simple_back package¶

A backtester with minimal setup and sensible defaults.

simple_back is a backtester providing easy ways to test trading strategies, often ones that use external data, with minimal amount of code, while avoiding time leaks.

Submodules¶

simple_back.backtester module¶

class simple_back.backtester.Backtester[source]¶

Bases: object

The Backtester object is yielded alongside the current day and event (open or close) when it is called with a date range, which can be of the following forms. The Backtester object stores information about the backtest after it has completed.

Examples

Initialize with dates as strings:

bt['2010-1-1','2020-1-1'].run()
# or
for day, event, b in bt['2010-1-1','2020-1-1']:
    ...

Initialize with dates as datetime.date objects:

bt[datetime.date(2010,1,1),datetime.date(2020,1,1)]

Initialize with dates as int:

bt[-100:] # backtest 100 days into the past

add_line(**kwargs)[source]¶: Adds a vertical line on the plot on the current date + event.

add_metric(key, value)[source]¶

Called inside the backtest, adds a metric that is visually tracked.

Parameters

key (str) – the metric name
value (float) – the numerical value of the metric

property balance¶

Get the current or starting balance.

Examples

Get the current balance:

bt.balance.current

Get the starting balance:

bt.balance.start

Return type: Balance

log(text)[source]¶

Adds a log text on the current day and event that can be accessed using logs after the backtest has completed.

Parameters: text (str) – text to log

property logs¶

Returns a pd.DataFrame for logs collected during the backtest.

Return type: DataFrame

long(symbol, **kwargs)[source]¶

Enter a long position of the given symbol.

Parameters

symbol (str) – the ticker to buy
kwargs – one of either “percent” as a percentage of total value (cash + positions), “absolute” as an absolute value, “percent_available” as a percentage of remaining funds (excluding positions) “nshares” as a number of shares

property metrics¶

Get a dataframe of all metrics collected during the backtest(s).

Return type: DataFrame

property pf¶

Shorthand for portfolio, returns the backtesters portfolio.

Return type: Portfolio

price(symbol)[source]¶

Get the current price of a given symbol.

Parameters: symbol (str) – the ticker
Return type: float

run()[source]¶: Run the backtesters strategies without using an iterator. This is only possible if strategies have been set using BacktesterBuilder.strategies().

short(symbol, **kwargs)[source]¶

Enter a short position of the given symbol.

Parameters

symbol (str) – the ticker to short
kwargs – one of either “percent” as a percentage of total value (cash + positions), “absolute” as an absolute value, “percent_available” as a percentage of remaining funds (excluding positions) “nshares” as a number of shares

show(start=None, end=None)[source]¶

Show the backtester as a plot.

Parameters

start – the start date
end – the end date

property strategies¶: Provides access to sub-strategies, returning a StrategySequence.

property summary¶

Get a dataframe showing the last and overall values of all metrics collected during the backtest. This can be helpful for comparing backtests at a glance.

Return type: DataFrame

property timestamp¶: Returns the current timestamp, which includes the correct open/close time, depending on the calendar that was set using BacktesterBuilder.calendar()

class simple_back.backtester.BacktesterBuilder[source]¶

Bases: object

Used to configure a Backtester and then creating it with build()

Example

Create a new Backtester with 10,000 starting balance which runs on days the NYSE is open:

bt = BacktesterBuilder().balance(10_000).calendar('NYSE').build()

balance(amount)[source]¶

Required, set the starting balance for all Strategy objects run with the Backtester

Parameters: amount (int) – The starting balance.
Return type: BacktesterBuilder

build()[source]¶

Build a Backtester given the previous configuration.

Return type: Backtester

calendar(calendar)[source]¶

Optional, set a pandas market calendar to be used. If not called, “NYSE” is used.

Parameters: calendar (str) – the calendar identifier

Return type: BacktesterBuilder

clear_metrics()[source]¶

Optional, remove all default metrics, except PortfolioValue, which is needed internally.

Return type: BacktesterBuilder

compare(strategies)[source]¶

Optional, alias for BacktesterBuilder.strategies(), should be used when comparing to BuyAndHold of a ticker instead of other strategies.

Parameters: strategies (List[Union[Callable[[date, str, Backtester], None], Strategy, str]]) – should be the string of the ticker to compare to, but Strategy objects can be passed as well

data(data)[source]¶

Optional, add a DataProvider to use external data without time leaks.

Parameters: data (DataProvider) – The data provider.
Return type: BacktesterBuilder

live_metrics(every=10)[source]¶

Optional, shows all metrics live in output. This can be useful when running simple-back from terminal.

