示例#1
0
def backtest(
    strategy,
    data,  # Treated as csv path is str, and dataframe of pd.DataFrame
    commission=COMMISSION_PER_TRANSACTION,
    init_cash=INIT_CASH,
    plot=True,
    fractional=False,
    verbose=1,
    sort_by="rnorm",
    sentiments=[],
    strats={},  # Only used when strategy = "multi"
    return_history=False,
    return_plot=False,
    channel="",
    symbol="",
    allow_short=False,
    short_max=1.5,
    figsize=(30, 15),
    data_class=None,
    data_kwargs={},
    plot_kwargs={},
    fig=None,
    **kwargs,
):
    """Backtest financial data with a specified trading strategy

    Parameters
    ----------------
    strategy : str or an instance of `fastquant.strategies.base.BaseStrategy`
        see list of accepted strategy keys below
    data : pandas.DataFrame
        dataframe with at least close price indexed with time
    commission : float
        commission per transaction [0, 1]
    init_cash : float
        initial cash (currency implied from `data`)
    plot : bool
        show plot backtrader (disabled if `strategy`=="multi")
    verbose : int
        Verbose can take values: [0, 1, 2, 3], with increasing levels of verbosity (default=1).
    sort_by : str
        sort result by given metric (default='rnorm')
    sentiments : pandas.DataFrame
        df of sentiment [0, 1] indexed by time (applicable if `strategy`=='senti')
    strats : dict
        dictionary of strategy parameters (applicable if `strategy`=='multi')
    return_history : bool
        return history of transactions (i.e. buy and sell timestamps) (default=False)
    return_plot: bool
        return the plot (if you want to save the plot) (default=True)
    channel : str
        Channel to be used for notifications - e.g. "slack" (default=None)
    symbol : str
        Symbol to be referenced in the channel notification if not None (default=None)
    allow_short : bool
        Whether to allow short selling, with max set as `short_max` times the portfolio value (default=False)
    short_max : float
        The maximum short position allowable as a ratio relative to the portfolio value at that time point (default=1.5)
    figsize : tuple
        The size of the figure to be displayed at the end of the backtest (default=(30, 15))
    data_class : bt.feed.DataBase
        Custom backtrader database to be used as a parent class instead bt.feed. (default=None)
    data_kwargs : dict
        Datafeed keyword arguments (empty dict by default)
    plot_kwargs : dict
        Argument for function cerebro.plot() (empty dict by default)
    {0}
    """
    # Setting initial support for 1 cpu
    # Return the full strategy object to get all run information
    cerebro = bt.Cerebro(stdstats=False, maxcpus=1, optreturn=False)
    cerebro.addobserver(bt.observers.Broker)
    cerebro.addobserver(bt.observers.Trades)
    cerebro.addobserver(bt.observers.BuySell)

    # Convert all non iterables and strings into lists
    kwargs = {
        k: v if isinstance(v, Iterable) and not isinstance(v, str) else [v]
        for k, v in kwargs.items()
    }

    # Add logging parameters based on the `verbose` parameter
    logging_params = get_logging_params(verbose)
    kwargs.update(logging_params)

    # Add Strategy
    strat_names = []
    strat_name = None
    if strategy == "multi" and strats is not None:
        for strat, params in strats.items():
            cerebro.optstrategy(
                STRATEGY_MAPPING[strat],
                init_cash=[init_cash],
                commission=commission,
                channel=channel,
                symbol=symbol,
                allow_short=allow_short,
                fractional=fractional,
                short_max=short_max,
                **params,
            )
            strat_names.append(strat)
    else:

        # Allow instance of BaseStrategy or from the predefined mapping
        if not isinstance(strategy, str) and issubclass(strategy, bt.Strategy):
            strat_name = (strategy.__name__
                          if hasattr(strategy, "__name__") else str(strategy))
        else:
            strat_name = strategy
            strategy = STRATEGY_MAPPING[strategy]

        cerebro.optstrategy(
            strategy,
            init_cash=[init_cash],
            commission=commission,
            channel=channel,
            symbol=symbol,
            fractional=fractional,
            allow_short=allow_short,
            short_max=short_max,
            **kwargs,
        )
        strat_names.append(strat_name)

