Perhaps you’ve hear about the “Hindenburg Omen” in the news lately. This technical signal is triggered when a significant number of stocks (about 2.6%) are hitting 52-week highs and a significant number of stocks (2.6%) are also hitting 52-week lows. There are other criteria involved, but you don’t have to be a technician to understand why traders care about this strange divergence in price action. This kind of a bipolar market movement is rare and indicates a lot of investor uncertainty.
Of course, we already know that the market is currently very fragile and uncertainty is high. What the omen is really supposed to be able to predict are market crashes. We are interested in this signal right now because it completed yesterday. But should you really be worried about a crash?
The omen includes additional criteria like volatility indicators, moving averages and a minimum number of 52-week divergences in a row. Critics would suggest that the pattern is over-fitted to historical data and that the actual predictiveness of the pattern is overstated.
Looking at the data in the past, I would definitely agree. It makes for dramatic news articles on Bloomberg or CNBC, but it is probably not predictive enough to use by itself. Fortunately the criteria are easy to quantify, and you can test it yourself.
This doesn’t mean that the market won’t crash. Stocks seem quite fragile right now and we have been advising very robust risk control for a while; however, the Hindenburg Omen is probably more of a coincidence than anything else. You can test this by building a model that copies the criteria and back-testing it. What you will find is that very small changes to the criteria, data, time periods or sample size will completely change your results. With certain criteria it is very predictive, but a very small change can make it appear nearly random. This problem means that the model is not “robust.” Statistics models are often not robust because there are large outliers (or surprises) in the data. Surprises and outliers are particularly large in the stock market.
I think this is a serious problem for traders. Too often they will apply a “probability” to a technical indicator without understanding the differences between market data and other kinds of data distributions. Systems and indicators that work really well on past data tend to fade as the market continues into the future because it was fitted to the past data without any tolerance to small changes in the market environment. You’ve no doubt heard the phrase “past performance is not necessarily indicative of future results.” That phrase is included in almost every prospectus you can find because it’s true. We have to be very careful about taking past data too far.
We don’t need to get into too much detail to understand that if you were to sell on every technical sell or buy on every buy signal, then you would be trading way too often. The Hindenburg Omen is interesting and is probably accurately showing a very high level of market uncertainty and anxiety in the market — which we already know — but running for the exit every time a signal like this hits the newswire will probably not do much more than increase your commissions. I would suggest that the same factors driving this indicator (risk and uncertainty) are reason enough to be trading very cautiously right now.
This article is brought to you by LearningMarkets.com.
Daily Trader’s Alert — Yours FREE! — In each issue, InvestorPlace’s Chief Technical Analyst Sam Collins gives you his take on what’s slated to impact your portfolio during the trading day. It also includes Sam’s Trade of the Day — his daily stock or ETF pick complete with chart and trading target. The Daily Trader’s Alert is yours free, sent right to your e-mail inbox each trading day before the market open. Click here to get started now.
