Comment on Streak just got a whole lot better
commented on
26 Dec 2018, 07:00 PM
Hi Vivek,
Optimisation of stoploss and take profit values can be done with various methods. The most suitable way is to use an exit condition which show fading of trend and momentum or start of reversal. If you absolutely want to do optimisation of stoploss and take profit values, we have listed some of them which require some degree of manual calculations and discretionary decisions.
1) Using ATR :
Average of the ATR value can be calculated for a certain period. Lets call it x [x = (ATR1+ATR2+ATR3….ATRn)/n]. To get the ATR value that accounts only for recent price move, we should use a smaller value for n. Use n = 14 to 20. Find out what percentage of the current price is ‘x’. Let’s call it y [y=(x/Current Price)x100]. A value just less than y can be used as take profit %. To account for risk/reward ratio, y/2 can be used as a stoploss value which will keep the risk/reward ratio as 1:2. Now different backtests can be carried out by varying stoploss and take profit % around these values and experimenting with different risk/reward ratios.
This method can also be tested with different ATR multipliers. To do this, x should be multiplied with a certain number, and then this value can be used instead of x in rest of the calculations. If you want to allow more room for the stock to move, a multiplier greater than 1 can be used. If less room needs to be given, a multiplier less than 1 should be used.
2) Using Standard Deviation :
Calculate the standard deviation of prices over a period. Greater the period, more robust will be the results. Let the standard deviation be x. Calculate what percentage of current price is 2x, because we will be using 2 standard deviations. Let’s call it y [y=(2x/current price)x100). A value that is just a little less than y can be now used as a take profit %. Stoploss can be calculated based on your desired risk/reward ratio (Suppose your desired risk/reward ratio is 1:2, we will use stoploss as y/2). Test this with various risk/reward ratios.
Note: It’s important to note that these values will be different for different stocks so make sure to carryout different calculations for different stocks. You also need to know that these methods rely on some discretionary decisions and hence will not produce the most optimum results that can be expected from an optimisation algorithm. Also, the values obtained through these methods uses the historical volatility at that moment of time and hence calculations will need to be carried out periodically to account for changing volatility. For best results, use both the methods and select the one that gives best performance.
We would also like to let you know that we will soon be releasing an option on streak that will be able to carry out optimisation automatically for different parameters in the algo being backtested.
Thanks
Streak