Multiple stop loss issued

Eduardo Menges Mattje

I have this code for a RSI strategy that I've tried to implement Stop Loss.

import backtrader as bt
from datetime import datetime


class firstStrategy(bt.Strategy):
    params = (
        ('stopLoss', 0.25),
        ('upperband', 75),
        ('lowerband', 25),
        ('period', 9)
    )

    def __init__(self):
        self.rsi = bt.indicators.RelativeStrengthIndex(self.data.close, period=9, safediv=True)
        self.order = None
        self.stopOrder = None

    def notify_order(self, order):
        if order.status in [order.Submitted, order.Accepted]:
            return

        if order.status in [order.Completed]:
            if order.isbuy():
                if self.params.stopLoss:
                    self.stopOrder = self.close(price=order.executed.price, exectype=bt.Order.StopTrail,
                                                trailpercent=self.params.stopLoss)
        self.order = None

    def next(self):
        if not self.position:
            if self.rsi <= 25:
                self.buy()
        elif self.rsi >= 75:
            self.close()

    # def next(self):
    #     if not self.position:
    #         if self.rsi <= 25:
    #             self.buy()
    #     elif self.rsi >= 75:
    #         self.close()


startcash = 100

cerebro = bt.Cerebro()

cerebro.addstrategy(firstStrategy)

data = bt.feeds.GenericCSVData(
    dataname='NEOUSDT-15m-data.csv',
    fromdate=datetime(2020, 2, 1, 0, 0),
    todate=datetime(2020, 6, 1, 0, 0),
    timeframe=bt.TimeFrame.Minutes,
    compression=15,
    dtformat='%Y-%m-%d %H:%M:%S'
)

cerebro.adddata(data)

cerebro.broker.setcash(startcash)

cerebro.addsizer(bt.sizers.PercentSizer, percents=90)  # Set the order size

cerebro.addanalyzer(bt.analyzers.DrawDown, _name='Drawdown')

cerebro.broker.setcommission(commission=0.0001, margin=False)

cerebro.run()

portvalue = round(cerebro.broker.getvalue(), 2)
pnl = round(portvalue - startcash, 2)

print('Final Portfolio Value: ${}'.format(portvalue))
print('P/L: ${}'.format(pnl))
cerebro.plot(style='candlestick')

The results were quite ridiculous, and the graph showed that multiple Stop Loss order were issued, even tho the order was supossed to be closed, as you can see:
alt text
So what can I do to stop it?

Eduardo Menges Mattje

I've changed the code, and now I'm trying to use a bracket order. However, the trailpercent arg doesn't do a thing. It can be 0.1, it can be 1, or even 100, and the results are same. This is the new code:

import backtrader as bt
from datetime import datetime


class firstStrategy(bt.Strategy):
    params = (
        ('stopLoss', 0.25),
        ('upperband', 75),
        ('lowerband', 25),
        ('period', 9)
    )

    def __init__(self):
        self.rsi = bt.indicators.RelativeStrengthIndex(self.data.close, period=self.p.period, upperband=self.p.upperband, lowerband=self.p.lowerband, safediv=True)

    def next(self):
        if not self.position and self.rsi <= self.p.lowerband:
                # self.buy()
                self.buy_bracket(exectype=bt.Order.Market, price=self.data.close, limitexec=None, stopexec=bt.Order.StopTrail, trailpercent=0.1, stopprice=self.data.close)
        elif self.rsi >= self.p.upperband:
            self.close()


startcash = 100

cerebro = bt.Cerebro()

cerebro.addstrategy(firstStrategy)

data = bt.feeds.GenericCSVData(
    dataname='NEOUSDT-15m-data.csv',
    fromdate=datetime(2020, 2, 1, 0, 0),
    todate=datetime(2020, 6, 1, 0, 0),
    timeframe=bt.TimeFrame.Minutes,
    compression=15,
    dtformat='%Y-%m-%d %H:%M:%S'
)

cerebro.adddata(data)

cerebro.broker.setcash(startcash)

cerebro.addsizer(bt.sizers.PercentSizer, percents=90)  # Set the order size

cerebro.addanalyzer(bt.analyzers.DrawDown, _name='Drawdown')

cerebro.broker.setcommission(commission=0.0001, margin=False)

cerebro.run()

portvalue = round(cerebro.broker.getvalue(), 2)
pnl = round(portvalue - startcash, 2)

print('Final Portfolio Value: ${}'.format(portvalue))
print('P/L: ${}'.format(pnl))
cerebro.plot(style='candlestick')

If this is intended behaviour, can I use another kind of order?

Eduardo Menges Mattje

BUMP....

ab_trader

Looks like your first script with StopTrail orders worked as expected. You used 25% stop from the price of order execution. Your first orders were executed at prices around 14.5 - 15.5 and you stops were set to approx 11.5 - 12.0. You can see them executed at that spike of the cash, 6 red triangles in a row.

ab_trader

@Eduardo-Menges-Mattje said in Multiple stop loss issued:

However, the trailpercent arg doesn't do a thing. It can be 0.1, it can be 1, or even 100

Just to clarify (Docs - Orders - StopTrail)

Percentage based distance: trailpercent=0.02 (i.e.: 2%)

You start with 0.1 which is 10% and your price range is much less than this value. You need to use much less % of stop in order to hit it before your self.close() closes position in regular way.

You may want to add more logging into your script, this will help you to understand what is going on.

