Agreed on the second part. It's quite revealing that most traders' concept of money management can be reduced to this: "use a stop" lol
So both of you do not encourage doing backtesting over the entire history of a market and instead restrict testing to recent data? lol
Not especially. My concept for backtesting is the following : 1) A system that works only on a single stock/future/FX has a very large probability to be just random and nothing else. If I have two stocks/futures/FX that seem to work with this system, everytime I optimize my parameters or try to change something to the system, I only accept theses changes if the adjusted performance is enhanced in both stocks/futures/FX. 2) A system that is not consistent troughout time is a bad system. In every case, I try to optimize the time consistency of my system, not its performance over a large time interval. Having a system that brings 5% a month with a monthly volatility of 1% is a better thing than a system that brings 8 months of negative returns and then one month with a 200% return ! In this last case, probabilities are very high your system is nothing more than random.
The true flaw in backtesting is thinking that there are a myriad of systems. In fact there are only two: reversal and continuation.
But do you backtest over the entire history of the market you're going to trade? For example, if you're building a system for the ES, would you backtest the system over data starting from the date ES started trading?
It depends on the strategy. You can expect that your system's life expectancy is roughly equal to your backtesting sample. But this is a very empirical rule of thumb. I personaly never go further than 95 in the past, except in some very precise cases I can't describe here.
You're also assuming at that the parameters of your sytem are static. If you system is self optimising, it should never have to stop working. Runningbear
If u modify the parameters (optimise), you're essentially making a new system. If the market keep changing and the parameters keep getting optimised to fit the current market, sooner or later u're gonna run out of parameters that work both in the past and now. The thing i'm getting at is "backtesting over entire history of a market". But I made the mistake of assuming people backtest as far back as they can go. If traders backtest over recent data and ignore historical data further back then yes, i would reckon that there are unlimited amount of systems.