Above is a screenshot of our internal EWAVES Live web interface, which we use as a front end to our server running EWAVES 3. We have not discussed EWAVES (Elliott Wave Analysis & Validation Expert System) in a while but the project has been racing along. The greatest innovations of the project’s history occurred during the development of version 3 and are now behind us. The research and development backing this version required a complete re-imagining of wave analysis from the ground up, breaking all structural ties with past versions.
We completed the analysis engine in 2020, and the web interface in January. Continual refinement will never end, but the program has reached the point at which we are now taking time to figure out its optimal application. And we are extremely excited.
What is so exciting? Let’s count (no pun intended) the ways:
We have reached a significantly higher level of analytical quality than ever achieved in machine wave analysis. One experienced in-house EWI analyst, upon seeing the output for the first time, said, “so far almost every single EWAVES count I’ve just looked at that the software considers ‘high quality’ is exactly how I would have labeled that market.”
The reason for this improvement is due to the extensive research we did into real-world Elliott waves. We spent several years doing a research & development deep dive, studying every aspect of wave behavior. What emerged from this process was a thorough statistical understanding of the Elliott Wave Principle (EWP), as well as the discovery of many new aspects of Elliott waves that have never been written about before.
Of course, understanding Elliott waves is only half the battle. Figuring out how to apply that knowledge into a working model is another task entirely. It has been a challenge, but the system we now have in operation has finally reached the point where it is now a “90% solution”. In other words, any further improvements are likely going to be relatively incremental compared to the progress we have achieved to date.
At the end of the day, the wave counts speak for themselves. We will take a look at some of them in this issue.
EWAVES can now quantify the extent to which individual markets adhere to the Elliott wave model, a measure we call elliotticity. This is a dynamic measure that shifts as time progresses, yet it displays persistence within individual markets and asset classes. Currencies are the most adherent to the model, while bonds are the least adherent. Among individual stock issues, as you might expect, high-volume and high-market-capitalization stocks are more likely to have high adherence than thinly traded issues. Elliott waves model herding, and herding occurs more reliably among the largest crowds.
We delineate multiple types of elliotticity. Total elliotticity takes into account the entire history of a market. Markets with the highest total elliotticity conform best to the Elliott wave model at all degrees of scale. Short term elliotticity helps us recognize markets that have recently traced out clear Elliott waves and thus are good candidates for near term forecasting. Total elliotticity is slower to change than short term elliotticity. Nevertheless, the short term picture is important even for long term investors because it offers a tool for risk mitigation. It is prudent to wait until both metrics reach high levels before establishing a long term stance.
Elliotticity is designed for practical use. We use a percentage scale of 1-100. Our percentage numbers are not percentiles, by which a 90% reading would represent the top 10% of all considered markets. They are independent measures of EWP adherence. Therefore, it is possible to come across many more markets than you might expect scoring above 90%, and conversely there are times at which all markets might be below 90%. Markets scoring at 90%+ are about as good as real markets get in terms of model conformance. So far, the highest number we have seen is 97%.
A market with very low elliotticity means the software is unable to label the chart in a useful way. Something rewarding emerged when we started investigating some of the lowest ranked markets: In many cases, the low elliotticity was due to corrupt data. Below is a real example of such an outcome occurring in HYG (a junk bond ETF), where the data providers did not adjust properly for a split. EWAVES rejected the data as non-adherent. So, amazingly, just by measuring markets’ adherence to Elliott-like behavior, we have found an effective tool to help find data errors.
On a related note, a few months ago we revisited our Real vs. Random study, which judges EWAVES’ ability to distinguish real market data from statistically compatible, randomized market data. Following statisticians’ lead, we generated fake data by scrambling the daily returns on the DJIA to generate a sequence that looks like real market data. Old versions of our software had trouble telling the difference and only got the answer right about half of the time. In our preliminary study, EWAVES 3 rated real market data as more model-conforming than every example of the fake data. We look forward to continuing this research and eventually publishing a formal paper, but our preliminary test suggests that the program is seeing something in real data that is not there in the fake data. We think that something is elliotticity.
Another way to think about elliotticity is as a quantitative formalization of what Elliott wave practitioners have been doing for years: identifying markets with a clear pattern. So, a key application of EWAVES is to use it as an elliotticity search engine. About a month ago, we got this idea up and running on the internal EWAVES Live site by ranking hundreds of markets based on their EWP adherence. By excluding markets that have low elliotticity from investment consideration, an investment strategist doesn’t have to worry about them. It is better to focus investment dollars where the waves are clear.
Aside from elliotticity, another important property of markets that we are researching is the overall degree of wave maturation. A market that is mature in its evolution has, by definition, a more complete picture, which implies less guesswork than a less mature market. Elliott wave analysis excels at identifying major peaks and troughs, because such turning points are preceded by a completed Elliott wave pattern at several degrees of trend.
