its usage. For more detailed options please check the
docstring of this strategy.
```python
from stratestic.backtesting import VectorizedBacktester
from stratestic.strategies import MachineLearning
symbol = "BTCUSDT"
trading_costs = 0.1 # This should be in percentage, i.e. 0.1%
ml = MachineLearning(
estimator="Random Forest", # The algorithm to use for training the model.
lag_features=["returns"], # A list of columns which we want to create lagged features.
nr_lags=5, # The number of lag periods.
test_size=0.2, # The proportion between test and training data.
verbose=True,
)
vect = VectorizedBacktester(ml, symbol, amount=1000, trading_costs=trading_costs)
vect.load_data()
vect.run()
```
By default, the model is saved in a directory as indicated by the parameter `models_dir`. If one wants to load an
existing model, the model filename must be passed at initialization. The model will then be loaded, instead of
trained.
```python
from stratestic.strategies import MachineLearning
ml = MachineLearning(load_model='<filename>')
```
The `MachineLearning` strategy also provides a method for visualizing learning curves out of the box, which can be
accessed either by the strategy object or the backtester:
```python
vect.learning_curve()
ml.learning_curve() # Same as the call above.
```
<p align="middle">
<img src="stratestic/utils/drawings/learning-curves.png" style="width: 40%" />
</p>
Raw data
{
"_id": null,
"home_page": "https://github.com/diogomatoschaves/stratestic",
"name": "stratestic",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": "",
"keywords": "backtesting,trading,algotrading,trading-strategies,algorithmic-trading",
"author": "Diogo Matos Chaves",
"author_email": "diogo_chaves@hotmail.com",
"download_url": "https://files.pythonhosted.org/packages/67/7b/e4b462cf97784a54f62e78405bc8c6e90914ceb51b97caf125f4e5390f8f/stratestic-1.6.0.tar.gz",
"platform": null,
"description": "# Stratestic \ud83d\udcc8\ud83d\udcca\ud83d\udee0\ufe0f\n\n[](https://codecov.io/gh/diogomatoschaves/stratestic)\n\n[](https://badge.fury.io/py/stratestic)\n\n\nThe `stratestic` module is a Python package for backtesting, analysing and optimizing trading strategies. \nIt includes a number of pre-implemented strategies, but it is also possible to create new strategies, as well as\nto combine them. It provides a general Machine Learning strategy, which can be further tweaked to your specific needs.\n\nThe basic usage is as simple as importing a backtesting class, a strategy and run the backtest. The backtest can \nthen be refined with optimizations or by incorporating leverage. \n\nOverall it offers a powerful way to explore and experiment with different strategies, \nand to evaluate the performance under different conditions. \n\nIf you are interested in a trading bot that integrates seamlessly with this library, check out \n[MyCryptoBot](https://github.com/diogomatoschaves/MyCryptoBot).\n\n## Installation\n\n $ pip install stratestic\n\n## Usage\n\n1. [ Vectorized Backtesting ](#vectorized-backtesting)\n2. [ Iterative Backtesting ](#iterative-backtesting)\n3. [ Backtesting with leverage and margin ](#leverage) <br>\n 3.1. [ Calculating the maximum allowed leverage ](#maximum-leverage)\n4. [ Optimization ](#optimization) <br>\n 4.1 [ Brute Force ](#brute-force) <br>\n 4.2 [ Genetic Algorithm ](#genetic-algorithm)\n5. [ Strategies ](#strategies) <br>\n 5.1. [ Combined strategies](#combined-strategies) <br>\n 5.2. [ Create new strategies](#new-strategies) <br>\n 5.3. [ Machine Learning strategy ](#machine-learning)\n\n<a name=\"vectorized-backtesting\"></a>\n### Vectorized Backtesting\n\nThe `VectorizedBacktester` is a backtesting class that allows you to test trading strategies\non historical price data. It has the advantage of being faster than the iterative backtesting, but at\na cost of flexibility, as it will be hard or outright not possible to accomplish this for some more \ncomplex strategies. For all the strategies provided by this library, vectorized backtesting is supported.\n\nBelow is an example of how to use it for the `MovingAverageCrossover` strategy:\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MovingAverageCrossover\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1 # This should be in percentage, i.e. 0.1% \n\nmov_avg = MovingAverageCrossover(50, 200)\n\nvect = VectorizedBacktester( # Initializes the VectorizedBacktester class with the strategy.