Kenneth Tropin, Chairman of Graham Capital, offered food for thought at a presentation every trader should absorb:

But most long-term trend following systems traders are always doing something that is very, very difficult to do discretionarily — which is, waiting for a market to make its new highest price ever and then get in it. It’s never this feeling of: Oh, I really bought this right! You always have a feeling on every trade of: I really bought this wrong! After 20-something years of doing this, that is a profitable thing to do, believe it or not. But nonetheless, systems traders are clearly not front-running market opportunities. They are following them; and that’s why they’re called trend-followers… Trend following systems are in general trying to do the following: they are trying to determine, in all of the market and macro events that happen every day, which of this price behavior is really a trend or is an emerging trend that’s happening, and which of what’s going on here is really noise. The art and science of a system is to be able to stay out of the noise and get involved in the right trends, hold those positions in the right trends until they ultimately fail. And every trend ever has always failed. And be able to make sure that when the trend fails, you have taken away some profit. By definition, trend followers never take peak profits, right? Because you’re always waiting for that market to tell you the trend is over. So of course you’re always giving back some unrealized gains… But nonetheless, there’s no question that you want to invest in trend following with a system. That system, and I’ll give you a very simple one, is: we have a model that we’ve designed and we can share with people who are interested…It’s a trading model that we applied to the Zurich index of long-term trend following. The basic model is so simple. You take your portfolio, you put it in equal thirds. I have $100 million to invest — one pot is $33 million and so on. Month one: I’m only looking at the $33 million and I’ll look back at the Zurich index and I want to see what the Zurich index has gone over the last 3 months. If that Zurich index is profitable by more than 2%, I don’t invest. I look back the next month — wait 30 days. Now let’s say that Zurich index is down more than 2% on the trailing 90 days, that’s when I add my capital. Once I get fully invested, if you always use this approach of looking back 90 days, and increasing your allocation when the trailing 90 day performance is negative, and slightly decreasing your allocation when the trailing 90 day performance is positive, you will improve — if you just look at the Zurich index — the nature of the return of that index, the nature of the return of that index by about 15-20% on a risk-adjusted basis. Real simple. Takes away all the emotion…

At a presentation he was asked, “Now, a computer has beaten the chess master, so we all have an appreciation of how much a computer can do. So my question is briefly — how much do each of you delegate to your computers and how much do you use your judgment?”

That’s a great question, and it’s one of the most widely misunderstood issues in the business. There are some important things that we use a computer to do, and there’s some important things that we don’t expect it to do. What we use a computer for is to test ideas. When we decide, as human beings — and human beings create those ideas — and we decided that the computer’s output of testing those ideas makes sense and it has to be fashioned into a trading program or system, we can then use that computer to automate the execution and administration of how it handles markets. But one of the things that is a sure and absolute guaranteed road to failure is to ask a computer to design a trading system. They don’t do it at all. They’ll give you such data garbage that you can’t possibly ever invest in something that was designed by a computer and make money. So all of your ideas have to be based on market experience of a human being who has come up with a way that I might be able to take advantage of something I’ve seen in the markets, develop a model around those ideas, then ask the computer to test it for me. Then if it tests well and you want to use it, have the computer administrate it [or automate it].

What Tropin Gets Right About Systematic Trading

Tropin’s “I always feel like I bought this wrong” observation is the most honest available description of what buying at new all-time highs feels like from inside the trade. Every breakout entry at a new highest price is uncomfortable. The price is higher than it has ever been. The purchase feels expensive. The instinct is to wait for a pullback. This discomfort is universal, persistent across 20-plus years of profitable trend following, and completely unrelated to whether the trade is correct. The profitability of the approach has nothing to do with whether each entry feels right.

The noise vs trend distinction is the core problem that every systematic approach is attempting to solve. Markets produce price movements every day. Most of them are noise: short-duration, reversing, unrelated to the sustained directional moves that trend following captures. The system’s value is in distinguishing sustained moves from noise with enough accuracy to produce positive expected value across a large sample of trades. Systems that confuse noise for signal enter and exit too frequently, accumulating transaction costs without capturing trend duration.

The Zurich index model Tropin describes is a practical implementation of the behavioral insight that trend following performs best when entered after a drawdown period rather than at peak performance. Buying after a trailing 90-day loss and reducing allocation after a trailing 90-day gain produces 15-20% improvement in risk-adjusted returns on the same underlying index. This is the behavioral finance principle applied to allocation timing: investors who chase recent performance enter at peak allocations and experience subsequent drawdowns fully; investors who add during drawdowns participate in subsequent recoveries more fully.

The computer question answer is the clearest available statement of the correct human/computer division of labor in systematic trading. The computer tests ideas. Human market experience generates ideas. Asking a computer to generate ideas produces data-fitted output with no market logic. Eckhardt’s 12-degrees-of-freedom limit and 1,800-trade minimum sample are the specific quantitative constraints on the same problem from the testing side. The human provides the market logic. The computer provides the testing power.

Frequently Asked Questions

Why does buying at new all-time highs always feel wrong even to experienced trend followers?

Because the purchase price is the highest the market has ever reached, which feels expensive relative to all prior prices. The instinct toward value investing, buying at a discount from historical range, is exactly opposite to what breakout entries require. Tropin’s observation that after 20-plus years this feeling persists but the approach remains profitable is the empirical answer: the feeling of buying wrong is irrelevant to whether the trade is correct.

What is the Zurich index portfolio model Tropin describes?

A three-tranche allocation approach to managed futures. The portfolio is divided into equal thirds, each deployed when the Zurich long-term trend following index has a trailing 90-day return worse than negative 2%, and reduced when trailing 90-day performance exceeds positive 2%. This counter-cyclical approach improves risk-adjusted returns by 15-20% compared to a static allocation by entering during drawdown periods rather than chasing peak performance.

Why can’t computers design trading systems?

Because computers optimize for statistical fit to historical data without market logic to constrain the optimization. A computer-designed system maximizes in-sample performance by finding parameter combinations that explain past price behavior, which is over-fitting by definition. The result has no conceptual basis for working in the future and no robustness against new market conditions. Human market experience provides the conceptual framework. The computer tests whether that framework is supported by the data.

Trend Following Systems
Want to learn more and start trading trend following systems? Start here.