Jack Schwager談系統開發

Janette: So, where would a person start?

Jack: First of all, you can’t just sit down and say I’m going to be a great technical or systematic trader and start designing strategies. You’ve got to watch markets in real-time and paper trade. You’ve got to get some real life experience to get some feel. From your experience, you draw ideas of what you think might work, and then take the ideas and systematize them in some way and test to see if that assumption bears out. It’s not a matter of taking someone else’s system and trading it. If you do that, you won’t have any confidence or feel for it. Every strategy is going to go bad at certain points, and when that happens you’ll end up bailing because you don’t have any reason to have confidence in it; its not yours, you haven’t done the analysis, and it doesn’t reflect anything that you’ve done. So I think it’s important for each person to develop his own methodology.

Janette: In your video series, Jack Schwager’s Complete Guide to Designing and Testing Trading Systems”, you talk about some of the pitfalls, specifically the “Super Razzle Dazzle”. What are you trying to get across?

Jack: What I was trying to say was, whether in articles about systems, ads for systems, or people giving presentations about systems, you’ll invariably find these examples of the strategy with phenomenal trading signals. It looks just great. The reason it looks great is because the example was chosen with the benefit of hindsight. Every conceivable strategy can be made to look great if you choose the right example.

Janette: Then what should a trader be looking for?

Jack: Well, it’s not the performance measures. It’s testing the system correctly—no matter what performance measures you’re using. Testing a strategy correctly means you’re not using hindsight—plain and simple. If you’re using hindsight, you’re doing it wrong. The question then becomes, “How do you do test a system it without hindsight?” There are two ways to do it. First, you can take a small amount of data, say one or two markets out of 30, or short segments of time for all markets and you use that data to develop the rules of your system. Then use those rules and test them on the unseen data.

The other way is a walk-forward simulation. You place yourself back in time. Say it’s 1985. You use data prior to 1985 to develop the system rules, and then test the system for 1985. You then repeat the process with data prior to 1986 and test 1986, and so on. The point is you use data up to a certain point of time and test on a point of time forward from that point and walk forward through time. Both of the approaches I’ve described are valid because they don’t use hindsight. The single most important thing about testing trading strategies is to avoid hindsight—if you don’t, then whatever you get is totally meaningless.

Janette: Okay, so once you’ve done the ‘walk-forward simulation’ or ‘small data sampling’ and you’ve got all this data back, how do you evaluate it?

Jack: You want to look at return to risk. This is particularly true of the futures markets. Return by itself is totally meaningless. Here’s an example: Let’s take a strategy that returns 20% a year and has periodic drawdowns of 40%, and another strategy that only makes 10% a year but its worst drawdowns were 5%. Well, you might look at the system with the 20% return and think,” well, it’s twice as good.” But, what you could do is take the 10% system and leverage it 4:1 and end up with a 40% return with a 20% worst drawdown. Thus the system with a 10% return (but only a 5% drawdown) is actually a tremendously superior strategy. Return by itself has no meaning because you can take any strategy and just by leverage, change its return. So the only thing that’s valid is return to risk.

Janette: That makes a lot of sense.

Jack: And the reason I said it was particularly true of futures is because there’s always excess capital (no reasonable person trades at full margin) and there’s no cost to borrow, so it becomes totally transformable and interchangeable. In that case it becomes an exact ratio—if the return to risk of Strategy A is 2:1, then its then it’s superior to Strategy B if Strategy B’s return to risk is anything less than 2:1, no matter what it’s return is, because you can transform the lower return system to a higher return through leverage.

Janette: If we want to oversimplify the steps of constructing a trading strategy, how would you define the steps?

Jack:

1. Come up with the ideas through market observation and research.

2. Develop a systematic set of rules based on these ideas.

3. Test the system without hindsight.

4. Implement the system. This is the simplest part. You just follow the signals.

Implementation is nothing-- it’s an afterthought. You just have to be consistent with what you develop. If you believe in it, then you will trade it exactly as it is.

Sometimes people will say “Yes, but…I know my strategy says I should be long, but you have to understand that this is a particular case, which is an exception for this and this reason and it’s not in my system”. Well, if it’s not in your system, make it part of the system. Then go back and without hindsight test to see whether it indeed helps the system. If it does, then incorporate the exception as part of your system. Then you have no reason to intervene and change what the strategy is telling you until the next time you think you have a reason, and when that happens, repeat the process. So, the simple answer is if anyone ever comes up with a reason for why they shouldn’t follow their strategy, then they should incorporate that exception as part of the strategy. If you find yourself second-guessing the system, it means you don’t truly have confidence in it, and you shouldn’t be trading it.

 4/5  首頁 上一頁 2 3 4 5 下一頁 尾頁