\n mov_avg,\n symbol,\n amount=1000,\n trading_costs=trading_costs\n)\nvect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'\nvect.run() # Runs the backtest and shows the results\n```\n\nThis will output the results in textual and graphical form.\n\n```\n************************************************************\n BACKTESTING RESULTS \n************************************************************\n Overview \n------------------------------------------------------------\nTotal Duration 4 years and 38 weeks\nStart Date 2018-05-23 13:00:00\nEnd Date 2023-02-13 01:00:00\nTrading Costs [%] 0.1\nExposure Time [%] 100.0\nLeverage [x] 1\nEquity - Initial [USDT] 1000\nEquity - Final [USDT] 3215.96\nEquity - Peak [USDT] 5356.87\n------------------------------------------------------------\n Returns \n------------------------------------------------------------\nTotal Return [%] 221.6\nAnnualized Return [%] 21.49\nAnnualized Volatility [%] 73.95\nBuy & Hold Return [%] 175.98\n------------------------------------------------------------\n Drawdowns \n------------------------------------------------------------\nMax Drawdown [%] -61.18\nAvg Drawdown [%] -8.2\nMax Drawdown Duration 1 year and 38 weeks\nAvg Drawdown Duration 3 weeks and 2 days\n------------------------------------------------------------\n Trades \n------------------------------------------------------------\nTotal Trades 267\nWin Rate [%] 32.21\nBest Trade [%] 87.77\nWorst Trade [%] -21.11\nAvg Trade [%] 0.44\nMax Trade Duration 5 weeks and 3 days\nAvg Trade Duration 6 days and 11 hours\nExpectancy [%] 5.9\n------------------------------------------------------------\n Ratios \n------------------------------------------------------------\nSharpe Ratio 0.33\nSortino Ratio 0.28\nCalmar Ratio 0.35\nProfit Factor 1.0\nSystem Quality Number -0.02\n------------------------------------------------------------\n************************************************************\n```\n\n<p align=\"left\">\n <img src=\"stratestic/utils/drawings/vectorized_results.png\" style=\"width: 100%\" />\n</p>\n\n<a name=\"iterative-backtesting\"></a>\n### Iterative Backtesting\n\nThe `IterativeBacktester` is a backtesting class that allows you to test trading strategies\non historical price data. It works by iterating through each historical data point and simulating\ntrades based on your strategy. This feature allows for a greater degree of flexibility, \nallowing you to add more complex logic to the strategies. Below is an example of how you would use this \nclass to backtest the `MovingAverageConvergenceDivergence` strategy. \n\n```python\nfrom stratestic.backtesting import IterativeBacktester\nfrom stratestic.strategies import MovingAverageConvergenceDivergence\n\nsymbol = \"BTCUSDT\"\n\nmacd = MovingAverageConvergenceDivergence(26, 12, 9)\n\nite = IterativeBacktester(macd, symbol=symbol) # Initializes the IterativeBacktester class with the strategy\nite.load_data() # Load the default sample data. You can pass your own DataFrame to load_data\nite.run() # Runs the backtest and shows the results\n```\nThis will output the results in textual and graphical form.\n\n```\n************************************************************\n BACKTESTING RESULTS \n************************************************************\n Overview \n------------------------------------------------------------\nTotal Duration 4 years and 39 weeks\nStart Date 2018-05-16 15:00:00\nEnd Date 2023-02-13 01:00:00\nTrading Costs [%] 0.0\nExposure Time [%] 100.0\nLeverage [x] 1\nEquity - Initial [USDT] 1000\nEquity - Final [USDT] 17145.68\nEquity - Peak [USDT] 29566.42\n------------------------------------------------------------\n Returns \n------------------------------------------------------------\nTotal Return [%] 1614.57\nAnnualized Return [%] 60.58\nAnnualized Volatility [%] 70.99\nBuy & Hold Return [%] 163.16\n------------------------------------------------------------\n Drawdowns \n------------------------------------------------------------\nMax Drawdown [%] -56.09\nAvg Drawdown [%] -5.46\nMax Drawdown Duration 1 year and 22 weeks\nAvg Drawdown Duration 1 week and 1 day\n------------------------------------------------------------\n Trades \n------------------------------------------------------------\nTotal Trades 3136\nWin Rate [%] 34.92\nBest Trade [%] 45.61\nWorst Trade [%] -12.84\nAvg Trade [%] 0.09\nMax Trade Duration 2 days and 14 hours\nAvg Trade Duration 13 hours and 15 minutes\nExpectancy [%] 1.72\n------------------------------------------------------------\n Ratios \n------------------------------------------------------------\nSharpe Ratio 0.84\nSortino Ratio 0.8\nCalmar Ratio 1.08\nProfit Factor 1.01\nSystem Quality Number 0.16\n------------------------------------------------------------\n************************************************************\n```\n<p align=\"left\">\n <img src=\"stratestic/utils/drawings/iterative_results.png\" style=\"width: 100%\" />\n</p>\n\n<a name=\"leverage\"></a>\n### Backtesting with leverage and margin\n\nBoth the Vectorized and Iterative backtesting classes provide users with the ability to incorporate leverage into a \nbacktest and visualize the margin ratio evolution during the backtest. This feature enables users to identify \ninstances where a margin call would occur, leading to a potential loss of all funds. The calculations follow the \nrules outlined by Binance, as detailed [here](https://www.binance.com/en/support/faq/how-to-calculate-liquidation-price-of-usd%E2%93%A2-m-futures-contracts-b3c689c1f50a44cabb3a84e663b81d93) and [here](https://www.binance.com/en/support/faq/leverage-and-margin-of-usd%E2%93%A2-m-futures-360033162192). \nIt's important to note that these calculations assume the selected margin is _Isolated_, and the position mode\nis _One Way_. Below an example is shown:\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MovingAverageCrossover\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.05\n\nmov_avg = MovingAverageCrossover(20, 150)\n\nvect = VectorizedBacktester(\n mov_avg,\n symbol,\n amount=1000,\n trading_costs=trading_costs,\n leverage=7 # Pass the desired leverage here\n)\n\nvect.load_data()\nvect.run()\n```\n\nThis will output the following results and plot:\n\n```\n************************************************************\n BACKTESTING RESULTS \n************************************************************\n Overview \n------------------------------------------------------------\nTotal Duration 4 years and 38 weeks\nStart Date 2018-05-21 11:00:00\nEnd Date 2023-02-13 01:00:00\nTrading Costs [%] 0.05\nExposure Time [%] 100.0\nLeverage [x] 7\nEquity - Initial [USDT] 1000\nEquity - Final [USDT] 0.0\nEquity - Peak [USDT] 17012.48\n------------------------------------------------------------\n Returns \n------------------------------------------------------------\nTotal Return [%] -100.0\nAnnualized Return [%] -100.0\nAnnualized Volatility [%] 144.81\nBuy & Hold Return [%] 157.53\n------------------------------------------------------------\n Drawdowns \n------------------------------------------------------------\nMax Drawdown [%] -100.0\nAvg Drawdown [%] -15.65\nMax Drawdown Duration 4 years and 7 weeks\nAvg Drawdown Duration 4 weeks and 7 hours\n------------------------------------------------------------\n Trades \n------------------------------------------------------------\nTotal Trades 82\nWin Rate [%] 28.05\nBest Trade [%] 350.39\nWorst Trade [%] -71.82\nAvg Trade [%] 0.3\nMax Trade Duration 2 weeks and 5 days\nAvg Trade Duration 4 days and 10 hours\nExpectancy [%] 26.08\n------------------------------------------------------------\n Ratios \n------------------------------------------------------------\nSharpe Ratio 0.2\nSortino Ratio -0.45\nCalmar Ratio -1.0\nProfit Factor 0.86\nSystem Quality Number -0.38\n------------------------------------------------------------\n************************************************************\n```\n<p align=\"left\">\n <img src=\"stratestic/utils/drawings/backtesting_with_margin.png\" style=\"width: 100%\" />\n</p>\n\nAs evident from the results, employing a leverage of `7` led to a margin call during the backtest, \nshowing that this particular strategy would have implied a total loss of the funds, unless more margin was\nadded to the positions in the meantime. \n\n<a name=\"maximum-leverage\"></a>\n#### Calculating the maximum allowed leverage\n\nThe backtesting class also offers an API to determine the maximum permissible leverage for a backtest, \nensuring that the margin ratio remains below a specified threshold. This can be accomplished by following the \nsteps outlined in the following example.