QuantConnect Community

Actually, QuantConnect has LimitOrders and StopLoss orders built in.

I use them like this, when placing an order:
LimitOrder(symbol, -quantity, currentPrice * (1m + profitTargetPercent))
StopMarketOrder(symbol, -quantity, currentPrice * (1m - stopLossPercent))

Sorry if this sounds like a dumb question, but how do I set the currentPrice parameter to mirror real time live price?

I create LimitOrders and StopMarketOrders at the same time that I make a purchase. Thus, currentPrice is the purchase price, and I set the stop loss or profit target relative the purchase price, exactly like what you are doing in your code example.

Maybe some sample code will help. Here's an example algo that buys stock, sets a StopLoss, then updates the StopLoss.

Thanks for the code, just another question. Based on the code that I'm reading does the stop loss mimic a "trailing stop" or is it hard set (static and non moving)? http://imgur.com/levofQV

The stop loss created in this line of code is a "hard set":
StopLoss = StopMarketOrder(Symbol, -quantity, currentPrice * (1m - StopLossPercent));

If you want to mimic a trailing stop, you have to manually update the stop loss order ticket like I did in the line of code you highlighted:
StopLoss.Update(new UpdateOrderFields{ StopPrice = currentPrice * (1m - StopLossPercent) });

Once I put the code you suggested into my program, my algorithm kept buying every second it was on instead of buying on a moving average cross. Heres a sample of my code. http://imgur.com/yDdtrd0

It's typically easier for others to help if you attach the full project, then we can edit it directly and re-attach the fixed project (you need to attach backtest as well). In response to your question, I see a couple things immediately. First, I think your buy/sell logic should be an if/else block so they can never fire on the same timestep. Secondly, it looks like you perform your market order via the Order function, but then you also call SetHoldings which will also submit a market order. Likewise, in your sell function you call Liquidate and then SetHoldings(0.0). The SetHoldings function tries to set your current holdings to the targeted percent of portfolio value (1 = 100% TotalPortfolioValue). So if you liquidate (close everything) and then call SetHoldings(0.0), then nothing will happen.

Another thing, when submitting your stop loss/take profit orders (via StopMarketOrder and LimitOrder) you need to set the quantity to negative, since you want those orders to close your position.

Heres the project that I'm working on Michael

Here's my fixes for your algo with StopMarketOrder and LimitOrder. I put some descriptions in the comments.

Hold up everyone. Question time:

What about this situation during backtesting.

CloseHour 1) I made a stopLoss order now:  StopLoss@165
CloseHour 2) Price drops to 168
CloseHour 3) Price drops to 10.

Now because of the way our system works, it seems to be my stop loss will now be called, WAY too late - I have gone way past my stoploss. So this means any kind of attemplt to Liquidate orders could happen way too late if done inside OnData.

How does QuantConnect deal with the concept of brokers taking care of StopLosses for us? If I am trying to set my risk at 1% its basically impossible because this is happening. Unless I decide to use tick data it seems impossible to emulate how stop losses will really work.

In backtest, order fills depend on resolution, simply because there is no higher resolution data points between two data points of the selected resolution. This is not an issue in live mode using QuantConnect paper trading account, because limit and stop orders are filled with the ticks that arrive in the live stream.
When we are trading with a real brokerage paper or real account, QuantConnect's order types are converted into the correspondent brokerage order type and placed. Their fills are, then, managed by the brokerage and, once they are filled, their API notifies our algorithm in QuantConnect.

Alex explained it well, it is as I suspected.

I supposed it would be a lot to ask for some better broker emulation. For example if we are only feeding hourly data, the system could calculate and emulate a straight line of "tick" data between the 2 hourly points. It's extreme, but would take care of this problem - perhaps as some kind of option.

Or I supposed we should only be downloading tick data.

Ian Worthington, you just need to request higher resolution data and you'll get a real fill price. You can use custom fill models if you'd like to mock up a fake data point between two bars but the fill won't be realistic. 

Levitikon made this gem a few months ago; I think it might help others in this thread:

Jared Broad, is there any document that briefly describes how this application may be used as a robo-trader or exchange which may include data flow and/or sequence diagrams?

Larry Smith, please open a new discussion, since your question is offtopic in this thread and other new users may want to look into the answer later.

Thanks Jon,

I tried to implement something similar with your code and combined it with top gainers data. I tried to run the backtest again and never got the same results. I'm quite new but I wanna focus on the risk portion correctly. I noticed my trailing stop might've never implemented a sell off. 

I'll look into Jared's suggestion of a stop loss & limit order cancelation when one or the other is executed. 

Regarding Jonathan Evans fixes to  StopMarketOrder and LimitOrder. 

This line gives me an error : 

Hi Christian,

I've attached a backtest in which I've taken your code and made the necessary modifications to make some basic orders. Generally speaking, the Python equivalent of a C# private variable is one that is defined using the self. prefix, which initializes it globally, such as
 

Thanks a lot! 