Parameters

every (int) – how often metrics should be updated
events ((in) –
10 = 5 days) (e.g.) –

Return type

BacktesterBuilder

live_plot(every=None, metric='Total Value', figsize=None, min_y=0, blocking=False)[source]¶

Optional, shows the backtest results live using matplotlib. Can only be used in notebooks.

Parameters

every (Optional[int]) – how often metrics should be updated (in events, e.g. 10 = 5 days) the regular default is 10, blocking default is 100
metric (str) – which metric to plot
figsize (Optional[Tuple[float, float]]) – size of the plot
min_y (int) – minimum value on the y axis, set to None for no lower limit
blocking (bool) – will disable threading for plots and allow live plotting in terminal, this will slow down the backtester significantly

Return type

BacktesterBuilder

live_progress(every=10)[source]¶

Optional, shows a live progress bar using tqdm, either as port of a plot or as text output.

Return type: BacktesterBuilder

metrics(metrics)[source]¶

Optional, set additional Metric objects to be used.

Parameters: metrics (Union[Metric, List[Metric]]) – one or more Metric objects
Return type: BacktesterBuilder

name(name)[source]¶

Optional, name will be set to “Backtest 0” if not specified.

Set the name of the strategy run using the Backtester iterator.

Parameters: name (str) – The strategy name.
Return type: BacktesterBuilder

no_live_metrics()[source]¶

Disables showing live metrics.

Return type: BacktesterBuilder

no_live_plot()[source]¶

Disables showing live plots.

Return type: BacktesterBuilder

no_live_progress()[source]¶

Disables the live progress bar.

Return type: BacktesterBuilder

prices(prices)[source]¶

Optional, set the DailyPriceProvider used to get prices during a backtest. If this is not called, YahooPriceProvider is used.

Parameters: prices (DailyPriceProvider) – The price provider.
Return type: BacktesterBuilder

slippage(slippage=0.0005)[source]¶

Optional, sets slippage which will create a (lower bound) strategy. The orginial strategies will run without slippage.

Parameters: slippage (int) – the slippage in percent of the base price, default is equivalent to quantopian default for US Equities

strategies(strategies)[source]¶

Optional, sets Strategy objects to run.

Parameters: strategies (List[Union[Callable[[date, str, Backtester], None], Strategy, str]]) – list of Strategy objects or tickers to BuyAndHold
Return type: BacktesterBuilder

trade_cost(trade_cost)[source]¶

Optional, set a Fee to be applied when buying shares. When not set, NoFee is used.

Parameters: trade_cost (Union[Fee, Callable[[float, float], Tuple[float, int]]]) – one ore more Fee objects or callables.
Return type: BacktesterBuilder

class simple_back.backtester.Portfolio(bt, positions=[])[source]¶

Bases: collections.abc.MutableSequence

A portfolio is a collection of Position objects, and can be used to liquidate() a subset of them.

attr(attribute)[source]¶

Get a list of values for a certain value for all posititions

Parameters: attribute (str) – String name of the attribute to get. Can be any attribute of Position.
Return type: List

property df¶

Return type: DataFrame

filter(func)[source]¶

Filters positions using any :class`.Callable`

Parameters: func (Callable[[Position], bool]) – The function/callable to do the filtering.
Return type: Portfolio

insert(index, value)[source]¶

S.insert(index, value) – insert value before index

Return type: None

liquidate(nshares=- 1, _bt=None)[source]¶

Liquidate all positions in the current “view” of the portfolio. If no view is given using [‘some_ticker’], filter(), Portfolio.long() or Portfolio.short(), an attempt to liquidate all positions is made.

Parameters: nshares (int) – The number of shares to be liquidated. This should only be used when a ticker is selected using [‘some_ticker’].

Examples

Select all MSFT positions and liquidate them:

bt.portfolio['MSFT'].liquidate()

Liquidate 10 MSFT shares:

bt.portfolio['MSFT'].liquidate(nshares=10)

Liquidate all long positions:

bt.portfolio.long.liquidate()

Liquidate all positions that have lost more than 5% in value. We can either use filter() or the dataframe as indexer (in this case in combination with the pf shorthand):

bt.pf[bt.pf.df['profit_loss_pct'] < -0.05].liquidate()
# or
bt.pf.filter(lambda x: x.profit_loss_pct < -0.05)

property long¶

Returns a view of the portfolio (which can be treated as its own Portfolio) containing all long positions.