\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MovingAverageCrossover\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.05\n\nmov_avg = MovingAverageCrossover(20, 50)\n\nvect = VectorizedBacktester(\n mov_avg,\n symbol,\n amount=10000,\n trading_costs=trading_costs,\n)\n\nvect.load_data()\nvect.maximum_leverage(margin_threshold=0.8) # The margin threshold will be the maximum margin_ratio allowed during the \n # backtest. If omitted, then the default value of 0.8 is used. Must be \n# # between 0 and 1.\n```\n\nWhich will output the maximum leverage without a margin call. In the example above, the result would be:\n\n```shell\nOut[2]: 5\n```\n\n<a name=\"optimization\"></a>\n### Optimization\n\n<a name=\"brute-force\"></a>\n#### Brute Force\n\nBoth the iterative and vectorized backtesters offer an optimization API, enabling the discovery \nof parameter combinations that yield optimal performance in a backtest, with optimization focused \non a specific metric.\n\nThe options for optimization metrics are: `Return`, `Sharpe Ratio`, `Calmar Ratio`, `Sortino Ratio`, \n`Win Rate`, `Profit Factor`, `System Quality Number`, `Expectancy`, `Volatility`, `Maximum Drawdown`,\n`Average Drawdown`, `Maximum Drawdown Duration`, `Average Drawdown Duration`. The default is `Return`.\n\nThe default optimization algorithm is brute force, entailing an analysis of all possible cases. \nBelow is an example demonstrating how to utilize this API:\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import Momentum\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1\n\nmom = Momentum(30) # Initialize the strategy object with any values. \n\nvect = VectorizedBacktester(mom, symbol=symbol, trading_costs=trading_costs) # The VectorizedBacktester class could also be used\n\nvect.load_data() # Load the default sample data. You can pass your own DataFrame to load_data\n\n# Pass as an argument a dictionary with the parameters as keywords\n# and with a tuple with the limits to test and the desired step. \n# In this case we are optimizing the strategy with the parameter 'window' \n# between the values of 1000 and 1500 with a step of 10\nvect.optimize(dict(window=(1000, 1500, 10)), optimization_metric='Sharpe Ratio')\n```\nThis will output the best parameters and show the corresponding best result. For this example, it would be:\n\n```shell\n100% (50 of 50) |########################| Elapsed Time: 0:00:26 ETA: 0:00:00\nOut[2]: ({'window': 1400.0}, 0.9786648787774422)\n```\n\n<a name=\"genetic-algorithm\"></a>\n#### Genetic Algorithm\n\nIn the prior illustration, we utilized the default optimizer\u2014a brute force optimizer\u2014which exhaustively \ntests all conceivable parameter combinations. However, as the number of parameters grows, the search space \nexpands exponentially, rendering real-time computation unfeasible for extensive input data and parameter \ncombinations. In such scenarios, employing a genetic algorithm can significantly reduce the time required \nto converge towards an optimum, albeit without guaranteeing the attainment of the global optimum.\n\nAccessing a genetic algorithm solver through this optimization API is exemplified below. This API leverages \n[geneal](https://github.com/diogomatoschaves/geneal) under the hood. For a comprehensive understanding of \nthe available parameters that can be passed to the solver, please refer to the documentation. \nAll the parameters are optional, but it is recommended to play with them to achieve the best results.