Written by Jeff Miller, A Dash of Insight
— this post authored by Mark D. Hines
Our previous Technical Thoughts asked the question: How Do You Filter Out Noisy Trade Signals? We considered the explosive growth in news and in data, and noted “information overload” is becoming an increasingly common complaint from traders. However, sticking to a disciplined trading process can help you do a better job of filtering out distractions and filtering in the data most important to your trades. A glance at your news feed will show that the key points remain relevant.
This Week: Ominous Vomiting Camel Formation
Please share this article – Go to very top of page, right hand side, for social media buttons.
If you are not familiar, the Vomiting Camel formation is one of the most ominous signals a trader can receive. For example, this dangerous formation recently formed in both Bitcoin and Gold, and the subsequent market moves were devastating, as shown in the following charts.
Gold:
Bitcoin:
However, to be clear, the Vomiting Camel formation is a joke. It’s not a real thing. And we are using it to highlight the dangers of building trading programs that over-fit the data. For example, some trading models over fit all of the data that is available to come up with meaningless patterns. Similarly, some humans draw absurd lines on charts, such as the vomiting camel. For more information on the origins of the Vomiting Camel, check out Katie Martin’s recent article:
Machine Learning and Over Fitting the Data:
We’ve written about the challenges of trading programs and machine learning/ artificial intelligence (“AI”) in the past, for example here:
And this continues to be a challenge both inside and outside the world of trading as the prevalence of artificial intelligence grows. For example, Thomas T. Hills does an excellent job of highlighting some of the common mistakes of algorithmic learning, ranging from autonomous vehicles misreading stop signs to Google Flu dramatically over forecasting the intensity of influenza outbreaks, in this article:
And while our trading models are not as complex as autonomous driving algorithms, we do find that human “supervision” is an important element of the process. Specifically, we find it better to “lock in” our trading approach, and then allow the humans to do periodic testing and revisions.
Another important consideration is “trading teamwork.” Having people around to discuss, test ideas, and provide support is helpful. This excellent recent article by Brett Steenbarger lends itself to the question “Do you have trading teamwork?”
Overall, we find that having checks and balances between trading models and human beings can help reduce the risks of over fitting the data, and it can help prevent both the machines and the humans from arriving at inappropriate results.
Model Performance:
Per reader feedback, we’re continuing to share the performance of our trading models.
We find that blending a trend-following / momentum model (Athena) with a mean reversion / dip-buying model (Holmes) provides two strategies, effective in their own right, that are not correlated with each other or with the overall market. By combining the two, we can get more diversity, lower risk, and a smoother string of returns.
And for these reasons, I am changing the “Trade with Jeff” offer at Seeking Alpha to include a 50-50 split between Holmes and Athena. Current participants have already agreed to this. Since our costs on Athena are lower, we have also lowered the fees for the combination.
If you have been thinking about giving it a try, click through at the bottom of this post for more information. Also, readers are invited to write to main at newarc dot com for our free, brief description of how we created the Stock Exchange models.
Expert Picks From The Models:
This week’s Stock Exchange is being edited by Blue Harbinger; (Blue Harbinger is a source for independent investment ideas).
Holmes: This week I purchased Petrobras (PBR). Are you familiar with this stock, Blue Harbinger?
Blue Harbinger: Yes, Petrobras is the big Brazilian integrated oil and gas company. Why do you like it, Holmes?
Holmes: As you know, I am a “dip-buyer.” I purchased the shares on April 25th, and I believe they have upside over the next 6-weeks – which is my typical holding period.
BH: Interesting pick, Holmes. I recall when Petrobras was trading at more than three times its current price. However, it’s sensitive to oil prices, and the Brazilian business environment makes me nervous. Here is a look at Petrobras’ Fast Graph.
Holmes: I am aware of the history of PBR’s stock price, but as I already stated, my typical holding period is about 6-weeks. I’ll be out of this trade before the fundamental story plays out. And to be clear, unlike you, I am a trading model, not a human, so I am not “nervous” about the Brazilian business environment. I have my price target for this trade, and I pay attention to macro factors and use stops for risk management.
