Stock Market Technical Analysis

[roger]

Well, what's laughable is it is 56 "positive results" with bad methodology, with no accounting for the file drawer effect. Research via wikipedia at it's best!

 

[corplinx]

Well, what's laughable is it is 56 "positive results" with bad methodology, with no accounting for the file drawer effect. Research via wikipedia at it's best!

You just broke rule 47: attack the argument, not the Wikipedia.

Any argument of yours that plotting price volatility on a chart makes price volatility worthless will continue to be dismissed out of hand as silliness. Sorry. I'm not going to entertain silly notions.

 

[roger]

You just broke rule 47: attack the argument, not the Wikipedia.

Any argument of yours that plotting price volatility on a chart makes price volatility worthless will continue to be dismissed out of hand as silliness. Sorry. I'm not going to entertain silly notions.

I have been attacking the argument, though really it'd really have to be called attacking the assertions in this case. Also, I find it odd that you think pointing out the flawed data in Wikipedia is somehow wrong.

As for your second paragraph, I can't parse what you mean. I don't recognize anything I said in it. Again, I find it odd that you refuse to even read research papers by leaders in the field. It's you vs people short listed for the Nobel et. al. published in the most prestigious economic journals.

Eugene Fama on price volatility:
http://www.e-m-h.org/Fama70.pdf

Paper on the flawed methodology on the 56 papers showing positive results (you have to pay for this one):
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1006275

Very extensive article that actually found some tiny positive results, given unrealistic assumtions about transaction costs (free for the downloading):
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=566882

Oh, I know, Journal of Finance. Silly. Not worth entertaining. Eugene Fama, short listed many times for the Nobel, research partners and co-author with winners of the Nobel. Just another world renowned, silly person who disagrees with you. They waste their time giving evidence for their arguments.

 

[bpesta22]

In particular, 56/95 is not statistically significant or distinguishable from "no result at all." (The actual p value is 0.0501 for a one-sided test (which is not justifiable) or 0.1002 for a two-sided test (which is justifiable).

That's elementary probability theory.

If I tried to publish a paper with a p value greater than 0.10, I'd get laughed at. I guess corplinx has less critical standards.

I'm not sure the probability calculation is as simple as what you imply here. If there really were no effect of TA on stock price picking (the null hypothesis) then getting 56/95 positive results would be highly surprising and would certainly lead one to reject the null (i.e., claim there is an effect).

If every individual study used the standard .05 level of significance, then finding an effect by chance alone would happen only 1 in every 20 studies. Here we have a rate much much higher than that.

So, something's going on; an effect exists.

The key though are other posts here that claim the studies are flawed and some file-drawered. It seems like an internal validity issue (some confound is explaining the significant results; not that TA works).

But again, I would think 56/96 would be highly significant, unless I am really mis-understanding how you evaluated it.

 

[corplinx]

Oh, I know, Journal of Finance. Silly. Not worth entertaining. Eugene Fama, short listed many times for the Nobel, research partners and co-author with winners of the Nobel. Just another world renowned, silly person who disagrees with you. They waste their time giving evidence for their arguments.

I am not sure if they disagree with me. I do not accept that technical analysis is a working tool for "picking" stocks. You go on ignore now for trying to pin that position on me again after being warned several times. Bye bye.

 

[drkitten]

If every individual study used the standard .05 level of significance,

That's not an assumption I'm willing to grant.

 

[roger]

I am not sure if they disagree with me. I do not accept that technical analysis is a working tool for "picking" stocks. You go on ignore now for trying to pin that position on me again after being warned several times. Bye bye.

I explained that "stock picking" was a miswording on my part. None of the papers linked involve using TA for stock picking, but for price prediction. Not only do you refuse to read the linked research, but you don't seem to read my posts either.

edit: for clarifications to other readers, not the person who is ignoring evidence: I think Corplinx misunderstood me, having just reread his #84 post. My statement was not that TA was used for stock picking, it was that stock pickers use TA. In fact, that was my wording "a stock picker using TA". Meaning someone who picks stocks, and then uses TA for pricing. I can see why that wording would be confusing and/or ambiguous. However, I explained to him earlier that I was not talking about using TA to pick stocks; in fact, I'm not aware of any use of TA for actual stock selection. Anyway, I find it odd that someone chooses to ignore somebody over a misunderstanding, and refuse to read the best scholars in the field.

 

[roger]

The key though are other posts here that claim the studies are flawed and some file-drawered. It seems like an internal validity issue (some confound is explaining the significant results; not that TA works).

To be fair, we don't really know which direction the file drawer effect would go. TA is hardly popular in academia, so one could imagine positive results getting file drawered by researchers looking to show that it is flawed. Who knows?

 

[egslim]

Correction. This is not roulette and we aren't betting on red because its been black 9 times in a row. I used to work at the HQ for the worlds' largest casino company. I know how that fallacy works.

Stock prices aren't magical nor the result of die throws. They are based on what people offer to sell them for and what other people bid to buy them for.

Indeed. So let's consider those stocktraders.

For every given stock there are many traders, they're different people almost every day, you don't know any of them, and like all humans they're often unpredictable even to those who do know them casually.

This isn't statistical physics, where all particles behave the same and obey relatively simple rules, so you can make statistical predictions. Each stocktrader is a complex individual.

With all those unknowns about the people who determine a stockprice, there's no rational reason to assume short term price fluctuations can be predicted. The irrational reason to believe that is because the human brain is wired to recognise patterns, and to make them up if they don't exist.

You're aware of that fallacy when related to gambling, but the exact same principle is at work with TA.

