Stock Market Technical Analysis

[roger]

Last I checked, TA was one of many ways of analyzing the potential gains/losses of a stock. It is just another tool in the toolkit. Nothing more, nothing less.

Except that all available evidence shows that it has a negative expectation.

I prefer that my tools actually fix things, not break them, to continue your analogy.

 

[drkitten]

One of the other posters noticed that this person kept buying in at times where his profits weren't optimal and recommended using TA as a tool for that person to find his buy-in point.

The only problem with that is that I have no confidence that any theory of TA with which I'm familiar would be any better at finding a buy-in point than random guessing.

 

[roger]

Is there any other part of this forum where we would accept anecdotes and glib statements as facts? Imagine:

• Sure, lots of homeopathy is hooey, but used correctly it can complement and improve on conventional therapies
• I know, most haunted house stories are woo, but I've been in many real ones.
• No, tarot cards won't solve all your problems, but they are a useful indicator.

etc.

The simple fact is that all available studies (that I know of) show that TA for stock selection does not work - it's a negative expectation act after you account for transaction costs, etc. So, yes, it is something to get "dogmatic" about, IMO.

 

[corplinx]

The simple fact is that all available studies (that I know of) show that TA for stock selection does not work

I didn't suggest that it was a method for selecting stocks. I pointed out that it is one of many tools for evaluating stocks. I pointed out that you can check the "charts" to pick your buy in point for a stock you already plan on buying.

I did not suggest the thing you are accusing me of suggesting. I have no problems putting words in my own mouth and do not need your assistance.

 

[roger]

I didn't suggest that it was a method for selecting stocks. I pointed out that it is one of many tools for evaluating stocks. I pointed out that you can check the "charts" to pick your buy in point for a stock you already plan on buying.

I did not suggest the thing you are accusing me of suggesting. I have no problems putting words in my own mouth and do not need your assistance.

Sorry, I used language too loosely - I meant predicting price movements of stocks. The literature is very clear on this point - it a negative expectation proposition.

I ask again with the above clarification - evidence for your assertions?

edit: I will go further with this and offer you 2 broad surveys of the literature: Malkiel's A Random Walk Down Wall Street, and Damodaran's Investment Myths, as general support for my assertions. They cite many individual studies, all pointing to the ineffectiveness of TA in predicting price movements in any profitable way.

A brief excerpt from the former:

The technical rules have been tested exhaustively by using stock-price data on both major exchanges going back as far as the beginning of the twentieth century. The results reveal that past movements in stock prices cannot be used reliably to foretell future movements. The stock market has little, if any, memory. While the market does exhibit some momentum from time to time, it does not occur dependably and there is not enough persistence in stock prices to overwhelm the substantial transaction cost involved in undertaking trend-following strategies

Let me point out another weakness in my use of language - "does not work" does not mean "does not make money". The measure of an investment system is that it performs better, risk-adjusted (though I think the whole "risk-adjusted" thing in academia is a can of worms), than the market as a whole. If the market is going up over time, which is always have (given a big enough time slice), it'd be a truly horrible system that actually lost money. But if over decades a passive style yields 6%, and some system yields 4.5%, well, that is "does not work" by any useful measure. It may perform better than some truly bad method - such as an investor going by their "gut", which can have many unconscious behavior patterns with very negative expectations. That in itself is no reccomendation. I bring this up not because you, corplinx, made this argument (you seem touchy on this point), but because I often hear retorts along those lines.

another edit: forgive me, it is late, but I'm hopped up on caffeine. The mind wanders a bit. But I also point out the book Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals by David Aronson for a truly sceptical look at TA. Conclusion: no evidence it works.

If you like, tomorrow, when I'm more cogent, I can dig up peer reviewed studies. But only if there is a reasonable expectation that you'll read them. Argument by citation dumps is pretty weak, even for a message board I'd like to avoid doing that. I do promise that if you (or anyone in this thread) have papers supporting your view, I will read them if they are accessible by me.

 

[Bodhi Dharma Zen]

That's hardly paranormal, nor even irrational.

Well, I was referring as TA being more woo than paranormal, and I believe it is irrational to believe that a derivative signal can give us a hint on the future price of the stock "if the momentum goes up the price will follow".

To a first approximation, the best predictor of the weather this afternoon is the weather this morning; the conditions that created bright sunshine today are unlikely to suddenly reverse themselves and produce a hurricane.

Continuation of similar conditions is not predicting the future. Sure, if it is a cloudy day you could say that there will be clouds for at least a brief period of time, but unless you have access to satellite data, you wouldn't know if the clouds are only above you and will clear in half an hour or would remain the whole day.

