So, after laying out the early steps and missteps of my trading journey (here and here), I left you yesterday when I was a technical analysis junky. Let’s begin there, and I’ll explain how I came to some very hard conclusions about technical analysis. (Apologies in advance for the length of this post and this series. I tried to give you a sense of my experience without using too many words, but perhaps I used more than necessary!)
I literally read hundreds of books on the subject, including as much of the original source material as I could find. (By the way, I still think the single best book on the subject is Jack Schwager’s big red book. Do not be put off by the “Schwager on Futures” part of the title if you don’t trade futures; it could just as well have been “Schwager on Stocks, Currencies, Futures for Daytraders, Intermediate term traders, and Long Term Investors.” This is, in my opinion, the book for TA as it represents the distillation of everything since Edwards and Magee.) I spent many months focusing on this material, and, by focusing, I mean 12 hour days, most days of the week.
Somewhere along the way, I started to pick up a lot more quantitative tools. I remember reading an article on the drunkard’s walk, and realized that I could pretty easily model that in Excel. So, I created a bunch of random price charts and price series, and, for whatever reason, I spent a lot of time looking at those charts. I don’t know quite what I thought I was doing or why I was doing it. As it turns out, this was probably time better spent than most of my TA research, but I didn’t know that at the time. I had always done a lot of “backtesting by hand”, which, again, I think is an excellent practice that more traders should do. (If I were creating a technical analysts’ certification program it would require a lot of work by the candidate on backtesting, analysis and statistics, rather than focusing on material from books. The market is your teacher. Learn from the market.) In retrospect, much of my early testing was highly flawed because, in doing it by hand, I was introducing a hugely subjective element: “Nah, that one doesn’t look so good. I would have skipped that one.” Really? Are you sure? But this did get me closely in touch with market patterns and I spent a lot of time looking at and thinking about details. I had, little by little, learned how to move toward more automated testing so, where I was doing hundreds of events in a day or two, now I could do thousands in a few minutes.
I had created some magical ratio formula, applied it to data, and found a huge statistical edge–or so I thought–with how often the levels were engaged. Now, I hope some of my readers are nodding their heads and smiling at my naiveté; I was about to have a life changing experience. I had a magic ratio system. It produced good stats on market data. I had random walk price data that I had generated. I had the idea to apply my magical ratio formula to a test on the random price data, and I was amazed: My magical ratio formula was so good that it also worked on random price data! This, certainly, is what Elliott was talking about with those ratios that are the foundations of the Universe. Not only was I going to make a lot of money, but I was probably looking at part of the blueprint of Creation, reading the mind of God… blah, blah, blah. It was a heady time, and I remember going to bed thinking that I had pretty much solved all the problems at once.
Well, I had my epiphany later that night when I realized that seeing the same results on random data as real market data meant that my tool was not finding a pattern in the market. I had, in fact, accomplished the exact opposite of what I thought a had proven. What I had just stumbled into was the concept of significance testing–looking at a statistical test against a random baseline. It’s pretty common to see technical patterns in books or on blogs with a comment like “It could be chance, or…” The authors, unintentionally, are making an important point: we are so easily deceived by patterns; humans have exceedingly poor intuition about randomness. Yes, any chart example, no matter how perfect, could be random–we should require a lot of convincing evidence to think something is not random. The burden of proof should always be on the pattern to prove itself non-random. This was the key question that technical analysts did not seem to be asking: given many occurrences of the pattern, is there something discernibly non-random about price movement following (or, possibly, before) the pattern? That, in a nutshell, is the essence of quantitative pattern analysis: a pattern has to produce non-random price movement. Otherwise, we are simply trading in the random noise and no edge is possible. (In my free trading course, I include a substantial module on the concept that “randomness is the enemy” we must overcome.)
Building a tool kit and asking hard questions
So, armed with this knowledge, I had some work to do and I had to start by educating myself in the discipline of inferential statistics.1 I started re-reading much of the TA literature, and now I was getting concerned. On my first pass through, I wanted to believe, but now I was reading critically and I saw something very different.
