It is getting interesting the discussion about tape reading x mechanized trading systems.
First, a definition: what I call tape reading isn’t only the observation of the “tape” (times and sales, bid/ask, etc) – it goes way beyond that – take into account intermarket relationships (indexes x currencies x gold x bonds x oil x etc), type of day, ”state” of the market (type of moment), day’s theme (proprietary study – perhaps not quite what you would understand by it’s name “day’s theme”), price information (not T.A. “formations”), volume information, a proprietary ”Sentiment” Index, volatility, time, trading psychology, news impact, and so on.
All these informations, all at the same time, crossing through your trading screen (I use two 22″ screens for brazilian market and my macbook for world market and home broker).
Jeff once said about the experience in the pit: “If a buy order came into the market, (and one could detect whether it was a buy or sell order by looking at the movement of the brokers eyes)”… Well, we don’t have how to do it behind a screen… So we try to get this “feel” of market action never forgetting these principles:
- Ever-Changing Cycles (or “regime shift”) – The markets are always changing, always evolving, always showing something new, just to show “classic themes” again). Although I’m working on parameterization, I find it very difficult to accept the idea that a “artificial intelligence” program will efficiently “read” this changes until it’s too late.
- Deception - Along with the principle above, one of the preferences of the market is to get people on the wrong side of expectation – and expectation is based on what “worked before”. In these terms, the situation is the worst when it appears the safest.
- Discontinuation – Most systems search “continuously” for the patterns to fit. My own experience tells me the market themes are not continuous. (This is, for instance, what kills many T.A. approaches).
In these terms, I believe one could implement mechanized trading systems on a daily basis, but after a human input of recognized market conditions.
It would be something like: “Ok. Today is recognized as “Day Theme # 1″, so the best solution is to run Program “x”.
Until someone is able to program a trading system with so many nuances.
Newton, that is correct. You cannot program experience into a trading system. However, you can gain experience in programming systems, which will reflect in your bottom line.
As I see it, there is no right or wrong way. It only depends on your psychological profile and your thinking patterns.
If you trade based on your feeling of the market, then discretionary trading is probably the way to go. If you can explain everything you think of in precise terms, then you can probably build a program that will trade in your place.
Dan,
One of the challenges of program trading when you have lots of things to parameterize (as in day-trading strategies) is WEIGHT. We can model all the influences of the market, but I realize that from time to time (sometimes even intraday) the market I trade gives more or less importance or weight to this or that influence. And, to read this weight shift is also tape reading.
I believe automated trading works better in multi-day strategies or in very ultra-high-frequency. But in regular day-trading, with all these influences, I find it difficult.
Newton, all systems face this problem of optimization. There are some powerful tools and algorithms that can help you, such as neural networks or the more recent support vector machines. These are very good at finding and learning patterns, but the real question is which patterns are relevant for future trading.
I don’t think there is a substitute for common sense in regard to this. You need to come in and think. If event X was followed by an uptrend, does it mean that event X has a positive influence on the market? There’s no simple solution to this question.
It’s also pretty ironic. It proves that any form of mechanical trading is actually very discretionary