There's even more support for the idea that stocks will continue to do this afternoon what they did this morning, because of the "natural" tendency of people to see patterns and expect them to continue. This means that people will naturally expect rising stocks to continue to rise, and therefore will buy into them and raise the price. Again, you will probably be right more often than you will be wrong if you simply follow the crowd.

I don't see how this follows. Hope? When some individuals are seeing a continuation of a rally, others are seeing the appropriate time to short because it is a confirmation of a bear market... Now this is not to say that the price goes indeed up when people is hoping it will continue to rise, but as in the case of local weather, unless you have access to the satellite data, you are blind in to how many people is on the bull side and how many on the bear side... oh yes.. and there is no satellite image for the market... all we have is the price.

The problem is that being right 51% of the time isn't very useful when your losses are, numerically, twice what your winnings are. Especially if you pay commission charges to play, which both reduce your winnings and increase your losses.

Well, you can be right about 40% of the time (as I said before, I don't believe I can predict the future) but if your earnings are twice your losses... you will make money.

So the question becomes -- and I think this is a quite legitimate question in both psychology and economics -- when is it safe to trust to the wisdom of crowds and when is it unsafe? I've seen some very interesting theories that try to apply theories of judgement to stock prices (with mixed results, as you might expect) and a whole bunch of voodoo from technical analysts. What I haven't seen are any reliable results that will beat the transaction costs.

Yes, this is the conundrum. Gauge the masses opinions...

Of course not. My father-in-law made money this past week in Atlantic CIty by following a methodology that was based on "betting on black." But a lot of people at the same table actively lost money following a methodology based on "red," and those same people might well have lost money on "black" the next day. Lady Luck has a notoriously poor memory.

Yes indeed, it can be called "luck". I have no problems with that.

 

[Bodhi Dharma Zen]

You've made claims that TA works. I'm asking for the peer reviewed studies. I can provide plenty of citatations that prove the contrary.

Lol. No, I didn't. I said that a proper trading methodology (in addition to some TA) will make money. It is different.

 

[roger]

Lol. No, I didn't. I said that a proper trading methodology (in addition to some TA) will make money. It is different.

"this is not to say that you can't make money on the market by following a methodology that is based (to a point) on TA. "

I'm not going to get into semantics, and I'll certainly accept your wording above.

Again, cite? In other words, show me that addition of TA is a positive expectation event. Why else would you add it? Your claim reduces to the claim that TA works (not works alone, granted, but works).

Also, if by "make money" you mean end up with more money you started with, then I grant that without hesitation. That is no big trick. However, that is not how we measure investment strategies. The best measure devised compare risk-adjusted returns to the underlying markets. I'm unhappy with the assumptions "risk-adjusted" are based on, but that is another topic. I'm pretty much happy with a comparision against a passive investment in an index fund over several decades as a starter - after that we can look at ways of adjusting for risk.

So, I'd honestly like to see the evidence. And, btw, this is not a Clausian "evidence" grilling, evidence to the contrary . I'm genuinely interested in the topic and will actively seek out any studies mentioned. It is my somewhat studied understanding that no such study exists. I'd like to be corrected if I'm wrong. It is also very much my understanding that the intellectual underpinnings of TA are impossible to realize in the real world. For that I'll refer to pages 651-654 of the 1951 edition of Security Analysis.

Against that I'll mention that I'm aware that Russian PhDs (et al) are very well employeed by Wall Street devising algorithms for programmed trading. Naturally, this IP is very well protected by the firms, so the publication data is thin. I'm extremely curious in what they do. My general impression is that what they do works until it doesn't, that it is relatively high risk behavior (after all, they work for companies rewarding quarterly results, not 10 year risk adjusted returns, and we've seen how many blow ups that engenders), and that they certainly aren't doing the chartists type things implied by TA. I note you don't think that future prices can be predicted from back data, so I wonder what you mean by TA.

There was an interesting paper cited by Francecsa (IIRC) awhile ago that I'm failing to find at the moment. The upshot was that a winning trading strategy can be backtested, and then subsequently make money. The rub? 1) there is no way to predict when it will fail; 2) there is no way to analyze current performance to predict whether you are starting to fiail or not; 3) there is no general way to form a new strategy; 4) past failures provide no information for future failues. Well, I can make identical claims for my coin flip prediction technique. My current strategy is to choose "heads" whenever my dog licks it's lips. It hasn't failed for the last 4 flips! Once it starts failing regularly, I'll search for some new scheme. I'm thinking sneezes is a good avenue for exploration. Yes, I'm being flip, but what value is that kind of strategy? "I don't know why it works, I don't know when it works, I don't know if it is working now, I don't know when to stop using it (except to notice I've lost more money than I've made), I don't know how to create a new strategy, and I don't know how to learn from the past events.

