Industry Benchmark Performance
Equities gains while futures lost in November, the opposite of the year-to-date returns where futures are well ahead of equities. Even with occasional rallies, equities have been a difficult place to show positive returns. Futures have benefited from selling short equities, buying the U.S. dollar, and trading energy on both sides.
Kaufman’sMost Popular Books (available on Amazon)
Trading Systems and Methods, 6th Edition. The complete guide to trading systems, with more than 250 programs and spreadsheets. The most important book for a system developer.
Kaufman Constructs Trading Systems. A step-by-step manual on how to develop, test, and trade an algorithmic system.
Learn To Trade. Written for both serious beginners and practiced traders, this book includes chart formations, trends, indicators, trading rules, risk, and portfolio management. You can find it in color on Amazon.
You can also find these books on our website, www.kaufmansignals.com.
Blogs and Recent Publications
Find other recent publications and seminars at the end of this report. We post new interviews and reference new articles each month, while the Timing program held flat.
November Performance in Brief
The Equity Trend and Equity Divergence programs finished higher, thanks to a rally at the end of the month. Although futures showed small losses, it did remarkably well given the alternating euphoria and disappointment in the interest rates, which also drives FX and ultimately the equity index markets.
Although it is unlikely that we’ll see a sustained rally in equities, the Fed’s new announcement may mean that we won’t see a new low.
The DOW, S&P, and Nasdaq all finished higher by 5% in November, dragging the S&P out of the “bear market” (defined as down 20%). The DOW is clearly the strongest, a suggestion that investors favor less risky high-cap stocks during hard times, with small caps (IWM) the least favorite. Except for the DOW, the other index markets still show a downtrend. It seems likely that we will see a bottom here, but it is still early to expect any sustained move higher.
CLOSE-UP: Validation, Not Optimization
When we decide on a new strategy, we want to know the chances that it will work in the future. Instead, many analysts look for the best parameters that worked in the past. I think that is the wrong approach.
Whether we want to admit it or not, optimization is an important part of system development. But it is the interpretation of the results, more than the process, that is most important. The process is important because some methods require a lot of data and still do not give you enough information to make the right decision.
Optimization
Optimization is the process of finding the best parameters for your strategy over a particular time period. But most analysts agree that the best parameters in the past are not likely to give you the best results going forward.
The reason is simple: the method has fine-tuned the parameters to specific price moves. It may have found the highest profit from a price shock that is not likely to happen in the same way next time.
To avoid that problem, theory says that you choose a test periods that has a variety of market action, bull and bear markets, high and low volatility, and a few prices shocks. The more data you use, the less chance of overfitting and the lower the results. On the other hand, if you choose a short test period that you think represents the current market, you are sure to be proved wrong.
But optimization is very clever. It can look past “robustness” and still find that combination of parameters that avoided the downturn and profited from a few upward price shocks.
I would like to use the same testing method to show that a system is robust. That is, it will generate profits and limit risk for many combinations of parameters.
Optimization Methods
There was a recent article in Technical Analysis of Stocks & Commodities, “Boost Your Strategy Development,” by Rene Koch, that summarized some optimization methods:
- Random search, picking parameters randomly. Requires many iterations.
- Hill climbing, you follow one parameter to its best value then move laterally to the next parameter and find the highest value. It can end up with a local maximum rather than the best value.
- Shrinking Window, using randomized method that hones in on the best. Sounds something like a genetic algorithm.
- Bayesian optimization uses prioritized parameters in a sequential search. It is best for small dimensions with continuous functions.
- Grid search, the beat-it-to-death approach of trying every combination. This is mostly what we use!
- Particle swarm, another method with similarity to a genetic algorithm, focusing on areas of good results. Can often find a local maximum.
- Sequential model-based optimization (SMAC), uses regression to explain parameter selection.
An impressive article, but personally, I don’t think we need anything this sophisticated.
They All Have Flaws, So Choose the Simplest
As sophisticated as they are, these methods often find a local maximum rather than the best result. So the grid search is going to be best. If you have a lot of tests, then spacing the values out to reduce the number of tests will give you a good idea of the solution. Then you can narrow it down and retest.
I had created a classic genetic algorithm solution and published it in detail in my book, Trading Systems and Methods, 5th edition. I removed it in the 6th edition because it seemed too much for readers. However, the results were excellent. But it too could find a local maximum. To avoid that, I “seeded” the test with random numbers and ran the test 10 times. If I got the same solution, or nearly the same, then I was satisfied that I had found the best.
