Attn: Experienced Backtesters -- Best Way to Model an Intraday Stop Loss?

[catmango]

Stop losses have always been difficult for me to model on backtesting systems, particularly on overnight holds. However, for a system that is based on intraday trades (I'm using 5 min bars), what's the best way to model a stop loss?

My approach would be the following (using a long position as an example):

1) Assuming that I enter long at P(E) and I set my stop at P(S), where P(S) < P(E). For each subsequent 5 min bar, perform steps 2 and 3.

2) If the 5 min bar open B(O) is <= P(S) then set the exit price P(X) at the open less slippage S --

P(X) = B(O) - S

3) If B(O) is > P(S), then if the 5 min bar low B(L) is <= P(S), then set the exit price at P(S) less slippage --

P(X) = B(L) - S

Since I'm backtesting an e-mini intraday trading system, I'm assuming that the slippage is one tick (.5 for NQ, .25 for ES).

Any comments? Would this be the best way to model a stop loss, or is there a better way?

Thanks!

 

[PetaDollar]

Whatever algorithm you program should act like you trade.
In this case, the algorithm would model reality very well if:

• after buying, you take no action until the open of the next bar
• if the open is less than or equal to your stop you instantly market sell
• if the open is higher than your stop, you instantly place a stop order

If this is the case then your algorithm is 'good'.
Also you can get rid of the slippage in this case.
I know i'll be taken out of context on this one, but with the emini
and the above rules, the backtest will model reality more closely without that fudge factor.