 

[drkitten]

I pointed out that you can check the "charts" to pick your buy in point for a stock you already plan on buying.

I can also check tea leaves to pick my buy in point for a stock I plan on buying,.

Or tarot cards.

Or the flight of birds, or the entrails of sheep.

Do you have any evidence that the performance of the first method is distinguishable from the performance of the other four?

 

[drkitten]

Continuation of similar conditions is not predicting the future.

The hell it isn't.

Sure, if it is a cloudy day you could say that there will be clouds for at least a brief period of time, but unless you have access to satellite data, you wouldn't know if the clouds are only above you and will clear in half an hour or would remain the whole day.

Empirically, in most parts of the world, it's a better bet that they will remain the whole day. Weather systems are big, and the visible part of the sky is small. Which is why predicting that the afternoon will be much like the morning is a pretty good prediction strategy in the first place.

I don't see how this follows. Hope? When some individuals are seeing a continuation of a rally, others are seeing the appropriate time to short because it is a confirmation of a bear market...

There are fewer contrarians out there than there are momentarians. Again, this is an empirical question, but it's also implicit in the notion of "contrarian"; if most people were contrarians, then we would have picked a different word, wouldn't we?

 

[Bodhi Dharma Zen]

Last I checked, TA was one of many ways of analyzing the potential gains/losses of a stock. It is just another tool in the toolkit. Nothing more, nothing less.

It is not something to get dogmatic about. For example, on one of the stock forums I read, I noticed that there was a poster good at picking beatdown stocks that were ready to rise again. One of the other posters noticed that this person kept buying in at times where his profits weren't optimal and recommended using TA as a tool for that person to find his buy-in point.

Exactly. It gives you a hint on when to act. Nothing more. In fact, there are no "magical" indicators. I can make a good trading system based on quite a few of them. MACD, RSI, Momentum, you name it, all of them are equally incapable of giving you entry points. They might not be statistically different as a voodoo selected one, but what matters is that it gives some people the confidence they need to hit the trigger.

Now, the only way to appropriately use an indicator (actually a strategy based on several of them) is to make a backtesting against different kind of markets (bearish, bullish and sidelines), then using it in the real market with carefully selected stocks (just with enough money to be able to buy a handfull of them), then compare the backtesting with the actual data and use a probabilities based scenario working with stuff like the Montecarlo Simulation... and ONLY THEN deciding if you are going to put big bucks on it. And yes, even if all of this works nothing can guarantee that you will make money out of the market.

 

[roger]

Evidence?

 

[Bodhi Dharma Zen]

I'm not going to get into semantics, and I'll certainly accept your wording above.

Then it is clear. Good.

Again, cite? In other words, show me that addition of TA is a positive expectation event. Why else would you add it? Your claim reduces to the claim that TA works (not works alone, granted, but works).

Cite? Why would I want to read a book when I have money in the markets, I have developed a scientific based strategy and I'm making money on a consistent basis? I do not claim that TA "works", I only say that it gives you a reason to pull the trigger. I have extensively researched several indicators and strategies based on many of them. I claim that you can't predict the future movement of a stock based on a signal that is calculated from the price. Still, a good strategy will try to cash in to some particular movements you are interested in.

To put in on other words, I don't care if the market goes up or down, all I care is that the movement will be big. I attempt to profit from fear and greed and I can do that with about 40% of my trades being in the right direction.

Also, if by "make money" you mean end up with more money you started with, then I grant that without hesitation. That is no big trick. However, that is not how we measure investment strategies. The best measure devised compare risk-adjusted returns to the underlying markets. I'm unhappy with the assumptions "risk-adjusted" are based on, but that is another topic.

There you go, you answered your own question. I also don't like risk adjusted measurements as they take for granted that past trades are a good way to measure future trades.

So, let me put it in other words.. if my equity runs up independently of the market going up or down.. would you say that my strategy is sound? I can answer that... yes... while it works.. will I get money forever? I do not know, I CAN'T know.

I'm pretty much happy with a comparision against a passive investment in an index fund over several decades as a starter - after that we can look at ways of adjusting for risk.

