[Andrew Bee] "Ah, there's the rub. You still have discretionary inputs to an otherwise automatic system. " Actually, I don't. I trade the same system day in and day out but I noticed volatility seems to affect it a lot so there is always room for improvement. "I've found it's best to find one set of parameters and stick with them day in, day out. The trick is finding the right ones that give you a consistent edge over a long period of time. Attempting to tune (my system at least) on a day by day basis kills its performance." Yeah, that's the trap. You can improve your system if your improvements are proven to be better on most of the long-term history, not just recent trades. --- "Volatility is variable ya know. Depends on timeframes ect. and the market can slow down on a dime. " Pa(b)st Prime, how do you gauge volatility in a day trading session and adapt to it ?
Exactly. I found that 3 months is too short of a period to go on to set my parameters. Tuning them to one 3-month period results in disastrous performance in the next 3 months. 6 months comes out better, but can still be thrown off. I went with 2 years in the end, mainly because that's all the data I had to test with. That was tick-level data and 2 years covers many different types of market conditions. Forward testing, and then live trading, performance was consistent with tested performance. I never change the system without verifying it historically first.
An initial two point hard stop set to break even after two points and a trailing stop set at either 1/2% or 1% of the price the index is trading at is AFAIK customary to catch intraday swings without suffering stops or chops no matter the volatility. Whichever is greater, percentiles or ATRs is what I use.
I use the VIX and the stock's ATR as parameters in my equations to determine position size, stop point, and exit point. Parameters change every day so my strategy is dynamic.
Xuanxue - Very good info. Thanks ! bespoke - That's exactly what I'm aiming at. I just have to use the right parameters then it's a matter of tweaking them. Thank you for your input. How many months of data have you tested your system on and how many months of forward testing have you done before going live ? My system has no negative months in the past 14 months but can have huge P/L swings during the month. Do your systems have similar swings ? Good comments guys, keep them coming
I optimized for 1 year of data (2006) and forward tested on 2002-5, and 2007-present. By using volatility parameters I noticed the equity curve became smoother. Without it, there were some really wild months like in 2002-3, and parts of 2007, and the rest of the months the strategy performed "okay" with a lot of flat months. But by using volatility parameters, there was less variance between each month which is what I like personally (for psychological sake, yes, I'm weak). It also reduced my max drawdown too. I think from something like 20% to 5% in backtested results. eidt: it should be 20% to 12% based on compounding profits
"But by using volatility parameters, there was less variance between each month which is what I like personally (for psychological sake, yes, I'm weak). " I know what you mean. Following a system can be very hard psychologically. When you're in a drawdown and you get stressed, just not taking one trade can highly handicap your P/L for the month if it's a big winner. "It also reduced my max drawdown too. I think from something like 20% to 5% in backtested results." That's amazing !
bespoke - Do you use the daily or intraday ATR for your day trading system ? Could you give an example of how you would use it. Also, could you elaborate on how you use the VIX ? Edit : PM me if you prefer.
"You don't need a weatherman to tell you which way the wind blows" wrote Bob Dylan. Fcuk the Vix. You can clearly tell when volitility increases in the ES by watching the price action and yes, when it gets hairy and scary, you need to increase your stops from initial entry as well as increase your expected exit. If you can't do this, then don't play. Sit back and wait for the flames to die down. You can get stopped out in a nanosecond with a 1-1.5 point stop (after entry) when the ES starts to get busy and that excludes, of course, scheduled news announcements.