3. Qualitative, not Quantitative, Properties
As mathematicians have demonstrated, financial market pricing is a fractal. A fractal is an object that is similarly irregular at all scales. Quant systems depend upon regularities in quantitative data. But the financial markets have no such regularities: no reliable periodicities, no reliable deviations from trend, no reliable extremes, and so on.
The market’s lack of adherence to quantitative norms is why we designed EWAVES to be a qualitative system rather than a quantitative one. To EWAVES, the form of market behavior is paramount, so absolute price distances and time durations are irrelevant. EWAVES does not care how long the average bear market is. EWAVES does not care what happened the last time the market went up or down for X straight days, weeks or months. Yet, when it sees a clear wave, of any size, it can recognize it. Programming a computer to “see” the market in this qualitative, form-based way has been a challenge. It required a different way of thinking.
One of the critical characteristics of our qualitative system is that it is scale invariant, which is the most important of several types of geometric transform invariants to which EWAVES adheres. The full list includes not just scale invariance but also translation invariance, reflection invariance and others. If you stop and think about it, this is so logical as to be axiomatic. After all, most pattern recognition a person does on a day-to-day basis is highly invariant to many geometric transforms: a tree is a tree regardless of its size, whether one inch or a hundred feet, or if it is moved from one place to another, or if it’s flipped upside down. So it is with the market’s patterns.
Scale invariance implies that markets create their own forms and have long-term memory. Apart from scale, a 100-year wave form has the same implications as a 10-minute wave form.
To take advantage of markets’ long term memory, EWAVES performs its wave counts with as much historical context as possible. EWAVES’ analysis on the major U.S. stock indices incorporates context going back to 1697, and it labels the data from the multi-century time frame all the way down to daily, combining all of it together, like a gigantic puzzle, into a single, progressing Elliott wave. By zooming in and out of the entire contiguous picture, the user gets an experience similar to that of the many Fractal Zoom videos posted on YouTube. But, instead of an artificially constructed fractal, we’re dealing with real Elliott waves, which are naturally occurring fractals generated by human herding behavior. These real-world forms have profound practical implications rather than just academic, theoretical ones.
How can we go back so far when the major US stock indices do not have data going back to 1697? We prepend data from preceding records to the stock index in question to achieve an approximate longer-term history. In house, we call this process stitching. The amount of independent datasets that can be stitched together is unlimited. So, what we usually do for the major indices, such as the DJIA, NASDAQ, S&P, NYSE, Russell 2000 or Dow Transports, is to create a stitch-chain composed of the index as far back as it goes and then earlier compatible datasets.
Looking at all the data together across multiple timeframes gives a different perspective than the often-seen focus in the media on individual “bull markets,” “bear markets,” “bubbles” or “crashes.” Most people define a bubble as an event featuring runaway speculation, credit expansion and irrational investing behavior. The EWAVES program is unfamiliar with these psychological and fundamental factors. It focuses on fractal structure. This is just what we want, because bubbles are not singular events, as popular theory would have you believe, but merely part of the hierarchical structure of the market. In this very real sense, the dot-com bubble, the housing bubble, and the current “everything bubble” are all just sub-bubbles within the larger uptrend since the 1980s, which is itself a larger uptrend since the 1930s, and so on. For a good grounding in this idea, read Chapter 23 of The Socionomic Theory of Finance, which is titled “Popular Bubble Theories vs. the Elliott Wave Model.”
4. Adaptive & Unbiased
A good Elliott wave practitioner should be as flexible as possible. But when a forecast goes awry, it can be hard for the human mind to change stance and recognize that an alternative, formerly less probable wave count is now in play. EWAVES, however, knows no uncertainty, worry or shame. It has no ego. It has no clients to please. It cares not what others may say. High real time elliotticity is critically important for a forecast, but it is never a guarantee, so the program is flexible. It goes where the probabilities lie, and it does so instantly. Because it is designed to interpret real time market data, it will change its mind instantaneously, sometimes radically, when new data changes the pattern’s analytical implication.
EWAVES operates in an emotional vacuum. It considers only its raw data input, comprising price and time data. It does not read news publications, and it knows nothing about anyone’s market opinion. It knows nothing about the latest government edicts, the latest Fed actions or economic fundamentals. It has been trained based on over a thousand real market examples, but it has no particular attachment to any market or wave count. It doesn’t even remember what its own opinion was yesterday. The way it operates is the epitome of unbiased wave analysis.
Now, let’s look at some examples of how to use EWAVES as a search engine across markets.