BH: By macro factors, do you mean the price of oil? It’s been showing some strength in recent months, and crude is up above $70.
Holmes: Thanks for that info too, but just pay attention to the price yesterday, when I bought, compared to about 6-weeks from now when I’m likely to sell.
BH: Ok. Thanks, Holmes. And how about you, Road Runner – any trades for you this week?
Road Runner: This week I purchased Proofpoint (PFPT). It operates as a security-as-a-service (SaaS) provider that enables large and mid-sized organizations to defend, protect, archive, and govern their sensitive data worldwide.
BH: I know the company, and I am aware that the stock price has basically been on fire over last year, increasing from around $40 to now over $120. I suspect this is driven by big IT spending budgets and spending appetites. It seems corporate IT departments were starved for budgets, until recently, and now they’re spending a lot, especially considering the economy has been strong. Here is a look at the Fast Graph.
Road Runner: Thanks for that background, but my typical holding period in only 4-weeks. I like to buy stocks at the lower end of a rising channel, and based on the chart I provide earlier, you can see that is exactly what I did.
BH: I tend to believe this stock has more room to run, so I actually like this trade, although my personal holding period would likely be longer than 4-weeks.
RR: It’s nice to know you approve. Anyway, how about you, Felix – any trades this week?
Felix: No trades to share this weeks, but I do have a ranking for you. Specifically, I ran the 30 Dow Jones stocks through my model, and I’ve listed the top 20 below.
BH: Interesting ranking. What is the criteria?
Felix: I am a momentum trader. However, my typical holding period is around 66 weeks, on average – much longer than the other traders.
BH: I see Boeing at the top of your list, and they’ve certainly had some momentum over the last 2-years. I also think Intel is attractive, based on the continuing strength of the semiconductor industry, although it seems Nvidia keeps gaining ground on them in terms of AI chips. Anyway, thanks for sharing these ideas. And how about you, Oscar – do you have anything for us this week?
Oscar: Yes, this week I ran a new universe through my model. Specifically, I ranked the High Liquidity ETFs with price volume multiples of over $100 million per day. Here are the top 20:
BH: I see Bitcoin (GBTC), ranked near the top of your list. This makes sense considering you are a momentum/trend-follower trader, and GBTC is up over 30% in the last month (although it is still down dramatically from its highs back in December). Aren’t you afraid of the “vomiting camel” pattern we saw earlier?
Oscar: Very funny. However, I am an objective trading model; I don’t draw camels on charts like some humans do. I typically hold for about 6-week and then rotate to a new ETF.
Conclusion:
Some trading models over fit all of the data that is available to come up with meaningless patterns. Similarly, some humans draw absurd lines on charts, such as the vomiting camel. We find that utilizing the data sifting power and objectivity of our trading models, combined with the testing and “supervision” of humans, can lead to superior results. We use a combination of models that allow us to keep market correlations low and expected returns high.
Background On The Stock Exchange:
Each week, Felix and Oscar host a poker game for some of their friends. Since they are all traders, they love to discuss their best current ideas before the game starts. They like to call this their “Stock Exchange.” (Check out Background on the Stock Exchange for more background). Their methods are excellent, as you know if you have been following the series. Since the time frames and risk profiles differ, so do the stock ideas. You get to be a fly on the wall from my report. I am usually the only human present and the only one using any fundamental analysis.
The result? Several expert ideas each week from traders, and a brief comment on the fundamentals from the human investor. The models are named to make it easy to remember their trading personalities.
Click for large image.
Getting Updates:
Readers are welcome to suggest individual stocks and/or ETFs to be added to our model lists. We keep a running list of all securities our readers recommend, and we share the results within this weekly “Stock Exchange” series when feasible. Send your ideas to “etf at newarc dot com.” Also, we will share additional information about the models, including test data, with those interested in investing. Suggestions and comments about this weekly “Stock Exchange” report are welcome.
Trade Alongside Jeff Miller: Learn More.