I realized that much of what I had read was fraught with error and superstition. Later generations of writers referred to the founding fathers of technical analysis with reverence, but so much of the early work centered around finding cycles and ratios. (Probably because there were not many analytical techniques that could be applied to market data at the time.) Now, I think I understand the value of working within a classical tradition better than most people. As a musician, I understand the importance of historically informed performance practice. You want to make me chew glass? Play a Bach fugue and start your trills on the principle note. I have been fortunate to have been formally trained in classical French cooking. There are ways we do things and reasons we do them, and, though there’s room for innovation, there’s also value in keeping the tradition alive: tourneed vegetables have seven sides and chicken Veronique has grapes, not raisins. I understand tradition, but the problem I saw with this concept in technical analysis was that it was discouraging rational analysis and critical examination of core concepts–don’t think. Don’t question. Just learn the patterns. When I spoke to established traders, authors, and educators about my doubts and the growing body of evidence I had contradicting much of traditional technical analysis, I got a very cold reception. I remember talking to one of the experts on Fibonacci as I was doing research and was completely unable to find any support for the ratios. Eventually, he said he didn’t have time for analysis–he was too busy writing books and selling courses to look at statistics.
The more I dug into the source material, the darker the picture was and the more depressed I became. There were glaring errors and inconsistencies. It seemed that most things had never really been tested, some were probably untestable, and too much rested on appeals to cycles and Biblical authority. (Yes, you read that right. Gann thought he had discovered a secret code in the Bible, and that was the source of his stock trading system.) Every legendary guru I investigated failed to live up to the hype. For instance, Alexander Elder has this to say about Gann:
They claim that Gann was one of the best traders who ever lived, that he left a $50 million estate, and so on. I interviewed W.D. Gann’s son, an analyst for a Boston bank. He told me that his famous father could not support his family by trading but earned his living by writing and selling instructional courses. When W.D. Gann died in the 1950s, his estate, including his house, was valued at slightly over $100,000. The “legend” of W.D. Gann, the giant of trading, is perpetuated by those who sell courses and other paraphernalia to gullible customers.
Modern experts fared little better. I remember one particularly famous chatroom trader with a following of thousands who eventually retired and told me that most of his or her2 trading was “on the simulator”. When I expressed concern that those trades had been the source of inspiration and teaching for many traders s/he said that it made no difference whether they were simulated or real. My trading world was imploding as one thing after another went up in flames. Nearly everything I tested–candlesticks, moving averages, most chart patterns, most indicators, most trading systems, ratios, etc.–nearly everything showed no edge, and the edges I found were very, very small. There appeared to be no 80% patterns in the market; I was finding, at best, 55% patterns. The more I understood human perception and randomness, the more I understood why trendlines, moving averages, or any line at all on a chart could be meaningful to traders. Everyone makes the argument that you can draw any trendline you want; well, it seemed they were right, just not in the way they meant. I held on to things that didn’t work for a long time, but, eventually, I stripped away one thing after another until almost nothing was left. Almost, but not quite.
The silver lining?
Now, I’m leaving one small detail out: I was making money trading, and had gone through a few periods where I made very good money (for my account size). What I was doing was pretty simple: I’d wait for a market to make a move and pause, get in when it started moving again, and get out if my stop was hit. I had also learned and created a simple and robust research methodology to understand price movements. (If I’m honest, I reinvented the wheel, and not very well. Having a formal quantitative education would have saved me a lot of time, but perhaps I grok it better since I bled to learn those lessons.) I realize this post has been pretty bleak, but it was a pretty bleak time in my life (if I ignore the fact I was making pretty consistent money trading.) Tomorrow, I will try to put all of this together into some lessons you can apply in your own analysis and trading.
If you need to educate yourself too, rest assured that it is not complicated. Any smart child could understand the core concepts; the key is that you need to learn them very well and perhaps go a bit deeper than you did in that college stats class you have mostly forgotten. ↩
gender deliberately obscured ↩