Example: The Stock of Walt Disney Company
EWAVES flagged Disney on January 13, 2021 as having potentially completed wave 4, with a very high ranking of 93% short term elliotticity. In the above chart, we are concerned with the final rise beginning at the orange dotted line. EWAVES’ analysis prompted us to issue a Flash recommendation on January 14.
The most conservative target for this forecast was for new highs above the wave 3 peak. A typical invalidation point for such a forecast would be if the market broke below the high of wave 1, because wave 4 cannot overlap 1 without invalidating the impulse wave. EWAVES was able to provide us with an even tighter invalidation point due to a combination of other wave relationships.
The market moved against us initially. Even though the most likely event was for wave 5 to begin, wave 4 continued lower.
Yet even though EWAVES was initially wrong on the very short term structure of the market, it quickly adapted to the market, and gave us precise information: EWAVES concluded that although wave 4 was abnormally large, it was still most likely a fourth wave, which had just terminated. It indicated that this was unusual behavior by recording a drop in short term elliotticity from 93% to 82%. Even though 82% elliotticity is lower than it was before, it is still above 80%, which is our benchmark reading for reliable patterns. So, we waited. If EWAVES had indicated otherwise, we would have decided immediately to cancel the recommendation.
Disney then rocketed higher, in a wave 5 advance, confirming that EWAVES’ notion that Disney was in a wave 4 position was correct, regardless of the precise timing of wave 4’s endpoint. We decided to exit after a strong up day as the stock made new highs above $190, a level that exceeded the minimal expectation for wave 5 (which, practically speaking, is just to get to a new high). After the market closed, we updated EWAVES’ count, which interpreted the high as wave 3 of 5, signifying that the market was most likely nearing the start of a larger correction, thus validating our conservative exit decision.
Example: Soybean Oil Commodity Future
On March 1, EWAVES was reporting nearly 90% elliotticity for Soybean Oil on multiple timeframes. The following day, we sent the chart shown in Figure A along with a buy recommendation to Traders Flash subscribers. Soybean Oil was positioned within wave 3 of 3 of 5, with further to go on the upside.
For analysts that rely on trend following indicators, trying to enter powerful moves such as these is a highly risky endeavor, because the stronger the move, the stronger the ultimate reversal. By definition, trend following cannot anticipate turning points.
EWAVES, however, uses contextual pattern recognition to understand the evolution of each wave from beginning to end, so it can anticipate moves rather than just react to them. It is designed to determine times when a wave is mature and to identify its reversal either in advance, at the time, or in swift retrospect.
By understanding the difference between impulsive and corrective price patterns, EWAVES can use countertrend moves as opportunities to enter the main trend. In contrast, large countertrend moves will often whipsaw trend followers. Trend following indicators are quantitative by nature, completely dependent upon pre-chosen numerical indicator settings. But the market’s trends do not adhere to quantitative norms, as anyone who’s ever tried to use a moving average figures out pretty quickly.
The most important advantage of EWAVES is that it is qualitative. It can identify the market’s patterns regardless of their quantitative duration, price change, speed, volatility or any other such aspects.
While most of our setups involve entering against the trend within a corrective formation (usually wave 2 or 4), the more aggressive style of recognizing an unfolding 3 of 3, while seemingly carrying more risk, is designed to provide a positive result almost immediately.
After we transmitted Figure A to subscribers, Soybean Oil immediately continued higher. We followed up with an exit recommendation at a higher price for Traders Flash.
Example: The Dow Jones Transports ETF
Most investors pay close attention to the three major U.S. stock indices: the NASDAQ, the S&P 500 and the Dow Jones Industrial Average. Few people have the time to regularly study all the slightly more esoteric indices, such as the Russell 2000, the NYSE or the Dow Transports.
A machine in our office, however, has nothing better to do than to sit and watch hundreds of markets every day on our behalf. The Transports caught our attention by hitting 79% short term elliotticity on March 5, 2021, a number supported by 90% total elliotticity. This means that the DJTA was highly conformant with the Elliott wave model at all degrees of scale. EWAVES ranked it as the clearest of the major US stock indices.
The short-term wave count indicated that we had just finished 2 of 2 of 5, and therefore the market was about to take off. We released an ETF Trader’s Flash the following day.
Figure D shows the evolution of the Dow Jones Transportation Average over roughly the next month, as it moved about 11.5% higher. Wave 5 had met several minimum targets and so, while the substructure had the potential to continue higher, we felt that it was prudent to exit the recommendation.
If you’d like to learn more about EWAVES, sign up for the open-access EWAVES newsletter at ewaves.com/newsletter, or follow Qualitative Analytics on Facebook at facebook.com/qualitativeanalytics. If you are interested in taking a look at Elliott Wave International’s Flash Services, which issues alerts after EWAVES scans hundreds of assets each day looking for charts with the highest elliotticity, please visit https://www.elliottwave.com/Trader-Analysis/Flash.