    # Apply Total, Average, Compound and Annualized Returns calculated using a logarithmic approach
    cerebro.addanalyzer(btanalyzers.Returns, _name="returns")
    cerebro.addanalyzer(btanalyzers.SharpeRatio, _name="mysharpe")
    cerebro.addanalyzer(btanalyzers.DrawDown, _name="drawdown")
    cerebro.addanalyzer(btanalyzers.TimeDrawDown, _name="timedraw")
    cerebro.addanalyzer(btanalyzers.TradeAnalyzer,
                        _name="tradeanalyzer")  #trade analyzer

    cerebro.broker.setcommission(commission=commission)

    # Initalize and verify data
    pd_data, data, data_format_dict = initalize_data(data, strat_name, symbol,
                                                     data_class, sentiments,
                                                     data_kwargs)
    cerebro.adddata(pd_data)
    cerebro.broker.setcash(init_cash)
    # Allows us to set buy price based on next day closing
    # (technically impossible, but reasonable assuming you use all your money to buy market at the end of the next day)
    cerebro.broker.set_coc(True)
    if verbose > 0:
        print("Starting Portfolio Value: %.2f" % cerebro.broker.getvalue())

    # clock the start of the process
    tstart = time.time()
    stratruns = cerebro.run()

    # clock the end of the process
    tend = time.time()

    if verbose > 0:
        # print out the result
        print("Time used (seconds):", str(tend - tstart))

    # Get History, Optimal Parameters and Strategy Metrics
    sorted_combined_df, optim_params, history_dict = analyze_strategies(
        init_cash,
        stratruns,
        data,
        strat_names,
        strategy,
        strats,
        sort_by,
        return_history,
        verbose,
        **kwargs,
    )

    # Plot

    if plot and strategy != "multi":
        # Plot only with the optimal parameters when multiple strategy runs are required
        if sorted_combined_df.shape[0] != 1:
            if verbose > 0:
                print("=============================================")
                print("Plotting backtest for optimal parameters ...")
            _, fig = backtest(
                strategy,
                data,
                plot=plot,
                verbose=0,
                sort_by=sort_by,
                return_plot=return_plot,
                plot_kwargs=plot_kwargs,
                **optim_params,
            )
        else:
            fig = plot_results(cerebro, data_format_dict, figsize,
                               **plot_kwargs)

    if return_history and return_plot:
        return sorted_combined_df, history_dict, fig
    elif return_history:
        return sorted_combined_df, history_dict
    elif return_plot:
        return sorted_combined_df, fig
    else:
        return sorted_combined_df
示例#2
0
def backtest(
    strategy,
    data,  # Treated as csv path is str, and dataframe of pd.DataFrame
    commission=COMMISSION_PER_TRANSACTION,
    init_cash=INIT_CASH,
    plot=True,
    verbose=1,
    sort_by="rnorm",
    sentiments=None,
    strats=None,  # Only used when strategy = "multi"
    data_format=None,  # No longer needed but will leave for now to warn removal in a coming release
    return_history=False,
    channel=None,
    symbol=None,
    allow_short=False,
    short_max=1.5,
    figsize=(30, 15),
    **kwargs,
):
    """Backtest financial data with a specified trading strategy

    Parameters
    ----------------
    strategy : str or an instance of `fastquant.strategies.base.BaseStrategy`
        see list of accepted strategy keys below
    data : pandas.DataFrame
        dataframe with at least close price indexed with time
    commission : float
        commission per transaction [0, 1]
    init_cash : float
        initial cash (currency implied from `data`)
    plot : bool
        show plot backtrader (disabled if `strategy`=="multi")
    sort_by : str
        sort result by given metric (default='rnorm')
    sentiments : pandas.DataFrame
        df of sentiment [0, 1] indexed by time (applicable if `strategy`=='senti')
    strats : dict
        dictionary of strategy parameters (applicable if `strategy`=='multi')
    return_history : bool
        return history of transactions (i.e. buy and sell timestamps) (default=False)
    channel : str
        Channel to be used for last day notification - e.g. "slack" (default=None)
    verbose : int
        Verbose can take values: [0, 1, 2, 3], with increasing levels of verbosity (default=1).
    symbol : str
        Symbol to be referenced in the channel notification if not None (default=None)
    {0}
    """

    if data_format:
        errmsg = "Warning: data_format argument is no longer needed since formatting is now purely automated based on column names!"
        errmsg += "\nWe will be removing this argument in a coming release!"
        warnings.warn(errmsg, DeprecationWarning)
        print(errmsg)