But be careful! If your test shows profits in only a small area, and you then focus on that area, you are ignoring the fact that the underlying strategy is not robust.
Searching for Robustness
A big problem with most optimizations is that they give you only one solution, the best. It’s up to you to then vary the parameters to find out how “sensitive” they are; that is, does the profit and risk change dramatically when the parameters only change a little? If so, the solution isn’t good.
You need to know the parameters that led up to the final choice, and the corresponding return and risk. The only optimization that records the results of each step is a simple grid search, where the results of each parameter combination are listed. If you sort the results best to worst, you can then see how the last 10 or 20 parameter combinations compare. If only a few parameters changed and the returns are similar, then you have a good solution. You don’t need a sensitivity test.
Even Better – The Average of All Tests
My own approach is different. Using a simple example, I test a moving average for a range of parameters that I consider “reasonable.” If I am a long-term trend follower, then I would test from 30 to 120 days. I use the average of all tests to judge the worth of the strategy.
Because I’ve traded for so many years, and tested markets over and over, I know that some markets are more trending that others. If you need to test that concept, apply my “efficiency ratio” to the markets. The higher the ratio, the more trending.
As an example, let’s look at the most trending market, Eurodollar futures. The least trending market will be EuroStoxx. In general, interest rates are the most trending. Of those, short-term rates trend more than long-term rates. FX (against the U.S. dollar) are the next best trending markets. Index markets have a long-term trend but are erratic in the short-term. They have periods of trending and non-trending, as do most other markets.
Chart 1. Ranking of futures markets by noise (appears in Technical Analysis of Stocks & Commodities, Jan 2023, “Matching the Markets to the System.”)
Let’s look at the best and worse in more detail. The Eurodollar test in Chart 2 shows that all moving averages from 10 to 150 days produce profits of nearly the same magnitude. I would call that robust.
Chart 2. Eurodollar test of moving average trends from 2000 (source: same as Chart 1).
In Chart 3 we see the same EuroStoxx test (long-only). The pattern shows no significant trend profits until we use a 90-day moving average. It is typical of equity index markets – noisy in the short term but an underlying upwards trend in the long-term.
Chart 3. EuroStoxx test showing trends favor the long-term (long-only).
These two examples show robustness in Eurodollars and less so in EuroStoxx. We learn that equity index markets have a long-term trend, which is also a valuable piece of information.
Multi-Parameter Testing Versus Building Blocks
Some analysts believe that you should test all parameters at the same time. That is, the choice of values for one parameter can affect the selection of other values and ultimately the total return. It is true. The most optimal set of parameters will come from a fully integrated test.
But I think that the problems offset the rewards. First, you get an extremely overfit solution. Even more important, when the system does not perform as well as expected, you have no idea which parameter(s) is(are) at fault.
Keeping with the idea of robustness, if a strategy has more than one parameter, I test each parameter separately, beginning with the most important (usually the trend calculation period), and moving down to the least important (whatever that is), perhaps profit-taking or volatility extremes.
In each test you want the choice of value to improve the average of all tests. You do not want to see a few values improve drastically while others are much worse. For those statistically minded, you can see that by applying kurtosis.
When each new parameter satisfies the criterion of robustness, then the final strategy is likely to be sound. If performance deteriorates, you can retest the individual parameters to see which has failed. You have control over the process.
Conclusion
Being able to test a range of parameters in your strategy is essential for success. But you are not looking for the best parameter. You are looking for the average of all tests, and a consistent pattern.
Once you have that, which parameter do you choose? You don’t know which will be best next month or year, but you know that most of them will be profitable. The answer is simple – choose at least 3 parameters across the span of tests, then trade them as three separate systems, each with 1/3 of your investment, netting out the signals. If two trends are up and one is down, you trade 1/3 of your investment on the long side. It also gives you the effect of scaling in and out of a trade.
A Standing Note on Short Sales
Note that the “All Signals” reports show short sales in stocks and ETFs, even though short positions are not executed in the equity portfolios. Our work over the years shows that downturns in the stock market are most often short-lived and it is difficult to capture with a longer-term trend. The upwards bias also works against shorter-term systems unless using futures, which allows leverage. Our decision has been to take only long positions in equities and control the risk by exiting many of the portfolios when there is extreme volatility and/or an indication of a severe downturn.
