This feels unnaturally harsh and likely incorrect. You seem to be suggesting that statistical tools are useless at predicting trends in time-series data.
If the assertion is true that this is useless for the stock market, that would imply the assertion is equally true for macro-economics, and extending further, climate change.
We are collectively deciding how to spend trillions of dollars based on the outputs of these models. Should we not bother then?
> You seem to be suggesting that statistical tools are useless at predicting trends in time-series data.
TA doesn't use "statistical tools" in any actually meaningful or predictive way. It's digital phrenology. Data-driven tea leaf reading. Programmatic palmistry. It deserves only scorn.
Textbook TA uses a lot of voodoo which, itself, is helpful for understanding market psychology. The fact that maybe 20% of market participants give significance to a simple moving average calculated from a certain number of days or weeks-- useful explanatory information!
But the core proposition of TA is that a scientific approach to analyzing past price movements will at least hint at the future, sometimes. Not all the time. That's not controversial, and wouldn't be in any other discipline either. Whether it can be profitably exploited at a particular scale or by a particular person-- another question entirely.
The interesting thing about applying TA to live markets is ... its adversarial. A pattern becomes known, and that leads it to change. If it doesn't change, we can consider it based in fundamentals. For example, markets are more volatile, on average, in the Fall. Why? Well, it's got to be fundamental because everyone knows this pattern and yet it very often repeats. TA is helpful because we would not identify the fundamental mechanism without first observing the historical cycles of prices.
You have phrased your comment very carefully and I agree, sometimes there might be some value in looking at simple trends, if only for the reason that other market participants do the same.
> But the core proposition of TA is that a scientific approach to analyzing past price movements will at least hint at the future, sometimes. Not all the time.
Will those "hints" be correct more than 50% of the time, though? I mean, if TA did beat coin flips, you could exploit this consistently with a profit. That, however, would be news to me.
You seem to be confused between descriptive, predictive, and explanatory models.
TA is not predictive. It can't help you anticipate the future. If it was, you could make consistent profits by using it, and no one has. If it's right, it's just as often wrong, in which case it's no better than a coin toss or throwing bones or reading tea leaves.
It's not explanatory. It provides no hypotheses for why the market behaves in certain ways. If it did it might have some hope of being predictive, but alas, as I already mentioned, it's not. And thus it can't teach us anything about market behaviours or their underlying causes.
TA might reasonably be thought of as descriptive, in that it gives a (voodoo) framework for describing observed market behaviours. As you say, we might observe the market is more volatile in the fall. But because it offers no explanatory power, we have no way to know why, and since it has no predictive power, it can't tell us if next year will be the same as this year. You're simply expected to believe that, well, it's always been that way, so I'll assume the future will be the same as the past.
As a result, it's frankly not that useful or interesting.
Stock price movements are literally text book examples of unpredictable (or more specifically, random walk) time series in university time series courses. A proper statistics course teaches you that you can predict with great certainty that the mean of the fraction of heads over many coin flips converges to 1/2, while you cannot predict anything about the outcome of the next flip. A good statistics course will also teach you how both can be true. TA is like trying to use statistics to infer something about the outcome of the next coin flip.
The standard model is not a plain random walk, but a random walk with drift. (Actually a geometric random walk with drift, but thats besides the point.) The drift term for the broader US stock market is usually assumed around 8% (long term average gain based on around a century of data). That works out to around 0.022% a day, which is not what TA traders are looking for.
Doesn't "technical analysis" mean pretending that you can ignore that it is Tesla or NVIDIA and just believe that the time-series itself tells you what comes next. I.e. you believe that there is some "nature" of stock ticker data independent of the financials and business environment of the particular company.
If stock prices are so easy predict, why aren't you putting your money where your mouth is? :P Or are you? Then I'd love to hear about that obscene fortune you've made so far. :)
In all seriousness, though, there is an entire field of research whose results substantiate GP's point. This is not to say that you can't beat the market but the challenge lies in doing so consistently.
They may have missed it this time, but if you look at a recent stock market history, it was impossible to not make a fortune out of a pretty obvious portfolio. /s
> I can say with relatively strong confidence that Tesla is going to be worth more than $1 trillion in three years.
How relative is your strong confidence? If you feel this is inevitable, you can become essentially infinitely rich. Why not do it?
Of course so could others. And all your collective trades move the market. At which point it's no longer true. It can't be, not everyone can become infinitely rich.
Technical analysis is based on statistics the same way astrology is: that is to say, it incorporates some of the jargon and features shallow, surface-level incorporation and features. TA is worse because it affects more people.
While we should be sceptical of any and all "models" of reality, we can't lump them all together.
Macroeconomic models have had spectacular fails but are fundamentally a semi-honest attempt to understand how the economy evolves (Only "semi" because ideology can be quite limiting).
Climate change models have a strong physical component that is neither made-up nor manipulated, but may suffer from not capturing biosphere dynamics all its complexity. Thats why there is uncertainty range around scenarios.
Long term investment decisions are in any case never based on TA. That technique is really a made-up pseudoscience tailored to provide comfort and talking points to the widest possible trading audience.
If the assertion is true that this is useless for the stock market, that would imply the assertion is equally true for macro-economics, and extending further, climate change.
We are collectively deciding how to spend trillions of dollars based on the outputs of these models. Should we not bother then?