    # Setting inital support for 1 cpu
    # Return the full strategy object to get all run information
    cerebro = bt.Cerebro(stdstats=False, maxcpus=1, optreturn=False)
    cerebro.addobserver(bt.observers.Broker)
    cerebro.addobserver(bt.observers.Trades)
    cerebro.addobserver(bt.observers.BuySell)

    # Convert all non iterables and strings into lists
    kwargs = {
        k: v if isinstance(v, Iterable) and not isinstance(v, str) else [v]
        for k, v in kwargs.items()
    }

    # Add logging parameters based on the `verbose` parameter
    logging_params = get_logging_params(verbose)
    kwargs.update(logging_params)

    # Add Strategy
    strat_names = []
    strat_name = None
    if strategy == "multi" and strats is not None:
        for strat, params in strats.items():
            cerebro.optstrategy(
                STRATEGY_MAPPING[strat],
                init_cash=[init_cash],
                commission=commission,
                channel=None,
                symbol=None,
                allow_short=allow_short,
                short_max=short_max,
                **params,
            )
            strat_names.append(strat)
    else:

        # Allow instance of BaseStrategy or from the predefined mapping
        if not isinstance(strategy, str) and issubclass(strategy, bt.Strategy):
            strat_name = str(strategy)
        else:
            strat_name = strategy
            strategy = STRATEGY_MAPPING[strategy]

        cerebro.optstrategy(
            strategy,
            init_cash=[init_cash],
            commission=commission,
            channel=None,
            symbol=None,
            allow_short=allow_short,
            short_max=short_max,
            **kwargs,
        )
        strat_names.append(strat_name)

    # Apply Total, Average, Compound and Annualized Returns calculated using a logarithmic approach
    cerebro.addanalyzer(btanalyzers.Returns, _name="returns")
    cerebro.addanalyzer(btanalyzers.SharpeRatio, _name="mysharpe")
    cerebro.addanalyzer(btanalyzers.DrawDown, _name="drawdown")
    cerebro.addanalyzer(btanalyzers.TimeDrawDown, _name="timedraw")

    cerebro.broker.setcommission(commission=commission)

    # Initalize and verify data
    pd_data, data, data_format_dict = initalize_data(
        data, strat_name, sentiments
    )
    cerebro.adddata(pd_data)
    cerebro.broker.setcash(init_cash)
    # Allows us to set buy price based on next day closing
    # (technically impossible, but reasonable assuming you use all your money to buy market at the end of the next day)
    cerebro.broker.set_coc(True)
    if verbose > 0:
        print("Starting Portfolio Value: %.2f" % cerebro.broker.getvalue())

    # clock the start of the process
    tstart = time.time()
    stratruns = cerebro.run()

    # clock the end of the process
    tend = time.time()

    if verbose > 0:
        # print out the result
        print("Time used (seconds):", str(tend - tstart))

    # Get History, Optimal Parameters and Strategy Metrics
    sorted_combined_df, optim_params, history_dict = analyze_strategies(
        stratruns,
        data,
        strat_names,
        strategy,
        strats,
        sort_by,
        return_history,
        verbose,
        **kwargs,
    )

    # Plot
    if plot and strategy != "multi":
        # Plot only with the optimal parameters when multiple strategy runs are required
        if sorted_combined_df.shape[0] != 1:
            if verbose > 0:
                print("=============================================")
                print("Plotting backtest for optimal parameters ...")
            backtest(
                strategy,
                data,
                plot=plot,
                verbose=0,
                sort_by=sort_by,
                **optim_params,
            )
        else:
            plot_results(cerebro, data_format_dict, figsize)

    if return_history:

        return sorted_combined_df, history_dict
    else:
        return sorted_combined_df