PORTFOLIO METHODOLOGY IN BRIEF
Both equity and futures programs use the same basic portfolio technology. They all exploit the persistence of performance, that is, they seek those markets with good long-term and short-term returns on the specific system, rank them, then choose the best, subject to liquidity, an existing current signal, with limitations on how many can be chosen from each sector. If there are not enough stocks or futures markets that satisfy all the conditions, then the portfolio holds fewer assets. In general, these portfolios are high beta, showing higher returns and higher risk, but have had a history of consistently outperforming the broad market index in all traditional measures.
PERFORMANCE BY GROUP
NOTE that the charts show below represent performance “tracking,” that is, the oldest results since are simulated but the returns from 2013 are the systematic daily performance added day by day. Any changes to the strategies do not affect the past performance, unless noted. The system assumes 100% investment and stocks are executed on the open, futures on the close of the trading day following the signals. From time to time we make logic changes to the strategies and show how the new model performs.
Groups DE1 and WE1: Daily and Weekly Trend Program for Stocks, including Sector Rotation, Income Focus, and DowHedge
The Trend program seeks long-term directional changes in markets and the portfolios choose stocks that have realized profitable performance over many years combined with good short-term returns. It will hold fewer stocks when they do not meet our condition and exit the entire portfolio when there is extreme risk or a likely drawdown.
We have been stuck in a sideways market for most of the year, with the Weekly Trend program looking a bit stronger than the Daily Trend. Small profits in the Daily (and the Weekly if the partial week was posted) keep these programs positive for the year. While we always prefer profits, these portfolios have done well controlling risk.
Income Focus and Sector Rotation
Some very volatile moves in high-yield interest rates resulted in gains for the month. Still, this program continues to fight with rising interest rates, favoring the short side. We expect a strong recovery early next year as the Fed’s interest rate policy leans towards slowing and finally ending. They it will be a mad dash to lower rates to avoid a recession.
Sector Rotation
A gain of nearly 3% has stopped the downtrend, but this program has the worst performance of our portfolios. Then again, it is just to show this classic strategy. To give it credit, it is not as bad as the Nasdaq returns this year. It has given back 50% of last year’s gains, which can be considered “normal.” We expect this to do better as the overall market stabilizes.
DowHedge Programs
A gain of nearly 4% for both the Daily and Weekly DowHedge parallels the strength of the DOW index compared to the other equity index markets. Investors tend to prefer large cap, reliable stocks when there is uncertainty.
Group DE2: Divergence Program for Stocks
The Divergence program looks for patterns where price and momentum diverge, then takes a position in anticipation of the pattern resolving itself in a predictable direction, often the way prices had moved before the period of uncertainty.
The nature of the market must be changing for the Divergence program to gain 8% to 9% in November. This program looks to buy during a pause in a trending move, which means that it sees an underlying trend (upwards). That should be good news.
Group DE3: Timing Program for Stocks
The Timing program is a relative-value arbitrage, taking advantage of undervalued stocks relative to its index. It first finds the index that correlates best with a stock, then waits for an oversold indicator within an upwards trend. It exits when the stock price normalizes relative to the index, or the trend turns down. These portfolios are long-only because the upwards bias in stocks and that they are most often used in retirement accounts.
Mixed results in November but this program is far ahead of our other equity programs and even farther ahead of the broad market indices. It holds very few positions at this time, looking for pullbacks within an upwards trend. Once we see a serious rally, this program should start to buy pullbacks.
Futures Programs
Groups DF1 and WF1: Daily and Weekly Trend Programs for Futures
Futures allow both high leverage and true diversification. The larger portfolios, such as $1million, are diversified into both commodities and world index and interest rate markets, in addition to foreign exchange. Its performance is not expected to track the U.S. stock market and is a hedge in every sense because it is uncorrelated. As the portfolio becomes more diversified its returns are more stable.
The leverage available in futures markets allows us to manage the risk in the portfolio, something not possible to the same degree with stocks. This portfolio targets 14% volatility. Investors interested in lower leverage can simply scale down all positions equally in proportion to their volatility preference. Note that these portfolios do not trade Asian futures, which we believe are more difficult for U.S. investors to execute.
Please read the report describing our revised portfolio allocation methodology. It can be found in the drop-down menu under “Articles.”
