Favorite Links

Thursday, March 15, 2018

Designing My Trading System: Part 3: Trading Metrics Beliefs by G

Feature Article
Designing My Trading System:
Part 3: Trading Metrics Beliefs
by G

Editor’s Note: G (who requested anonymity) is a client with an engineering background
who has been working on developing trading systems for the last several years. He
details his observations and insights from that process in this five-part series of articles.

All trading metrics need a job. The specific jobs assigned to a particular metric stems from one or more of the following four purposes:

Measuring Financial Gains and Losses
Enabling Process Control
Enabling Process Improvement
Enabling Process Comparison

The first purpose is simple: how do your trading results compare to your trading business plan? The other three purposes are process-oriented and eventually determine your financial results.

Examples of process metrics are expectancy and SQN score. With only these two metrics, you have the beginning of a trading metric system that produces confidence in both the near-term and the long-term which builds the foundation for process control, process improvement, and process comparison.

Process control of a trading system isn’t complicated. For your selected time frame, document the market type, win-rate, average winner, average loser, expectancy, and SQN. Process control is all about knowing which factors matter the most in your trading system and paying attention to how changes in those factors affect financial results. Create only enough variables to represent the factors that really matter. Those variables should provide normal ranges of variation to allow exceptional values to standout. The effects of any process experiments on your system should be caught by at least one of your few chosen variables.

The final process purpose for trading metrics is being able to compare one system to another, using clearly defined and widely accepted criteria as the basis for comparison. Van’s metric of SQN was primarily developed to help traders answer a basic question on the degree to which position sizing strategies can be used by traders in reaching their financial objectives. The secondary use of SQN for comparing different systems is a Van Tharp Institute (VTI) application of the metric.

In an effort to compare my trading system performance to non-VTI systems (those that don’t measure expectancy or SQN scores), the metric “Profit Factor” (PF) turned up as the most common system performance metric used by trading system designers for comparing various systems.

PF is really simple: divide the gains from all your winning trades by the losses from all your losing trades for a given period of time. The resulting number usually ranges between 1 and 4, where 1 is ‘breakeven’ and 4 is considered very high and difficult to sustain.

Beyond comparing the performances of various systems, PF can be used for both process control and for process improvement.

The formula for PF can be re-written into a very useful alternate form as:

PF = (winner% / loser%) x (average R winner / average R loser)

With a 67% win-rate and a Reward-to-Risk Ratio (RRR) of 2, you have:

PF = (.67/.33) x 2 = ~ 4

These two ratio-metric factors, i.e. the ratio of win-rate to lose-rate, and the RRR, have very different ranges of possible values, ranges I found enlightening in thinking about process improvement possibilities.

The literature and on-line sources mention professional systems running win-rates as low as 30% and still making good profits. Likewise, some scalpers brag about running 90% winning trades. Given these stats, what’s the dynamic range for win-rates?

One the low end: (0.3/0.7) = 0.43, while on the high end: (0.9/0.1) = 9.0

Hence, the dynamic range for win-rates is: 9/0.43 = ~ 20

RRR ranges are tougher to determine. On the low end, excluding the Ultra-High-Frequency traders, the lowest I’ve heard for very-high-win-rate scalping is 0.2R, so we have (0.2R/1.0R) = 0.2 On the high end, I have measured daily RRR’s running above 20R using advanced position-sizing strategies.

Hence, the dynamic range for RRR is: 20/0.2 = 100

The difference in RRR and win-rate dynamic ranges is: 100/20 = 5x

So, if you wanted to dramatically improve your PF, there is 5 times more dynamic range available to you in working on increasing RRR’s than in trying to increase win-rates. Hence, this alternate form of the PF metric suggests where to allocate attention if the goal is to make major improvements in PF, and also provides an easy way to measure those improvements once implemented.

If you doubt the logic and math above, consider the three notable quotes from successful traders make the same point a different way.

“It’s not whether you are right or wrong that’s important, but how much money you make when you’re right and how much you lose when you’re wrong.”

George Soros

“It turns out that it is much easier to make money when you are wrong most of the time.”

Curtis Faith, author, Way of the Turtle

“Looking back, about one-third of our trades have been winners and about two-thirds losers. That’s been true for a long time.”

William Eckhardt, co-organizer of the Turtles experiment

If you’re still not buying the math or the quotes above, don’t worry … your skepticism shows you are human. Evolution has wired us to be intrinsically risk averse, perhaps an inherited artifact from our ancestors who had to survive the scarcity and uncertainty of food and water for untold millennia. Even worse, our successes in school and in sports ingrain the premiums placed by our modern world on scoring highly on tests and winning at everything we do. Rational thought and the logic of math doesn’t stand a chance when pitted against our drive to fulfill our emotional needs to “be right” and to win. Let’s face it: winning feels good, and RRR is an abstract concept … there’s simply no contest! Most traders can’t lose on most of their trades and still feel good about trading … our wiring and upbringing rule us.

In 2005, trading legend Chuck LeBeau published a thoughtful article in Van’s newsletter: BENEFITS OF SYSTEMS WITH A HIGH WINNING PERCENTAGE, in which he outlines several psychological reasons why traders should prefer high win-rate systems.

My final take on trading metrics: what matters the most in a trading is not a system’s statistics but the system’s fit to the trader. “Useful” metrics derive from a good fit between trader and system … it’s just that simple.

In the upcoming article, Part 4 will describe several psychological lessons I learned that materially improved my trading and my life.

– G

No comments:

Post a Comment