Creating a Simple Strategy

Open In Colab

In this tutorial, we will build a crossover strategy that exclusively buys and sells JNUG. It will enter a long position whenever the price goes above its 20-day MA and enter a short position when it goes below.

Building a Backtest

Use .BacktesterBuilder to configure and create a .Backtester object.

[2]:

from simple_back.backtester import BacktesterBuilder

builder = (
   BacktesterBuilder()
   .name('JNUG 20-Day Crossover')
   .balance(10_000)
   .calendar('NYSE')
   .compare(['JNUG']) # strategies to compare with
   .live_progress() # show a progress bar using tqdm
   .live_plot() # we assume we are running this in a Jupyter Notebook
)

Running the Backtest

We can now treat the bt object like an iterator with a date index.

[3]:

bt = builder.build()
for day, event, b in bt['2019-1-1':'2020-1-1']:
    pass
    # code here will be called
    # on 'open' and 'close' events on trading days
../_images/intro_quickstart_6_0.png

Note that we wrote .compare(['JNUG']) before so we could compare our strategy to just buying and holding JNUG. Because we do not do anything so far, the JNUG 20-Day Crossover strategy results in a flat line.

Note:

.BacktesterBuilder.compare normally only accepts a list of .Strategy objects, but for each string that is passed, it automatically creates a .BuyAndHold strategy for the symbol represented by the string.

Creating a Strategy

We are now ready to create the strategy. Although not necessary for this strategy, we will only act on market open by checking that event == 'open'

We then calculate the 20-day moving average of JNUGs close prices.

[4]:

bt = builder.no_live_plot().build()
for day, event, b in bt['2019-1-1':'2020-1-1']:
    if event == 'open':
        jnug_ma = b.prices['JNUG',-20:]['close'].mean()

100%|██████████| 504/504 [00:02<00:00, 179.12it/s]

Note:

The second indexer of b.prices (implemented in .DailyPriceProvider) can be a date range or single date that allows you to use int, dateutil.relativedelta.relativedelta or datetime.date values.

We can check if the current price is above or below the moving average using b.price('JNUG').

[5]:

bt = builder.no_live_plot().build()
for day, event, b in bt['2019-1-1':'2020-1-1']:
    if event == 'open':
        jnug_ma = b.prices['JNUG',-20:]['close'].mean()
        if b.price('JNUG') > jnug_ma:
            pass
            # price is above MA
        if b.price('JNUG') < jnug_ma:
            pass
            # price is below MA

100%|██████████| 504/504 [00:01<00:00, 285.25it/s]

Now we should only need to buy when the price is above, and sell when the price is below the MA.

[6]:

bt = builder.no_live_plot().build()
for day, event, b in bt['2019-1-1':'2020-1-1']:
    if event == 'open':
        jnug_ma = b.prices['JNUG',-20:]['close'].mean()
        if b.price('JNUG') > jnug_ma:
            b.long('JNUG', percent=1) # as percent of total value
        if b.price('JNUG') < jnug_ma:
            b.short('JNUG', percent=1) # as percent of total value

  1%|          | 3/504 [00:00<00:06, 79.13it/s]

InsufficientCapitalError:
                        not enough capital available:
                        ordered 1 * 10563.648455632008
                        with only 449.31090733719975 available

But as you might already have expected, this fails with an .InsufficientCapitalError This is because we repeatedly try to invest 100% of our assets into JNUG, even when we already hold JNUG shares.

Note:

.Backtester.long and .Backtester.short allow for use of the following arguments: percent: order as percent of total value; percent_available: order as percent of the available cash; absolute: order as absolute cash value; nshares: order as number of shares.

To fix the previous error we now check if the corresponding positions are already in our portfolio, and liquidate positions of the wrong kind. We liquidate long positions before we go short and vice versa:

[7]:

bt = builder.build()
for day, event, b in bt['2019-1-1':'2020-1-1']:
    if event == 'open':
        jnug_ma = b.prices['JNUG',-20:]['close'].mean()

        if b.price('JNUG') > jnug_ma:
            if not b.portfolio['JNUG'].long: # check if we already are long JNUG
                b.portfolio['JNUG'].short.liquidate() # liquidate any/all short JNUG positions
                b.long('JNUG', percent=1) # long JNUG

        if b.price('JNUG') < jnug_ma:
            if not b.portfolio['JNUG'].short: # check if we already are short JNUG
                b.portfolio['JNUG'].long.liquidate() # liquidate any/all long JNUG positions
                b.short('JNUG', percent=1) # short JNUG
../_images/intro_quickstart_19_0.png

While running, the .Backtester.portfolio (with the shorthand .Backtester.pf) object keeps track of positions. If we hold a position at the very end of the backtest, it will remain in this object. .Backtester.portfolio.df holds a dataframe of said positions.