\n\n\n```python\nfrom stratestic.backtesting import IterativeBacktester\nfrom stratestic.strategies import MovingAverageCrossover\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1\n\nmov_avg = MovingAverageCrossover(30, 200)\n\nite = IterativeBacktester(mov_avg, symbol, amount=1000, trading_costs=trading_costs)\nite.load_data()\n\nopt_params = {\n \"sma_s\": (800, 1200),\n \"sma_l\": (1200, 1600)\n}\n\nite.optimize(\n opt_params,\n optimizer='gen_alg',\n pop_size=10, # population size (number of individuals)\n max_gen=20, # maximum number of generations\n mutation_rate=0.1, # mutation rate to apply to the population\n selection_rate=0.6, # percentage of the population to select for mating\n selection_strategy=\"roulette_wheel\", # strategy to use for selection. see below for more details\n fitness_tolerance=(1E-5, 10), # Loop will be exited if the best fitness value does not change more than\n # 1E-5 for 10 generations\n verbose=False, # Whether to print best fitness at every iteration\n plot_results=True, # Whether to plot the results at the end\n)\n```\n\n```shell\n\\ | # | 191 Elapsed Time: 0:06:38\nOut[2]: ({'sma_s': 1030, 'sma_l': 1206}, 1206.09)\n```\n\nIn the above example, we allowed a population of 10 individuals to evolve over 20 generations, \nresulting in approximately 200 calls to the backtester. Had we opted for the brute force algorithm, \nthe number of calls would have skyrocketed to 160,000 (400 * 400), translating to a significantly \nlonger processing time on a standard machine. \n\n<a name=\"strategies\"></a>\n### Strategies\n\n<a name=\"combined-strategies\"></a>\n#### Combined strategies\n\nIt is possible to combine 2 or more strategies into one, by means of the `StrategyCombiner` class. The options\nfor combining the strategies are `Unanimous` or `Majority`. The `Unaninmous` option signals a buy or a sell\nif all the individual strategy signals agree (unanimous), whereas the `Majority` method provides a buy a \nor sell signal if the majority of the individual strategy signals points in one direction. \n\nHere's an example of how that could be achieved:\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MovingAverageCrossover, Momentum, BollingerBands\nfrom stratestic.backtesting.combining import StrategyCombiner\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1\n\nmov_avg = MovingAverageCrossover(30, 200)\nmomentum = Momentum(70)\nboll_bands = BollingerBands(20, 2)\n\n# The strategies are passed on to StrategyCombiner as list.\ncombined = StrategyCombiner([mov_avg, momentum, boll_bands], method='Unanimous')\n\nvect = VectorizedBacktester(combined, symbol, amount=1000, trading_costs=trading_costs)\nvect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'\n\nvect.run()\n```\n\nThis strategy combiner class can also be optimized using the same API, with the difference that the \noptimization parameters have to be passed in an array. See the next example:\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MovingAverageCrossover, Momentum\nfrom stratestic.backtesting.combining import StrategyCombiner\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1\n\nmov_avg = MovingAverageCrossover(30, 200)\nmomentum = Momentum(70)\n\n# The strategies are passed on to StrategyCombiner as list.\ncombined = StrategyCombiner([mov_avg, momentum], method='Majority')\n\nvect = VectorizedBacktester(combined, symbol, amount=1000, trading_costs=trading_costs)\nvect.load_data() # Load the default sample data. You can pass your own DataFrame to 'load_data'\n\n# The optimization parameters are passed as an array of dictionaries containing the \n# parameter intervals and steps for each individual strategy.\nvect.optimize([dict(sma_s=(20, 40, 2), sma_l=(100, 200, 1)), dict(window=(60, 80, 1))])\n```\n\n<a name=\"new-strategies\"></a>\n#### Create new strategies\n\nThis module comes with some default strategies ready to be used, but chances are you will want\nto expand this and create your own strategies. This can be easily achieved by using the template class below, \nwhich inherits the `StrategyMixin` class:\n\n```python\nfrom collections import OrderedDict\nfrom stratestic.strategies._mixin import StrategyMixin\n\n\nclass MyStrategy(StrategyMixin):\n \"\"\"\n Description of my strategy\n\n Parameters\n ----------\n parameter1 : type\n Description of parameter1.\n parameter2 : type, optional\n Description of parameter2, by default default_value.\n\n Attributes\n ----------\n params : OrderedDict\n Parameters for the strategy, by default {\"parameter1\": lambda x: x}\n\n Methods\n -------\n __init__(self, parameter1, parameter2=default_value, **kwargs)\n Initializes the strategy object.