That cleard up quite a bit, however when I am running a backtest, it seams to fill all orders at once, and it does not seam to reackt at SL ot TP? resaulting in a buy and hold.

Hi Christian,

Here are a few places that could go wrong:

1. in line 112-113, you might want to reconsider the use of stop order and limit order. Here is some information as reference.

2. since you defined some percentage (self.tpPercent and self.slPercent), does it make more sense to multiply by (1 +/- percentage) instead of directly add/substract them?

3. in line 129, filledOrderid => filledOrderId

4. in line 131 & 134, I believe you want to compare order ids. This would work as you expected:

And you could use self.Liquidate() to cancel any unfilled order.

Here is my attempt. Hope it helps!

Hi Jack,

Regarding your post above with the Python backtest, how does the script deal with SL and TP executed in the same bar? I think I read in one of Jared's posts somewhere that anything triggered in the OnOrderEvent method will only affect the backtest in the next OnData cycle. Therefore - in the scenario a SL and TP trigger on the same bar, both will be triggered (before one is cancelled) causing you (in your case above) to go negative on the trade in the opposite direction instead of neutral (0).

Any help greatly appreciated - i'm just starting out with QC. Thanks, Blake

Hi Blake,

If both the SL and TP are triggered in the same bar, then our position will go negative, and I've shown an example in the attached backtest. To address this, we can set our SL and TP farther out, increase granularity (i.e. use Minute Resolution over Daily), or constantly check our positions to make sure any negative positions are liquidated.

Best,
Shile Wen

Hi Shile, thank you. This has more or less confirmed my thinking.

One question - when you say constantly check positons - do you mean there is a way to cancel an order in the same bar on the basis of another order filled in the same bar? My understanding is we cannot do this but I want to be 100% clear.

Thanks

Hi Blake,

No, Hi Blake,No, I meant constantly check positions as in check if self.Portfolio[symbol].Quantity < 0 to make sure we don't have a negative position.

Best,
Shile Wen

Hi,

I understand this is an old thread but is the triple order method to have the stop loss and take profit compatible with Crypto?

I am asking this because when I insert the orders below I get the following error:
Order Error: id: 1096, Insufficient buying power to complete order (Value:-XXX), Reason: Your portfolio holds [holdings] [symbols], [holdings] [symbols] of which are reserved for open orders, but your Sell order is for [holdings] [symbols]. Cash Modeling trading does not permit short holdings so ensure you only sell what you have, including any additional open orders.

SetHoldings(symbol, targetSize)  # Enter the position
holdings = Portfolio[symbol].Quantity
LimitOrder(symbol, -holdings, takeProfitPrice)  # Set the take profit
StopMarketOrder(symbol, -holdings, stopLossPrice)  # Set the stop loss

I assume the last order, for the stop loss, gives error because then there would be 2 open orders for -holdings quantity which could be interpreted as a short even though the 2 orders should not trigger together.
Any idea how to solve this?

Thanks!

Hi Francesco Baldisserri,
The above problem is occurring because the cash model will not allow us to have a negative quantity. We have 3 ways to implement this.
Set the Stop Market, handle the take profit manually
 

crypto = self.Securities[symbol]

quantity = self.CalculateOrderQuantity(symbol, targetSize)

min_notional = crypto.SymbolProperties.MinimumOrderSize

if abs(quantity) * crypto.Price > min_notional:

    self.entryTicket = self.MarketOrder(symbol, quantity)

    self.StopMarketOrder(symbol, -quantity, stopLossPrice)

    self.limit_price = ...

or, set the Limit, handle the stop loss manually
 

crypto = self.Securities[symbol]

quantity = self.CalculateOrderQuantity(symbol, targetSize)

min_notional = crypto.SymbolProperties.MinimumOrderSize

if abs(quantity) * crypto.Price > min_notional:

    self.entryTicket = self.MarketOrder(symbol, quantity)

    self.LimitOrder(symbol, -quantity, takeProfirPrice)

    self.stop_price = ...

or place only half of quantity for the Stop and Limit
 

crypto = self.Securities[symbol]

quantity = self.CalculateOrderQuantity(symbol, targetSize / 2)

min_notional = crypto.SymbolProperties.MinimumOrderSize

if abs(quantity * 2) * crypto.Price > min_notional:

    self.entryTicket = self.MarketOrder(symbol, quantity * 2)

    self.LimitOrder(symbol, -quantity, takeProfirPrice)

    self.StopMarketOrder(symbol, -quantity, stopLossPrice)

in this case, call Liquidate whether one or the other is filled, because Liquidate closes the remaining half and cancels the other order. We have created a demo algorithm implementing the above. Refer to the attached backtest.
Best,
Varad Kabade

Hi Varad,

I suspected we could not escape from the need to check at least one of these conditions manually. I suspect this would be tricky for the performance if working with a large universe but the mechanism is clear.

Thanks,
Francesco