Return type: Portfolio

property short¶

Returns a view of the portfolio (which can be treated as its own Portfolio) containing all short positions.

Return type: Portfolio

property total_value¶

Returns the total value of the portfolio.

Return type: float

class simple_back.backtester.Position(bt, symbol, date, event, nshares, uid, fee, slippage=None)[source]¶

Bases: object

Tracks a single position in a portfolio or trade history.

property initial_value¶

Returns the initial value of the position, including fees.

Return type: float

property nshares¶

Returns the number of shares in the position.

Return type: int

property order_type¶

Returns “long” or “short” based on the position type.

Return type: str

property price¶

Returns the current price if the position is held in a portfolio. Returns the last price if the position was liquidated and is part of a trade history.

Return type: float

property profit_loss_abs¶

Returns the profit/loss associated with the position (not including commission) in absolute terms.

Return type: float

property profit_loss_pct¶

Returns the profit/loss associated with the position (not including commission) in relative terms.

Return type: float

property value¶

Returns the current market value of the position.

Return type: float

property value_pershare¶

Returns the value of the position per share.

Return type: float

class simple_back.backtester.StrategySequence(bt)[source]¶

Bases: object

A sequence of strategies than can be accessed by name or int index. Returned by Backtester.strategies and should not be used elsewhere.

Examples

Access by str:

bt.strategies['Some Strategy Name']

Access by int:

bt.strategies[0]

Use as iterator:

for strategy in bt.strategies:
    # do something

simple_back.backtester.display_side_by_side(bts)[source]¶

simple_back.data_providers module¶

class simple_back.data_providers.CachedProvider(debug=False)[source]¶

Bases: abc.ABC

clear_cache()[source]¶

get_cache(key)[source]¶

get_key(key)[source]¶

Return type: str

in_cache(key)[source]¶

Return type: bool

abstract property name¶

rem_cache(key)[source]¶

set_cache(key, val, expire_days=None)[source]¶

class simple_back.data_providers.DailyDataProvider(debug=False)[source]¶

Bases: simple_back.data_providers.CachedProvider

abstract property columns¶

Return type: List[str]

abstract property columns_order¶

Return type: List[int]

abstract get(symbol, date, event)[source]¶

Return type: DataFrame

set_date_event(date, event)[source]¶

class simple_back.data_providers.DailyPriceProvider(highlow=True, debug=False)[source]¶

Bases: simple_back.data_providers.DailyDataProvider

property columns¶

property columns_order¶

abstract get(symbol, date, event)[source]¶

Return type: DataFrame

class simple_back.data_providers.DataProvider(debug=False)[source]¶

Bases: simple_back.data_providers.CachedProvider

abstract dates(symbol)[source]¶

abstract get(datetime, symbol=None)[source]¶

abstract property name¶

class simple_back.data_providers.SpProvider(debug=False)[source]¶

Bases: simple_back.data_providers.WikipediaProvider

property name¶

process_html(html, symbol)[source]¶

transform_symbol(symbol)[source]¶

class simple_back.data_providers.WikipediaProvider(debug=False)[source]¶

Bases: simple_back.data_providers.DataProvider

dates(symbol)[source]¶

get(datetime, symbol)[source]¶

get_revisions(title)[source]¶

property name¶

abstract process_html(html, symbol)[source]¶

transform_symbol(symbol)[source]¶

class simple_back.data_providers.YahooFinanceProvider(highlow=False, adjust_prices=True, debug=False)[source]¶

Bases: simple_back.data_providers.DailyPriceProvider

get(symbol, date, event)[source]¶

Return type: DataFrame

property name¶

simple_back.exceptions module¶

exception simple_back.exceptions.BacktestRunError(message)[source]¶: Bases: Exception

exception simple_back.exceptions.InsufficientCapitalError(message)[source]¶: Bases: Exception

exception simple_back.exceptions.LongShortLiquidationError(message)[source]¶: Bases: Exception

exception simple_back.exceptions.MissingMetricsError(metrics, message)[source]¶: Bases: Exception

exception simple_back.exceptions.NegativeValueError(message)[source]¶: Bases: Exception

exception simple_back.exceptions.PriceUnavailableError(symbol, requested_date, message)[source]¶: Bases: Exception

exception simple_back.exceptions.TimeLeakError(current_date, requested_date, message)[source]¶: Bases: Exception

simple_back.fees module¶

class simple_back.fees.Fee[source]¶

Bases: abc.ABC

Abstract Callable that calculates the total cost and number of shares given an asset price and the capital to be allocated to it. It returns the total cost and the number of shares aquired for that cost.