\n update_data(self)\n Retrieves and prepares the data.\n calculate_positions(self, data)\n Calculates positions based on strategy rules.\n get_signal(self, row=None)\n Returns signal based on current data.\n \"\"\"\n\n def __init__(\n self, \n parameter1: <type>,\n parameter2: <type> = <some_default_value>,\n data=None,\n **kwargs\n ):\n \"\"\"\n Initializes the strategy object.\n\n Parameters\n ----------\n parameter1 : type\n Description of parameter1.\n parameter2 : type, optional\n Description of parameter2, by default default_value.\n data : pd.DataFrame, optional\n Dataframe of OHLCV data, by default None.\n **kwargs : dict, optional\n Additional keyword arguments to be passed to parent class, by default None.\n \"\"\"\n self._parameter1 = parameter1 # Each specific parameter that you want to add to the strategy\n # must be initalized in this manner, with a _ followed by the name \n # of the parameter\n self._parameter2 = parameter2\n\n self.params = OrderedDict(\n parameter1=lambda x: <type>(x),\n parameter2=lambda x: <type>(x)\n ) \n\n StrategyMixin.__init__(self, data, **kwargs)\n\n def update_data(self, data):\n \"\"\"\n Updates the input data with additional columns required for the strategy.\n\n Parameters\n ----------\n data : pd.DataFrame\n OHLCV data to be updated.\n\n Returns\n -------\n pd.DataFrame\n Updated OHLCV data containing additional columns.\n \"\"\"\n super().update_data(data)\n\n # Code to update data goes here. Check the given strategies for an example.\n \n return data\n\n def calculate_positions(self, data):\n \"\"\"\n Calculates positions based on strategy rules.\n\n Parameters\n ----------\n data : pd.DataFrame\n OHLCV data.\n\n Returns\n -------\n pd.DataFrame\n OHLCV data with additional 'position' column containing -1 for short, 1 for long.\n \"\"\"\n data[\"side\"] = # Code to calculate side goes here\n\n return data\n\n def get_signal(self, row=None):\n \"\"\"\n Returns signal based on current data.\n\n Parameters\n ----------\n row : pd.Series, optional\n Row of OHLCV data to generate signal for, by default None.\n\n Returns\n -------\n int\n Signal (-1 for short, 1 for long, 0 for neutral).\n \"\"\"\n # Code to generate signal goes here\n\n return signal\n\n```\n\nYou would replace `MyStrategy` with the name of your strategy, and replace \"Description of my strategy\"\nwith a brief explanation of what your strategy does.\n\n`__init__()` is where you initialize your strategy parameters. In the case of our example strategy outlined \nabove, `parameter1` and `parameter2` would be replaced with the actual names of your strategy's parameter(s), \nand `<type>` would be replaced with the appropriate data types of your parameters. \nThis is very important for appropriate type checking on the frontend.\n\nThe `params` attribute is an `OrderedDict` that specifies the default parameters for your strategy. \nThe key is the parameter name, and the value is a lambda function that converts the user's input\ninto the appropriate data type.\n\nFinally, we need to call StrategyMixin.__init__(self, data, **kwargs) in order to initialize the parent class.\n\n`update_data()` should contain code to retrieve and prepare the data for your strategy. This is where you can \nadd indicators or manipulate the data and create new columns that will then be used to calculate a signal. \nAnd example if you were developing a momentum strategy would be to calculate the moving average for the selected window.\n\n`calculate_positions()` should contain code to calculate the positions for your strategy based \non the current data. This is where you input the logic of your strategy in a vectorized way. For the same example \nof the momentum strategy, here you'd add the logic for getting the signal of when it was a BUY or a SELL.\n\nNote that this may not be possible if your strategy is very complex. In that this method can \nbe ignored, and only the IterativeBacktester can be used.\n\n`get_signal()` should contain code to generate the signal for a given row of data. The signal \nshould be an integer, where -1 represents a short position, 1 represents a long position, \nand 0 represents a neutral position.