[9]:

b.portfolio.df # shorthand: b.pf.df

[9]:
date event initial_value nshares order_type price profit_loss_abs profit_loss_pct start_price symbol value value_pershare
0 2019-12-24 open 4238.278808 6.0 long 832.202148 754.934082 0.178123 706.379801 JNUG 4993.212891 832.202148

Inspecting Metrics

Now that our Backtest is complete, we can use .Backtester.metrics and .Backtester.summary to get more details. Both of these will return DataFrames.

Metrics

[10]:

bt.metrics

[10]:
Max Drawdown (%) Annual Return Portfolio Value Total Value Total Return (%) Daily Profit/Loss
Backtest Date Event
JNUG 20-Day Crossover 2019-01-02 open -71.121124 0.525525 0.000000 10000.000000 0.000000 0.000000
close -71.121124 0.525525 9696.819855 10146.130762 1.461308 0.000000
2019-01-03 open -71.121124 0.525525 10114.337548 10563.648456 5.636485 563.648456
close -71.121124 0.525525 10751.050232 11200.361139 12.003611 1054.230377
2019-01-04 open -71.121124 0.525525 10197.839717 10647.150624 6.471506 83.502169
... ... ... ... ... ... ... ... ...
JNUG (Buy & Hold) 2019-12-27 close -49.921226 1.799052 16358.200195 16807.511103 68.075111 -718.145142
2019-12-30 open -49.921226 1.799052 16615.727700 17065.038607 70.650386 -343.369146
close -49.921226 1.799052 17553.713379 18003.024286 80.030243 1195.513184
2019-12-31 open -49.921226 1.799052 18221.607392 18670.918299 86.709183 1605.879692
close -49.921226 1.799052 17476.245117 17925.556025 79.255560 -77.468262

1008 rows × 6 columns

Summary

[11]:

bt.summary

[11]:
Max Drawdown (%) Annual Return Portfolio Value (Last Value) Total Value (Last Value) Total Return (%) (Last Value) Daily Profit/Loss (Last Value)
Backtest
JNUG 20-Day Crossover -71.121124 0.525525 4993.212891 5276.118290 -47.238817 -22.133789
JNUG (Buy & Hold) -49.921226 1.799052 17476.245117 17925.556025 79.255560 -77.468262

Strategies

You can also use .Backtester.strategies which returns a .StrategySequence, with their own .Backtester.metrics and .Backtester.summary.

[12]:

bt.strategies['JNUG 20-Day Crossover'].metrics

[12]:
Max Drawdown (%) Annual Return Portfolio Value Total Value Total Return (%) Daily Profit/Loss
Backtest Date Event
JNUG 20-Day Crossover 2019-01-02 open -71.121124 0.525525 0.000000 10000.000000 0.000000 0.000000
close -71.121124 0.525525 9696.819855 10146.130762 1.461308 0.000000
2019-01-03 open -71.121124 0.525525 10114.337548 10563.648456 5.636485 563.648456
close -71.121124 0.525525 10751.050232 11200.361139 12.003611 1054.230377
2019-01-04 open -71.121124 0.525525 10197.839717 10647.150624 6.471506 83.502169
... ... ... ... ... ... ... ... ...
JNUG (Buy & Hold) 2019-12-27 close -49.921226 1.799052 16358.200195 16807.511103 68.075111 -718.145142
2019-12-30 open -49.921226 1.799052 16615.727700 17065.038607 70.650386 -343.369146
close -49.921226 1.799052 17553.713379 18003.024286 80.030243 1195.513184
2019-12-31 open -49.921226 1.799052 18221.607392 18670.918299 86.709183 1605.879692
close -49.921226 1.799052 17476.245117 17925.556025 79.255560 -77.468262

1008 rows × 6 columns

The advantage of using using .StrategySequence is that we can also use it for plotting.

[13]:

bt.strategies[0].show()
../_images/intro_quickstart_30_0.png