I would like to ask you two questions on your opinion of my thoughts on developing a trading strategy. I have two ways I am going to try to develop a strategy and would like to know your opinion of which one is best.
1) Is this the best way to develop a strategy? I want to build a simple system. I want it to build the trigger that gets me in at the best time. My first thought is to find two triggers that either say to be in the market long, short, or flat. Then by combining the two I will come up with far fewer trades that have a higher success rate. It is my idea to test a couple of basic strategies this way and see if they improve. For the sake of argument let’s assume that I can identify 6 zones for trading with each basic idea I am testing with. For example, I might decide that the stochastics are to be used and tested for the 6 conditions of rising or falling above or below 0 or being in the oversold and overbought zone. That is each wave would have two rising pieces, a peak area, two falling parts and a trough. ***If*** I found the best time to buy was in the rising between oversold and 0 section do you think that could be assumed to be the best one to combine with another method such as using MACD somehow or might another section when combined be the best bet. That is overall the best method might be to buy when coming out of oversold, but when combined with MACD in some way the best method might be to buy while in the oversold zone before the breakout. Obviously there will be exceptions, but would the best way to trade a signal on its own be also the best one to add to another signal or might there be some goofiness of combining them that makes a completely new test required, that is must I test all 30 different ways to find the best or is it almost always going to just be the combination of the best two when tested individually. The reason I ask is that it would be much easier to test everything individually and then combine them, but it might not work very well if it doesn't give the most likely best solution.
2) Or, is the best way? Use a single simple method and then try and determine if the market conditions are right. It is easy to look back on a chart and say it was rising during this time, so the buy-a-trend strategy will work best. And here it is range bound so the rubber band trade system that rides extremes back to the center are going to work best. But if a decent system is developed for those given conditions, ideally one that breaks even in other conditions, then the step two of determining what condition is prevalent is the key. According to a book I read by Tharp, something with Financial Freedom in the title, he says that there are 6 conditions, trending up, trending down, and range bound, both volatile and non-volatile. It is easy enough to mark off these zones in the past, but how do you tell what day a new one starts. I can historically find a time, a long period of time, of any of these conditions, but measuring them in the moment is like adding a new indicator that is sometimes wrong and makes the system take the wrong type of trades when it is faked out. Another way this type of system is often defined is to trade your system but only take trades in the direction of the long term trend. The idea being that by looking at a long term chart it is easy to tell which way to go. I would suggest trying to put it in an algorithm isn't so easy.
I would love to get your thoughts on which system building idea is the best approach.
Comments
I think a great trading system is developed through discipline. Finding one inicator is good, but I do think it is best when combining multiple indicators as mentioned in strategy 2. As technical analyisis is the voice of fundamentals, (ie: the fundamental news of a company is expressed through the charts candle) looking at various indicators for conformation can be helpful. Timeframes as well.
My system uses, Price Trend and Volume along with a cross of Moving Averages (whether MACD, Stochastic, 10/20 day it's up to you) and a Bollinger Band to help identify extension levels, through in a few FIBO retracements and you have covered a lot of ground technically, that is being expressed fundamentally. Refer to some of my comments on "3 golden rules of trading" and "buy with higher high from previous day sell for lower low from previous day."
Commet back and I'll write again. Thanks!
The point of strategy 1 is to determine the type of market and turn off the system when it is in a bad market. For example, if you build a trend following system and then turn off the system when you somehow measure that the market is more likely to be stuck in a range. I like this idea, but feel that trying to determine what type of market is going to be is just as difficult as getting a good strategy in the first place. I guess doing it this way I would think you go elsewhere to other data or timeframes to help determine the type of market. I can find the direction of the market on a 5-minute chart easily when its a simple thing to see it on the daily chart.
The other idea, #2, is what I have been working on and am having a heck of a time getting a strategy that works for long periods of time. I am working in forex so I have no volume.
You mention that you use price trend in your system. If you would be willing to expand on how you do that I would love to hear it.
Another problem I have been struggling with is how to get price targets. I am trying to build the strategy into tradestation so I can backtest it which is of course some of my problem having to define it so precisely that a computer can figure it out. One of the worst things for a system performance as far as I can tell is to correctly pick the direction and a move and then only get a small percentage of it. I am not talking about selling at the maximum or anything, but if the system correctly enters a move and then has a target of 25 pips and is out but the actual move was 100 pips then the strategy is suffering because of the target.
I would love to hear any comments about this idea that I am trying: I am looking at momentum to find the peaks and make decisions at these points. I am averaging it out a bit so it is smoother and then when it gets over a certain level and starts to drop I can be pretty sure it is going to head back down. Instead of just looking at the value of the momentum which seems to not help me, I am trying to find the price peak associated with the momentum peak and keep track of several. So far I don't have anything that works over long periods of time because of the levels that work best seem to change. I am working on making them adaptive with great difficulty.
I also look at the time it took to form a particular rally and pullback, and allow the same amount of time for the next rally. (ie: stock rallies for 10 days, pulls back 4, my entry on that 4th would be a ten day timeframe to make it's next move higher).
As for price targets, I scale out, selling a portion at the previous high of a stock (because I bought the pullback remember), and then usualy a 127.1 Fibonacci extension as the next price target (again RIG back a few months put in a large hammer and ran right up to the 127.1 Fibo extension).
Another way to asses a price target is to take the difference between the high of the pattern and the low of the pattern and add it to the high. So a stock hits $50, pulls back to $45 and starts to turn higher, I would buy as close to $45 (with conformation the stock is now moving higher) and sell a portion at $50 leaving the rest to run to a price target of Fibo Extension 127.1 (which may be $53) and then an my $55 target last.
Buying into a security is the easy part, knowing when to sell is what takes skill I've found. If you sell a portion whether it be half or 1/4 or whatever on a big up day in the stock or at a previous high or just selling into strength, you limit your risk to the downside. I always net positive using this scale out strategy over a statistical set of trades. You can never pick EXACT tops and BOTTOMS, my as well take some off the table when you have a profit.