Optimization of sharpe ratio based on arithmetric return

[stefanmsy]

Hi,

I am trying to optimize the sharpe ratio calculated based on daily arithmetric return rather than return based on trade. Though I believe I managed to do that successfully, it seems that I need to calculate that dynamically with next() inside the strategy. As per what I read in other discussion, it is not recommended as it can be very slow. Any other recommendation to handle this?

Thanks!

'''
Arithmetric return on each day
Default to close price change but will be set to 0 if there is no position
'''
def arith_return(p_values, n):
_a = pd.Series(p_values).pct_change()
_a.iloc[:(n-1)] = 0
return _a

'''
Cumulative arithmetric return for benchmark and
empty series to store cumulative arithmetric return
'''
def arith_cum_return(p_values, n):
_bm = pd.Series(p_values).pct_change().cumsum()
_blank = _bm.copy()
_blank.iloc[:] = 0
return _bm, _blank

'''
Buy when the difference between price and sma > threshold
Short when the difference < - threshold
'''
class SmaDiff(Strategy):
n_sma = 70 # look back windows for MA
th_diff = 0.125 # threshold for MA diff

def init(self):
    self.sma = self.I(sma, self.data.Close, self.n_sma)
    self.maDiff = self.I(ma_diff, self.data.Close, self.n_sma, plot=True)
    self.pnl = self.I(arith_return, self.data.Close, self.n_sma, plot=False)    # placeholder for arithmetric return
    self.cum = self.I(arith_cum_return, self.data.Close , self.n_sma, plot=True)  # placeholder for arithmetric cumulative return

def next(self):
    '''
    Trade with fixed amount
    '''
    _price = self.data.Close[-1]
    _size = ORDER_SIZE // _price

    '''
    Calculate arithmetic return and cumulative return
    '''
    if self.position.size == 0:
        self.pnl[-1] = 0
    elif self.position.size < 0:
        self.pnl[-1] = -self.pnl[-1]
   '''
      cum[0] is cumulative return for benchmark and 
      cum[1] is empty series to store calculated cumulative return for strategy based on return on that day
   '''
    self.cum[1][-1] = self.cum[1][-2] + self.pnl[-1]

    if self.maDiff > self.th_diff:
        if self.position.size == 0:
            self.buy(size=_size)
    elif self.maDiff < -self.th_diff:
        if self.position.size == 0:
            self.sell(size=_size)
    else:
        self.position.close()

df contains the ohlc data

bt = Backtest(df, SmaDiff, cash=CAPITAL, trade_on_close=True)

stats = bt.run()

sharpe_ratio = lambda s: s._strategy.pnl.mean() / s._strategy.pnl.std() * (365 ** 0.5)

stats, heatmap = bt.optimize(
n_sma=range(20, 100, 8),
th_diff=np.linspace(0.0, 0.2, 8).tolist(),
maximize=sharpe_ratio,
return_heatmap=True)