Well, the decades part is an issue. All I need is a system that works for three kind of markets (arbitrary definitions of course) a bear market, a bull market and a sidelines one... Now, the system that is working for me right now is useless in a choppy market, but as soon as volatilty expands so my equity

So, I'd honestly like to see the evidence. And, btw, this is not a Clausian "evidence" grilling, evidence to the contrary . I'm genuinely interested in the topic and will actively seek out any studies mentioned. It is my somewhat studied understanding that no such study exists. I'd like to be corrected if I'm wrong. It is also very much my understanding that the intellectual underpinnings of TA are impossible to realize in the real world. For that I'll refer to pages 651-654 of the 1951 edition of Security Analysis.

I can see that you are genuinely interested. The sad part is that I won't discuss any further my strategy or specific techniques, and of course you are free to think that I'm making all this up.

Against that I'll mention that I'm aware that Russian PhDs (et al) are very well employeed by Wall Street devising algorithms for programmed trading. Naturally, this IP is very well protected by the firms, so the publication data is thin. I'm extremely curious in what they do. My general impression is that what they do works until it doesn't, that it is relatively high risk behavior (after all, they work for companies rewarding quarterly results, not 10 year risk adjusted returns, and we've seen how many blow ups that engenders), and that they certainly aren't doing the chartists type things implied by TA.

Yes, I believe you are right. 1) people making money will not tell you how they do it. 2) yes you can find algorithms that will work, at least for a period of time 3) yes they only work until they dont (but you can always design another trading strategy 4) yes it is risky but what is not 5) and yes they are not seeing the typical "magical" indicators that you can see in books about "how to get rich in the market".

I note you don't think that future prices can be predicted from back data, so I wonder what you mean by TA.

Only what I said. It gives you a reason to enter a stock at a determinate price. Sometimes the direction will be good for you, sometimes it wont.

There was an interesting paper cited by Francecsa (IIRC) awhile ago that I'm failing to find at the moment. The upshot was that a winning trading strategy can be backtested, and then subsequently make money. The rub? 1) there is no way to predict when it will fail; 2) there is no way to analyze current performance to predict whether you are starting to fiail or not; 3) there is no general way to form a new strategy; 4) past failures provide no information for future failues. Well, I can make identical claims for my coin flip prediction technique. My current strategy is to choose "heads" whenever my dog licks it's lips. It hasn't failed for the last 4 flips! Once it starts failing regularly, I'll search for some new scheme. I'm thinking sneezes is a good avenue for exploration. Yes, I'm being flip, but what value is that kind of strategy? "I don't know why it works, I don't know when it works, I don't know if it is working now, I don't know when to stop using it (except to notice I've lost more money than I've made), I don't know how to create a new strategy, and I don't know how to learn from the past events.

I disagree with 2 and 3.

 

[roger]

I disagree with 3. You can always make new strategies. I mostly agree with everything else, except for the "I do not know when to stop using it". If your strategy is sound, you will have a calculation on when your system is no longer working.

My comments on this referred specifically to the paper, not anything you or anyone else might be doing. The point was there was no system, no method of analysis to find a new strategy for the method described in the paper. Just make one up, back test it, use it until you decide it's not working (which can't be defined, given the constraints). But this is kind of a pointless discussion, as I can't provide the paper at the moment.

As for the rest, I don't know what to say. I respect your right not to share IP. I will note that this is how it has gone every single time I have ever discussed these things on this or any other forum. Never any citations. Never any evidence. In any individual instance I grant the poster's right not to post data, though I reserve the right to point out that they have provided zero evidence, and will strenuously object if they end up giving financial advice to others. But it's the sum of all those actions that interests me. We have ample studies showing no positive expectation over appropriately long time periods, balanced against claims of efficiacy that quickly fade away (every time!) when the light of reason is shone on them.

I never have that problem with proponents of efficient markets, or proponents of value investing, or proponents of some sort of index investing. Citations are readily proffered. I'll happily share with you any of my thoughts on investing. Not individual stocks, I'm not licensed, and don't give stock tips, but IP? It's all yours for the asking. Backed up with citations to the literature, if necessary. Of course, I am a nobody, but you will see the same behavior by respected value investors: Graham, Dodd, Greenblatt, Schloss, Fischer, Eveillard, Berkowitz, Buffet, Munger. You will see the same by EMT proponents and others: Fama, Malkiel, Bogle, etc., etc., etc. But no TA folks. Strange, huh?

Here's a link to a paper cited by Wikipedia as positive evidence for simple TA rules. They find some rules that generate excessive returns on back tested data. Cool! Except, read it. The rules stop working: "However, we also find that the best technical trading rule does not provide
superior performance when used to trade in the subsequent 10-year post-sample period."