Eduardo Menges Mattje

@ab_trader said in Multiple stop loss issued:

Just to clarify (Docs - Orders - StopTrail)

Percentage based distance: trailpercent=0.02 (i.e.: 2%)

You start with 0.1 which is 10% and your price range is much less than this value. You need to use much less % of stop in order to hit it before your self.close() closes position in regular way.
Even with trailpercent=0.01, the result is the same as 0.1 or any other value. As I said, it is as if trailpercent is completely disregarded.

You may want to add more logging into your script, this will help you to understand what is going on.
Ok. This is the code with the logging:

import backtrader as bt
from datetime import datetime


class firstStrategy(bt.Strategy):
    params = dict(
        stopLoss=0.01,
        upperband=75,
        lowerband=25,
        period=9
    )

    def log(self, txt, dt=None):
        dt = dt or self.data.datetime[0]
        if isinstance(dt, float):
            dt = bt.num2date(dt)
        print('%s, %s' % (dt.isoformat(), txt))

    def notify_order(self, order):
        if order.status in [order.Completed]:
            if order.isbuy():
                self.log(
                    'BUY EXECUTED, Price: %.2f, Cost: %.2f, Comm %.2f' %
                    (order.executed.price,
                     order.executed.value,
                     order.executed.comm))

            else:
                self.log('SELL EXECUTED, Price: %.2f, Cost: %.2f, Comm %.2f' %
                         (order.executed.price,
                          order.executed.value,
                          order.executed.comm))

        self.order = None

    def __init__(self):
        self.rsi = bt.indicators.RelativeStrengthIndex(period=self.p.period, upperband=self.p.upperband,
                                                       lowerband=self.p.lowerband, safediv=True)

    def next(self):
        if not self.position and self.rsi <= self.p.lowerband:
            # self.buy()
            self.buy_bracket(exectype=bt.Order.Market, price=self.data.close, limitexec=None,
                             stopexec=bt.Order.StopTrail, trailpercent=self.p.stopLoss, stopprice=self.data.close)
            self.log('BUY CREATE, exectype Market, price %.2f' % self.data.close[0])
        elif self.position and self.rsi >= self.p.upperband:
            self.close()
            self.log('BUY CREATE, exectype Close, price %.2f' % self.data.close[0])


startcash = 100

cerebro = bt.Cerebro()

cerebro.addstrategy(firstStrategy)

data = bt.feeds.GenericCSVData(
    dataname='NEOUSDT-15m-data.csv',
    fromdate=datetime(2020, 2, 1, 0, 0),
    todate=datetime(2020, 6, 1, 0, 0),
    timeframe=bt.TimeFrame.Minutes,
    compression=15,
    dtformat='%Y-%m-%d %H:%M:%S'
)

cerebro.adddata(data)

cerebro.broker.setcash(startcash)

cerebro.addsizer(bt.sizers.PercentSizer, percents=90)  # Set the order size

cerebro.addanalyzer(bt.analyzers.DrawDown, _name='Drawdown')

cerebro.broker.setcommission(commission=0.0001, margin=False)

cerebro.run()

portvalue = round(cerebro.broker.getvalue(), 2)
pnl = round(portvalue - startcash, 2)

print('Final Portfolio Value: ${}'.format(portvalue))
print('P/L: ${}'.format(pnl))
cerebro.plot(style='candlestick')

This is the log with trailpercent=0.01:
https://controlc.com/90e3b126
This is the log with trailpercent=0.10:
https://controlc.com/80136684

ab_trader

I'll take a look. Also please log ohlc prices and rsi values.

Eduardo Menges Mattje

@ab_trader said in Multiple stop loss issued:

I'll take a look. Also please log ohlc prices and rsi values.

This is the csv file: https://drive.google.com/file/d/1zx6Ywdts38sHVJqr9ZYfYV5Izxh_MBVk/view?usp=sharing

Eduardo Menges Mattje

BUMP....

Eduardo Menges Mattje

So it turns out that this is intended behaviour.
StopTrail compares the latest price, not the executed price.

ab_trader

Good you figured this out. Maybe this is true in your case since you use Market order as a first order.For this type of order the execution price is unknown at the moment when order is issued. You passed close prices and maybe stoptrail started to work from that value.

I didn't have a chance to go thru all the info you dropped here. Especially having no RSI values and having outputs started from the middle of the whole price array. But I am still thinking that my guess above is true. Even 1% stop trail is large for your set of prices.

Eduardo Menges Mattje

@ab_trader said in Multiple stop loss issued:

Good you figured this out. Maybe this is true in your case since you use Market order as a first order.For this type of order the execution price is unknown at the moment when order is issued. You passed close prices and maybe stoptrail started to work from that value.

I didn't have a chance to go thru all the info you dropped here. Especially having no RSI values and having outputs started from the middle of the whole price array. But I am still thinking that my guess above is true. Even 1% stop trail is large for your set of prices.

Is there a way for StopLoss to trigger only when executed price - current price <= percentage?

ab_trader

bt's stop order triggers when the current price is lower than the stop price. so stop price need to be defined as executed price * ( 1 - stop percent).

i've played with your data and script, and it seems to me that the issue is in the stoptrail side of the bracket order. I've changed it to the regular stop order, it worked as expected but with the StopTrail the results were weird.

In your case if you want to do the following:

issue market buy order based on rsi value
than issue stoptrail order with the position open
and issue closing order based on rsi condition with the cancellation of the existing stoptrail order

than you may want to take a look on the OCO orders.