This is not how this system tests and it is not giving you results for symbols. That is the point. The system is testing the event not the stocks and it is the only one that I know of that does it. You have no idea what the system is going to trade tomorrow. The notion of a system working for 2 years is absurd to say the least. The market is way too efficient. With the OM you get put in the right direction and then continue to refine as you trade as long as real results are similar to what you have been testing. You are talking about a cybertrader or tradestation type of backtest and I agree with you on that. What I am doing is not looking for sma cross and finding the stock that it works well with. I am looking for is an event that has been working across a broad range of stocks in the last 3 weeks. And its not easy.
My test results are never the same as my tested results. Sure, quite a few of them are within an expected result... but never the same. Similar vs. Same.
My system's actual performance equals backtested performance, but that's because I tested with 2 years of tick level data (avg. 25,000 price level changes per day, including off hours). The reason for testing with such a large data set is that as we all know the market changes over time. What works for one 3-month period may perform horribly in the next one. 2 years includes a bull market and a bear market, and periods of high and low volatility, pre- and post-credit crunch, etc. I got my data sets at opentick.com. The quality of the data isn't perfect, but it works for my purposes.
I get better results in actual trading vs backtesting as I avoid obvious risky trades (visually) ie..at resistance, overall market condition and direction, risk/reward etc... There are certain things you just can't code into a mechanical system effectively.
Then you have an edge on us system traders. The brain is the best trading machine going, if well trained.
I am a systems trader. I don't take trades outside my system. I just don't take all of the buy signals for the reasons I explained. Maybe take a look at your system in detail and analyse the losing and winning trades, you may find some clues there to help you identify which ones will be a winner, or at least improve your win rate and reduce drawdowns.
Totally agree... "one severe theoretical limitation [of backtesting] arises from the fact that when a trading system is back tested it is not a real market participant, and thus its effect on market prices is not reflected in the results."
Actually your system has little or no affect on market prices unless you are trading very big size relative to the volume.
1. Little or big... the affect exists. 2. Considering that all system traders work under a same universe of past data, meaning the basis of everyone's decision is the same. There's going to be an obvious correlation (synchronicity) between traders and that little affect can lead up to a big one.
Do you have an idea of what big size relative to volume means? I ask because I want to trade some $3 stocks. I'm wondering what % of the daily volume I can trade. 2? 5? I like to use market on open orders because I'm often at work during trading hours, so in that case maybe I could trade even less. Some stocks open on small volume and I'm afraid if I put in 5% of the volume it'll drive the price up.