Small losses this month seem to be a good outcome as the key markets, interest rates, FX, and equity index – flop around with high volatility. This could be the formation of a bottom, yet more interest rate hikes will come. The only way to trade this is with a system and risk control, waiting for the big reversal.
Group DF2: Daily Divergence Portfolio for Futures
After a run of losses, the Divergence program gained nearly 5% in all portfolios. As with the equity program, this strategy buys pauses in an uptrend and sells pauses in downtrends. At the moment the uptrend seems to be winning.
Blogs and Recent Publications
Perry’s books are all available on Amazon or through our website, www.kaufmansignals.com.
Coming in January and February 2023
Another year! Technical Analysis of Stocks & Commodities will publish “Matching the Markets to the Strategy,” It is a look at why certain markets do best with specific strategies and why that can materially improve results.
Alejandro de Luis will publish Perry’s article “Living Off Profits,” translated into Spanish, in the January issue of his magazine, Hispatrading. It’s a good way to practice your Spanish!
Perry will give a webinar to Eduardo Lopez’ Robotrader students in Spain on Wednesday, February 8, noon (New York). I’m sure you can log in.
October 2022
An interview with Perry was featured in the October anniversary issue of Technical Analysis of Stocks & Commodities. The interviewer is his wife, Barbara Diamond, giving a different perspective on his career.
September 2022
“The Real Risk of System Trading” can be found in the September issue of Technical Analysis of Stocks & Commodities. It summarizes the many way we can measure risk and suggests ways that will help you.
July 2022
The basis for this month’s Close-Up was posted on Seeking Alpha June 16. This month, Perry posted “3 Ways to Reduce Risk and 2 Ways to Increase Profits.”
June 2022
The July issue of Technical Analysis of Stocks & Commodities has Perry’s latest article, “Is It Too Volatile to Trade?” It is important to understand when the risk is greater than the reward.
Perry posted “How To Tell When the Bear Market Has Ended” on Seeking Alpha. You might find it useful if you are thinking about getting back in.
May 2022
Perry’s webinar on risk, given to the U.K. Society of Technical Analysts, can be seen using the following link: https://vimeo.com/708691362/04c8fb70ea
The presentation for MetaStock is available on Youtube using the link:
The May issue of Technical Analysis of Stocks & Commodities has a new article by Perry, “In-Sample Test Data, Out-of-Sample Data – Does It Really Matter?” It is a different look at testing.
March 2022
The 2022 Bonus Issue of Technical Analysis of Stocks & Commodities published Perry’s latest article, “50 Years On. What Have I Learned?” It is a summary of the most important trading and development lessons he has learned.
Sunny Harris (MoneyMentor.com) interviewed Perry on Saturday, March 26. Her approach combines both personal and technical questions, having known Perry for many years. You should find it interesting. Go to MoneyMentor.com
Four articles have been posted on Seeking Alpha in March. They are
- How To Control the Risk of Cryptos in Your Portfolio
- How To Find Low-Volatility Stocks That Outperform the Market
- The Best Balance of Stocks and Bonds Will Surprise You
- Determining Whether Crisis Alpha Is A Good Idea Or A Flash In The Pan
Book Interview
Mr. Kaufman appears as a chapter in Mario Singh’s book, Secret Conversations with Trading Tycoons, published by FXI International.
Older Items of Interest
For older articles please scan the websites for Technical Analysis of Stocks & Commodities, Modern Trader, Seeking Alpha, ProActive Advisor Magazine, and Forbes. You will also find recorded presentations given by Mr. Kaufman at BetterSystemTrader.com, TalkingTrading.com, FXCM.com, systemtrade.pl, the website for Alex Gerchik, Michael Covel’s website, TrendFollowing.com, and Talking Trading.com.
In May 2021, Mr. Kaufman gave a 30-minute presentation, “Lagged Trends,” for The Money Show on Tuesday, May 11. You can see it using the following link: https://youtu.be/bh2fA8oBwBk
November 1, 2020, Mr Kaufman taped a session with Andrew Swanscott’s BetterSystemTrader.com.
“The 1st and 2nd Cross” has been very popular with readers. It was published in Technical Analysis of Stocks & Commodities in the March 2020 issue. It is based on an idea of Linda Raschke and captures small but reliable pieces of a trending move. You can find it online.
You will also find back copies of our “Close-Up” reports on our website, www.kaufmansignals.com. You can address any questions to perry@kaufmansignalsdaily.com.
© November 2022, Etna Publishing, LLC. All Rights Reserved.