Why popular methods of technical analysis fail to work but quant analytics or statistical analytics may
how right you are. I agree 100% with all your findings and conclusions. I have reached the same conclusions years ago. The expected value of playing heads or tails is zero, no matter what time frame is used or time horizon considered. It has been known for centuries that the outcome of a 50/50 game is zero expected gains.
You conclusions could also be viewed with the added twist that there is no harm in using a random entry, or any other decision surrogate, since your expected profit will still remain zero; which is another form of saying your expected loss is also zero. No gain, No loss.
Appreciated reading your paper, great work.
The naive and unduly complicated methods is ineffective. Statistical tests show it. Efficient methods such as log-regression, being built on 10 years of historical data on multiple instruments (4-50 trading instruments with similar behavior covered by the general universal model). In this case, statistical tests show a dependence. But even then, you can not expect low-risk, incredibly high returns. Must expect loss-making quarters or half-years, and long periods of low trading activity.
You can not limit yourself to a few trading instruments. Need all the time to look for other more predictability trading instruments. For example, last year, well predicted precious metals, and this year they are harder to predict.
The problem is that all this is difficult to explain to beginners, it is difficult to sell the analyst honestly showing all the risks. Our company makes a market analyst (forecasts, trading signals) and is often sees the naive expectations of clients.
I have read hundreds of those academic papers and usually agree with their conclusions as I agreed with yours. But somewhere you say: “The problem is that all this is difficult to explain to beginners”. I’ll pass on that one.
I’ve posted some remarkable charts on this thread with let’s say performance results somewhat higher than the Buy & Hold strategy. I’ve said many times that I do not predict prices, that I often use random entries to get in a position. But all that is not the point.
My methods are the execution of administrative procedures, investment policies that you set from the inception of a portfolio, nothing more. As in this is what we will do if this happens. And the major point I am making is that simply by re-investing the accumulating profits, instead of doing nothing with them, one can easily achieve an exponential Jensen ratio and increase his/her portfolio performance above the Buy & Hold by default due to administrative procedures. Oh, I do make a prediction using these methods and that is in 20 years from now, the markets will probably be higher than today.
I’ve just posted on my web page a more elaborate research note:
I agree with you for the most you are saying in your paper but I have different opinion in some points. Some things I see in different way.
For example in your article (page 6) “Summing up the results”, “2.Simple technical Analysis tools, including popular indicators and instruments…prevent traders from making money” – I party agree, but for “Popular tools do not allow a trader to stand out from the crowd to get bigger piece of the trader’s pie.” I fully agree.
Regarding the role of the technical analysis in trading, before 1980s most researchers criticised it and concluded that is not able to produce as good results as the buy-hold strategy (Alexander, 1961; Jensen and Bennington, 1970; Fama, 1970). However, later studies have proven the opposite (Pruitt and White 1988, Bessembinder and Chan 1995). Trading rules/systems (using technical indicators) based on past data could create excess returns and prove that usefulness of technical analysis. Such conclusions had Brock, Lakonishok and Lebaron (1992), later Bessembinder and Chan (1995, 1998), Lo, Mamaysky and Wang (2000) and many others.
Another view is from psychological aspect as behavioural finance. The premise of behavioural finance is that conventional finance theory ignores how real people make decisions and that people make a difference (Constantinides, et al., 2003). While conventional finance presumes that investors are rational, behaviour finance starts with the assumption that they might not be. As a result, we can discover market anomalies such as mispricing (psychological bias could cause prices to deviate from their correct level) and take the advantage to apply proper strategies ( such as arbitrage) to make profit.
Generally, behavioural biases may be consistent with technical analysis (Bodie, et al., 2009, pp. 395). There are chart/price patterns in technical analysis that are repeated many times in the past depicting specific investors behaviours. For example, the process of price underreaction and overreaction to public information might become visible in the form of the technical analysis as abnormal (upwards or downwards) trends.
“technical analysis” vs “statistical analysis”
“Technical analysis” is pseudoscience (numerology). This is only a “good” selection of the conditions and facts.
“Statistical analysis” is a science. This is an exact description of the restrictions, the definition of hypotheses and their mathematical proof or disproof. Statistical analysis was successfully applied in many fields.
Let’s use a statistical analysis for the markets.
Here we are all to exchange ideas and say our opinion.
I did not say that Technical Analysis (TA) is science. I implied that TA is useful as assistant tool to make trading decisions. Lately, most academics agree with that.
We all know the debate between academics and practitioners regarding TA (art or science). Most say that TA is art not science, ok? Let’s say technical analysis is ‘Art’. I will present only the following:
John R. Kirby, seeking to clear-up the art vs. science debate, quoted highly regarded Technical Analyst Ralph Acampora: “‘Art’ means a skill acquired by experience, study,
or observation. ‘Science’ is a body of knowledge with its own axioms, rules, and language (Technically Speaking, 2005, January).”.
TA is not a standalone “art’ – it requires the help of statistics. Statistics give an indication of the strength of a trading system or set of trades at any point in time. Technical stock analysis and trading statistics are important to identify areas of your trading which require improvement and help to illustrate your trading characteristics. They inform you if you are on the right road to success.
Finally, the goal is to make profits with low risk in a consistency way. Every one has developed his own style (TA, statistics, data mining, AI, etc).
Statistics give an indication of the strength of a trading system or set of trades at any point in time.
Can you please give examples of TA articles, which assess the probability of non-randomness of the results of testing of the trading systems? I have not seen the TA where the notion of “probability” is not in the most naive form of the arithmetic mean.
There are pockets of non-randomness caused by trader effects and other events such as news events. These are what the pro discretionary trader seeks to exploit.
Standard statistic methods are far too crude a tool to apply to price action.
Consider the Brownian processes. Without statistical analysis, we find a lot of false “are pockets of non-randomness”.
The markets are not a true random brownian process. You need to learn about ‘trader effects’. A skilled trader can find an edge.
Many people have been down the statistical analysis road. It leads nowhere.
Models simplify reality and require assumptions and /or simplifications – so one model cannot capture all aspects of the real situation (world). Brownian motion, allthough it helps, is just one theory that attempts to explain stock market fluctuations, along with the random walk theory and Markov process.
If the stock market is truly random, as described by Brownian Motion, then the market is unpredictable and efficient. Efficient market is that it is impossible to ‘beat the market’ since all relevant information is already reflected in the market price.
I agree that the market does not Brownian process. Brownian process is an idealization, and all real processes different from Brownian process.
But, warning! First experiment. If you observe and analyze the time series is actually the ideal generated by the Brownian process (but you do not know it), you “find” many “patterns” But these patterns are false and not will be confirmed by new data.
In the second experiment, you’ll observe the nonideal Brownian process with a small deterministic components. But in this case, the randomly noise can hide the real pattern.
Market data is like that. The market is a composition of two processes – the strong randomly proccess and weak deterministic process. Without statistics, you’ll never be able to distinguish between false and true patterns. Technical Indicators result of systematic detection of false patterns.
FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!