I recently developed a strategy that seemingly works quite well when tested on EOD data for the past 500 days. Unfortunately, the results are less impressive when viewed against longer time frames. Obviously markets and the nature of markets change over time. No two markets will be the same, though arguably some will share enough similarities to be of some use in backtesting. So here are my questions: 1. What past markets (general equities, dates please) behave in a comparable manner to todays markets. Please don't say 1974... 2. How can we identify / quantify changes in market types? Anyone?
As strange as this may seem, the SPX is correlating very closely to the extended move off of the lows in October, 1998. If you can do a simple "raw" overlay between the two time periods, you will see this quite clearly...
oh man, i knew it! i knew the last couple years were just a temporary pullback! vulture, i sure hope you're right man! october 98 again! oh man, how i wish you are right!! (cos we all know what happend after 10.98 - gentelman, hold onto your hats!)
What period are we in? A confusing one. To really understand the stock market, you also have to look at the other markets that influence equity pricing, i.e. the securities market and the currency markets. Big money moves the markets, and there is more money in bonds than in stocks. We are at historic low levels for bond yields. Normally, low levels for bond yields are seen at the tops of equities, not at the bottom. Are we really going to see money flowing out of the bond market into the stock market like we did in the past, to fuel the movement upwards? The major bull market that began in 1983 in stocks was fueled by declining yields in bonds, thus making stocks yields more attractive. We are now going to have increasing yields, as the budget deficit grows, so yields on stocks have to keep pace with bonds. A tough chore, given the inherit weakness in the dollar based on trade deficits, budget deficits, and low interest rates. A period of the 60's, where we congested for a decade, with a downward bias to digest the gains of the 17 year bull market makes sense to me. The money has to come from somewhere, doesn't it?
Thanks for the insight 777. Personally, I think you have very nicely explained the current and future picture as it is and will be be, respectively. A beneficial difference for traders between this period and the one you reference is the level at which equities are. Higher overall prices allow for greater profit opportunity day to day. A decade-long bear market beginning with the Dow at 11500 will afford traders continued opportunity for livelihood than one at the levels in the 60's. Of course the availability of point-and-click trading, direct access, etc etc helps the cause as well.
I think you can do some kind of intermarket backward testing, in order to define market stages. I suggest you use neural networks to evaluate current x past market stages using intermarket indexes in stocks, bonds, currencies, commodities, economic indicators, etc... You can start with simple indicators (like moving averages) and then move to something more complex.
The number of data points is less important relative to what the data set(s) actually respresent. It's important that all extremeities are represented (bull/bear/sideways markets). In order to establish some sort of statistical evidence for your system, you must be much more rigorous in your testing approach. If the system is robust and designed to work long term, its performance numbers should hold steady even when factoring in some random data points and random slippage. And this is just the start. This is not meant to be captious, but by subjecting your system to datasets that are only representative of today's market dynamics, you're engaging in curve fitting. The system is not going to survive for the long term unless it is designed in a manner where it can be profitable under all market dynamics. Take a look at the following book : Trading Systems that Work. If you utilize these statistical methods when testing your system, you'll be well under way to creating a robust system.