All you have to do to be as rich as Buffet. If you've ever looked at Berkshire Hathaways Book Value growth rate, you'll see that all it was doing was beating the S&P by 10% annually. Specifically, this was my opinion, which may be a little biased for obvious reasons: <i>Anyway, my opinion is stick with me on Pairs Trading QID QLD Scalper or go to www.collective2.com/go/pairsqidqld . Look at the S&P, and then decide if that's where you'd like to take your trading, because that's what an S&P beating system's curve looks like, which comes as a surprise to the paper traders on the site. Not that subscribing will give you instant riches, but that I am confident I'll beat the S&P significantly over the next 20 years. </i> Some people had asked me what system I'd recommend, but I'm not going to really recommend anything. The reason is that finding a system that beats the S&P500 by a wide margin isn't that hard to do. It's accepting that you will be correlated significantly with the S&P that gets most people in a hang up. System's are either death by thousands of papercuts or the jumbo loss at once. If you build one to not be either of those, you should be all right and attain your goal of beating the S&P by 10% annually. In 40 years, you'll be as rich as Buffet. Good luck, and I'm open to comments.
It depends on what your edge is. Correlation to the S&P can be a very good thing if your edge works towards defining relative value. By minimizing the downside of correlation and maximizing the upside a good relative value system should exaggerate upside moves and dampen down moves. This is done by positioning when the S&P is not anywhere near value (or, trading the relatively "undervalued" components of the S&P during downtrends). IMO, edges that do not rely on correlation are some of the easiest to screw up and, and can also potentially be the most effective. The reason most non-correlation type systems fail is that the developer makes false assumptions about the conditions during which the system should work. High volatility, for example, should be a blessing to both correlated and non-correlated systems. Trendiness should work well for correlation based systems and possibly not so much for non-correlated systems. As a system developer myself, I would be most skeptical of systems that do not perform well during high volatility. Looking at your system, I believe your results could be significantly improved by looking at a greater variety of products relative to your base market (QQQQ). Also, the DD during the last few months is disconcerting. A correlation based model should be doing well during high vola... Mike
Sure we'd all like no DD. But I don't see any systems that haven't drewdown significantly during the last 9 months. Negatively correlated systems obviously will be up, suggesting they only do well in short markets. And we're not talking large correlations, either. You see any just -0.01 means they're negatively correlated, and mostly a short system. Mine at 0.07 means I'm slightly correlated with the market, and mostly long.
On this I don't trade correlations, and haven't found evidence correlation based models are anything but statistical anomalies when they do happen.
So what you're saying is that, your system, Up 50% since Oct 2007, and mine from the same time frame up 12% to the S&P's 50% loss and nearly 56% drawdown is bad?
Statistical anomalies are the heart of this business... whether those anomalies occur with reliable enough frequency so as an edge can be gained is what I try to determine. For example, I can create numerous systems that trade only a few times per year and have them show very good historical results. Going forward, I would have little faith in such systems due mainly to the sample size. That said, given the massive amount of market data available, one's edge can be stress-tested across thousands - if not tens of thousands of possible trade setups/outcomes. What I'm getting at is that reliable market edges demonstrate a certain type of historical behavior - one which usually shows better performance with higher volatility due primarily to the fact that higher volatility implies less efficient price discovery as well as active (i.e. moving) markets. I'm curious as to the fundamental price movement the C2 system you have is trying to exploit. If it's a pair's approach, then you're doing something I haven't seen before because you don't trade a correlation. If its an indicator based approach then you're either doing some sort of vola. breakout/trend ID approach or a mean reversion approach. In either case, higher volatility should help your system. If higher volatility is presenting a problem then I suggest you look at the individual components that compose the QQQQ and test out your model on those individual components. The results from the components should mimic those from the composite - if they don't then find out why. Mike
unfortunately, that and a buck will still not get you a good latte. An edge is only as good as its next trade. The future matters, backtesting is not a real edge. Working well in the past guarantees little, especially as the markets are constantly shifting their market geometry
Not working in the past guarantees something as well. Maybe I'm missing something - but - I've found price behavior to be fairly consistent despite "geometry". Markets are either doing 1 of 2 things - reverting to prior value, or finding new value. If by geometry you are referring to patterns, then you've lost me. Patterns only exist where we want to see them. Mike P.S. Do you have a public track record/credentials posted anywhere? I'm not trying to be rude, I just wan't to get a sense of your knowledge/experience base.
20 years trading, worked at 2 investment banks. Market geometry means all the current behavior of a market: trending, choppy, volatility, etc. etc. This behavior is a continuous killer of perceived"edges" because it is continuously changing.