\n\n**In any case it is highly recommended to check the existing [strategies](https://github.com/diogomatoschaves/stratestic/tree/main/stratestic/strategies) to get a better \nidea of how to implement these methods.**\n\n<a name=\"machine-learning\"></a>\n#### Machine Learning Strategy\n\nThis library implements a machine learning strategy, which can be used for backtesting \nin the same way as for the other strategies, but it has the difference that first a model is trained on \nthe training part of the data, and subsequently the backtest is performed on the test set. \nThis ensures that we're performing out-of-sample backtesting, but it also means that more data is required for \na meaningful backtest and good model fit.\n\nWhat follows is a simple example to demonstrate its usage. For more detailed options please check the \ndocstring of this strategy.\n\n```python\nfrom stratestic.backtesting import VectorizedBacktester\nfrom stratestic.strategies import MachineLearning\n\nsymbol = \"BTCUSDT\"\ntrading_costs = 0.1 # This should be in percentage, i.e. 0.1% \n\nml = MachineLearning(\n estimator=\"Random Forest\", # The algorithm to use for training the model.\n lag_features=[\"returns\"], # A list of columns which we want to create lagged features.\n nr_lags=5, # The number of lag periods.\n test_size=0.2, # The proportion between test and training data.\n verbose=True,\n)\nvect = VectorizedBacktester(ml, symbol, amount=1000, trading_costs=trading_costs)\n\nvect.load_data() \nvect.run()\n```\n\nBy default, the model is saved in a directory as indicated by the parameter `models_dir`. If one wants to load an \nexisting model, the model filename must be passed at initialization. The model will then be loaded, instead of \ntrained. \n\n```python\nfrom stratestic.strategies import MachineLearning\n\nml = MachineLearning(load_model='<filename>')\n\n```\n\nThe `MachineLearning` strategy also provides a method for visualizing learning curves out of the box, which can be \naccessed either by the strategy object or the backtester:\n\n```python\nvect.learning_curve()\nml.learning_curve() # Same as the call above.\n```\n\n<p align=\"middle\">\n <img src=\"stratestic/utils/drawings/learning-curves.png\" style=\"width: 40%\" />\n</p>\n\n\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Library for testing and analysing trading strategies.",
"version": "1.6.0",
"project_urls": {
"Homepage": "https://github.com/diogomatoschaves/stratestic",
"Repository": "https://github.com/diogomatoschaves/stratestic"
},
"split_keywords": [
"backtesting",
"trading",
"algotrading",
"trading-strategies",
"algorithmic-trading"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "079957de51b7e0ab071cd1041ebcf8d0a98bcf936c7387da050748832164d144",
"md5": "d4952cb55883c5aaae1fe8c4a375e0af",
"sha256": "2c47d04c77e8055f832b60dd9b3da7673acb556977dc6ee8577729e0543a074e"
},
"downloads": -1,
"filename": "stratestic-1.6.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "d4952cb55883c5aaae1fe8c4a375e0af",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 4195796,
"upload_time": "2024-03-06T02:34:40",
"upload_time_iso_8601": "2024-03-06T02:34:40.565686Z",
"url": "https://files.pythonhosted.org/packages/07/99/57de51b7e0ab071cd1041ebcf8d0a98bcf936c7387da050748832164d144/stratestic-1.6.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "677be4b462cf97784a54f62e78405bc8c6e90914ceb51b97caf125f4e5390f8f",
"md5": "1746d19409df569d62e88a7114333454",
"sha256": "86e18f91061f29b70e1d5bcff608b5323efe179954a6aa2f96a89e9d3f6604da"
},
"downloads": -1,
"filename": "stratestic-1.6.0.tar.gz",
"has_sig": false,
"md5_digest": "1746d19409df569d62e88a7114333454",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 4179617,
"upload_time": "2024-03-06T02:34:43",
"upload_time_iso_8601": "2024-03-06T02:34:43.476271Z",
"url": "https://files.pythonhosted.org/packages/67/7b/e4b462cf97784a54f62e78405bc8c6e90914ceb51b97caf125f4e5390f8f/stratestic-1.6.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-03-06 02:34:43",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "diogomatoschaves",
"github_project": "stratestic",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "stratestic"
}
JSON
