Duplicating results in Way of the Turtle

[LeviF]

I duplicated the Bollinger Breakout results that are described in the WotT book. The position sizing he used were 0.5% equity / ATR of a given market. This has nothing to do with the stops however, because positions are exited at the 350 day SMA. The actual risk taken on typically varies from 2% to 5%, as much as 10% of equity. Isn't this a bit excessive?

[sluggo]

The only person who can tell you what the author had in mind when he wrote that section, is the author himself. Unfortunately he hasn't posted here in quite a while. He appears to be focusing on other things, and so he might not answer your question for a long time.

[Tim Arnold]

Actually, the question

Isn't this a bit excessive?

can only be answered by each trader for him/her self

[LeviF]

Than can I ask what % of their account most traders here risk on any one trade?

[Roger Rines]

Hi "levijean",

Than can I ask what % of their account most traders here risk on any one trade?

The reality of this question is best answered by Tim's answer:

Actually, the question [snip] can only be answered by each trader for him/her self

In our case we find our answer from our system's performance results, and through “what-ifâ€

[LeviF]

Roger, i'm curious:

In your trading, how have your realized results compared with your system backtests?

[Roger Rines]

In your trading, how have your realized results compared with your system backtests?

Yes. We track each system's net equity change daily and reset it at the start of each year. This is a process we use to measure execution performance. In our case we have been using a full service broker (MF GLobal was EDF Man) and he has been able to keep our market results on most trades very close to what the system is reporting. In more than a few cases he had been able to beat the performance of the systems by how he was handling the size, but I've noticed with the electronic side being dominant now, the number of times he can beat the system's result has been less frequent. Still, he most often finds a way to pay for his effort during roll-forwards and that is hard to dislike.

What made you ask this question?

[LeviF]

What i meant was, how has your future performance differed from your expectations from backtesting? Returns, drawdown % and length, etc.

[Roger Rines]

Our current major systems began trading in 1994 and their results in the market turned out to be a little better than what we had seen in testing before going live. That improved performance continued on up to around year 2000 before dropping. During the period of 2000 to the fall of 2003, results flattened out a lot as our average trade durations shortened. In late 2003 the results took off like a rocket and brought in the best year we had in 20-years of trading. In 2004 to mid-2007 results began to look like early this century performances with the equity swings showing increasing larger deviations.

In looking back at that entire period of results, it is fair to say that on over average results didn’t fall outside of what was possible, but recent performance didn’t show the same kind of gains we saw during system testing and then achieved for the next 6-years afterwards. In looking forward I don’t think we’ll get back to our earlier performance results unless the trendiness improves, or we get smarter with how our systems adapt. In summary I don’t see it as reasonable to bet that the markets will get better, so we’ve begun to implement some adjustment features that will improved how the system deals with the shorter trends.

If there is a message in any of this it would be that nothing in system testing results promises anything no matter who does the development work, or how they worked in the past. My experience has shown the best we can hope for is to be sure the pieces being used in our systems are sound and provides some form of advantage when assembled into an approach that fits our personality.

If you can get that far with your systems, you’ll need to find a way to believe in your systems because of what they demonstrate from a lot of angles. You’ll also need to be honest about knowing and accepting what your system’s weaknesses are while you still can do something about them. What that means to you is only something you can decide. For certain if you don’t believe in the methods you employ for trading, you’ll give up or increase your account risk at exactly the wrong time.

[nodoodahs]

A technical, specific (and yet still strangely non-specific) question about your test vs. your actual results. I say specific and non-specific because it will give me (and perhaps, you) a very important insight into the results, but won't disclose anything sensitive.

Calculate your annual backtested returns for every eligible day in backtest. Over a hypothetical ten-year test, you would have nine eligible years worth of days or about 2300 days. I don't want to know that number.

From that sample, calculate your median, average, minimum, and maximum annual returns from test. Also calculate your standard deviation of returns, and your maximum drawdown in test. I don't want to know those numbers, either.

Now do the same calculations, on the same basis, for your actual results. I don't want to know those numbers.

I do want to know this number. Take your average actual return and calculate how many backtested standard deviations it is from your backtested average return. That's a number I'd like to know, at least in a broad sense.

Did any of the following actuals exceed the benchmarks set in test (yes or no)? Minimum annual return, maximum annual return, maximum drawdown?

Thanks!

[sluggo]

... Calculate your annual backtested returns for every eligible day ...

... Now do the same calculations, on the same basis, for your actual results. I don't want to know those numbers. ...

Bill, do you have daily account equity values for ten years' of your actual trading, saved in a single computer data file ready for quick processing? I would be very surprised if you did. I doubt that private traders would bother to perform this level of detailed bookkeeping and keep well-organized historical files of it in an easily accessible form.

Which means it just won't be possible to perform the backflips you've requested. But heck, I might be wrong, there might be seasoned traders who are also Intuit Quicken fanatics, who've been doing this very thing since the dawn of Windows 95. I wouldn't bet on it, though.

[nodoodahs]

Eh, I'm not trying to be difficult. Or make you do backflips.

The question boils down to: do you have any STATISTICAL IDEA how your actuals compare to the statistics you developed in test? With the majority of the post explicated one method of developing that.

Daily? Weekly? Monthly? Quarterly? Annually? I'm sure you have a test file to do at least one of those against.

No backflips, please, but I'm also reasonably sure you have a CAGR since inception and a maximum drawdown number since inception, readily available. That would be a good indication of whether the system matched the test.

I don't CARE what your returns are. I AM interested in whether your test experience set a good boundary for your actual experience, if you don't mind sharing that.

[nodoodahs]

I just realized that sluggo responded to a question for roger.

FYI and FWIW, I have weekly returns from 77 weeks back and end of month since inception, and monthly from my start of trading, although I spent a lot of that time as either discretionary or as working on a system, settling on the ones I'm trading just a few months ago. For tests, I have either monthly or daily returns, depending on the test, for the systems I'm trading now.

I'm small fry, don't use Quicken or Intuit, it's just on a spreadsheet.

I fully expect that anyone managing money professionally and using a system like Blox has detailed test data, and any CTA will have very detailed actual data. Maybe not daily, but weekly or monthly, and have it handy.

I find it strange, sluggo, that you seemed to take such great offence at a question that (1) wasn't for you and (2) was aimed at providing statistical analysis of system vs. test performance, which is something I think every mechanical systems trader would be desirous of obtaining.

[sluggo]

Funny. No offence taken, just surprise. I will stand aside so you can beseech Roger, unimpeded. Good wishes towards your project.