It's classic data mining, IMO. You can always find some algorithm to fit data. It's guaranteed (I leave DrKitten to provide the math ) That is in no way evidence that something works. Furthermore, when it stops working just as soon as new data starts coming in, well, you ain't predicting the market, you are data mining.

The conclusion is particularly laughable. They note that the rules they developed in fact do not work beyond the data set they used to test the rules, but then hypothesize this could be due to nonrepresentative data during the post-sample period, or that the markets have become more efficient. They didn't even mention the possibility that they are ex post facto fitting algorithms to data. I had this problem in my first job with a scientist at NIH, where he kept changing the rules until he got some data hits on cancer incidences and mortality in US Hispanic populations. No hit on this idea? Well, look at just Cuban Americans. No hit there. How about the ones ages 25-34. No hit? How bout 25-34 female Cuban Americans. A hit! Yay! Let's publish! (and let's ignore we are now looking at 6 data elements). Back testing data is not evidence. It's a suggestion.

That's just one study, but a routinely quoted one.

It sure is interesting that despite the number of claims of efficacy, no one can say "roger, this is the way to do it" or "roger, here is a paper with no special pleading".

edit: thank you for an enjoyable conversation.

 

[Bodhi Dharma Zen]

Back testing data is not evidence.

Indeed, as I have said. You need a lot more, for instance, the steps I proposed above. And even then you do not have a clue on if its going to end nor when.

The only other hint I´m willing to let go is that I do not care about if the price is going up or down, in fact, I don't even use a time series (price against time) as an input. There are other ways to make sense of the data. And, again, the same as you understand why I will not discuss any further I understand your feelings.

 

[jimbob]

The only problem with that is that I have no confidence that any theory of TA with which I'm familiar would be any better at finding a buy-in point than random guessing.

And if it did work better, and was sufficiently reliable, then I'd be even better off trying to predict where the TA as opposed to the market was going next.

A simplified example:

Devil's Advocate: "
If a stock shows a particular pattern of behaviour, its price will rise by 5%.

If everyone believes this rule, then it almost certainly will, but that "knowledge" will then be factored into the price.


"

 

[Bodhi Dharma Zen]

If a stock shows a particular pattern of behaviour, its price will rise by 5%.

If everyone believes this rule, then it almost certainly will, but that "knowledge" will then be factored into the price.

Allow me to complete the scheme:

If, and only if, a sufficient number of participants is aware of that particular behavior. What some people neglect to see is that there are countless of possible trading systems built on signals. Countless. The odds about someone else developing the same one are close to zero.

This is, if somebody writes a simple system ("generate a buy at next bar if momentum from previous bar is higher than two bars ago"), we could enter this system on 1 minute charts, or 5 minute charts, or daily weekly or anual graphs... We can choose a trailing stop loss, a monetary based stop loss, a percentage based. We can implement money management techniques, we can average our entry price by buying by the same rule without exiting for any number of trades, we can apply it only to stocks from a particular sector, or determinate historical volatility, and so on.

How on earth would your friendly neighborhoods systems will equate yours, thus canceling each other on the market?

 

[roger]

You are right. If people are just doing random ****, then yes, they'll more or less cancel each other out.

But we are not positing random ****, we are positing a system that actually, demonstrably works.

Furthermore, if the # of systems is nearly infinite, what are the odds of finding a system that actually, demonstrably works?

I leave the latter as an exercise for the reader.

 

[jimbob]

BDZ:

I think you are missing my point. (My hypothetical example hypothetically worked, ...at first)

If a system works and becomes known to work, it will become part of the information that will be factored into the marke price.

If a system is known to be popular it could be worth investigating, in order to speculate against someone using the system.

A large numbet of people use system x and so if they see a particular pattern of trades, they will buy at a price which I can now calculate. They don't know what I will accept, but I know what they would give.

 

[roger]

From Security Analysis, 1951 edition, page 652:

1.Chart Reading Not a Science That chart reading cannot be a science is clearly demonstrable. If it were a science, its conclusions would be as a rule dependable. In that case everybody could anticipate tomorrow's or next week's price changes, and hence everyone could make money continuously by buying and selling at the right time. This is patently impossible. A moment's thought will show that there can be no such thing as a scientific prediction of economic events under human control The very "dependability" of such a prediction will cause human actions that will invalidate it.... It follows that there is no generally known method of chart reading that has been continuously successful for a long period of time. If it were known, it would be speedily adopted by numberless traders. This very following would bring its usefulness to an end.

I don't think anything has changed in the ensuing 60 years to change Graham's or jimbob's conclusions.