 

[Tippit]

Indeed. So let's consider those stocktraders.

For every given stock there are many traders, they're different people almost every day, you don't know any of them, and like all humans they're often unpredictable even to those who do know them casually.

This isn't statistical physics, where all particles behave the same and obey relatively simple rules, so you can make statistical predictions. Each stocktrader is a complex individual.

With all those unknowns about the people who determine a stockprice, there's no rational reason to assume short term price fluctuations can be predicted. The irrational reason to believe that is because the human brain is wired to recognise patterns, and to make them up if they don't exist.

You're aware of that fallacy when related to gambling, but the exact same principle is at work with TA.

I must say I agree with this logic. Having said that, how do you value commodities or other assets without cash flows that we can compare with other, similar assets? At some point in our forecasting of the future we have to quantitatively identify trends, and how do we do this without some form of price/volume analysis?

 

[Francesca R]

I must say I agree with this logic. Having said that, how do you value commodities or other assets without cash flows that we can compare with other, similar assets?

I do not think of commodities as being assets, and they are not capable of being valued using more than informed guesses that do well if they predict the future better than average. There is no case for holding commodities passively in any portfolio. For actively trading them, knock yourself out with any technique you like--the notion of fundamental analysis with these prices is tenuous.

 

[roger]

I must say I agree with this logic. Having said that, how do you value commodities or other assets without cash flows that we can compare with other, similar assets? At some point in our forecasting of the future we have to quantitatively identify trends, and how do we do this without some form of price/volume analysis?

There's no shame in saying "I don't know" at some point. People endlessly try to forecast the future, yet there is no real evidence that they can. I am speaking in broad terms - the talking heads on CNBC, Barron's articles, etc. There's very slight evidence that technical trading rules can produce slightly positive returns, for a short period of time, for some commodities. But there is no intellectual basis for these rules, I have never seen a paper that tells you how to tell when your rule has changed (other than the obvious: you lose your shirt) so it is not longer profitable. So, I dispute your " have to quantitatively identify trends" phrase. We don't have to. Some things we just dont know, and can't know. There's nothing wrong with concluding that.

 

[egslim]

I must say I agree with this logic. Having said that, how do you value commodities or other assets without cash flows that we can compare with other, similar assets? At some point in our forecasting of the future we have to quantitatively identify trends, and how do we do this without some form of price/volume analysis?

Stocks can be valued to some extent by fundamental analysis. However, it must be recognized a large margin of error remains due to our fundamental inability to predict the future. We can compensate for that by introducing a margin of safety.

In theory commodities can be valued if we know their supply and demand-curves. In practice we generally don't. Though I expect the price of oil to rise over the next 20 years, as the most accessible supplies run out and demand continues to increase from developing nations. However, the margin of error for estimating the value of a commodity seems too large to compensate for.

If the stockmarket is like a casino where you know the longterm-average odds are in favour of the players, the commodity-market is a place where the odds are unknown. Which is why investors shouldn't go there, and leave the place to speculators.

 

[roger]

Stocks can be valued to some extent by fundamental analysis. However, it must be recognized a large margin of error remains due to our fundamental inability to predict the future. We can compensate for that by introducing a margin of safety.

In theory commodities can be valued if we know their supply and demand-curves. In practice we generally don't. Though I expect the price of oil to rise over the next 20 years, as the most accessible supplies run out and demand continues to increase from developing nations. However, the margin of error for estimating the value of a commodity seems too large to compensate for.

If the stockmarket is like a casino where you know the longterm-average odds are in favour of the players, the commodity-market is a place where the odds are unknown. Which is why investors shouldn't go there, and leave the place to speculators.

That's a good point, and makes me want to revise my last post, which was written with more short-term movements in mind. In the very long term, general trends can probably be predicted. Oil prices will almost certainly go up as supply dwindles (short of a technological revolution), the stock market of a stable country as a whole will go up, etc, but these aren't the kinds of price movements that are generally "predicted". And there are shorter trends that I think can be predicted - for example, last year it was pretty apparent that the crack spread was going to contract because we have very accurate #s on the number of refiners on and offline, schedules for when repairs will be completed, etc., and we knew that a bunch of facilities would be coming online. Not foolproof info, as conditions can change, but reasonably reliable. But even there you have to take it with a huge grain of salt as you can usually find "experts" coming down on both sides of an issue. IIRC, though, the crack spread contraction was pretty much universally expected.

How that plays out in stock prices, I don't really know or understand. For example, I thought Buffett was crazy for buying COP when he did. Sure enough, the stock prices went down after he bought as in fact the crack spread did contract. OTOH, if it was so obvious, why wasn't this knowledge already priced into COP? Well, partly I don't believe the market is oh-so-efficient. And then I could be wrong about how likely it was that the crack spread would contract. But, the data on refinary outages is very reliable. My gut feeling is that the whole system is too complex to predict - too many differently motivated parties. Oil funds that have to buy COP when new investors come in, people buying to hold for 20 years, people trying to buy for short term price swings, buyers on rumors, etc. Trying to predict all of that to produce short term price predictions doesn't strike me as a plausible act.

 

[egslim]

And there are shorter trends that I think can be predicted - for example, last year it was pretty apparent that the crack spread was going to contract because we have very accurate #s on the number of refiners on and offline, schedules for when repairs will be completed, etc., and we knew that a bunch of facilities would be coming online. Not foolproof info, as conditions can change, but reasonably reliable. But even there you have to take it with a huge grain of salt as you can usually find "experts" coming down on both sides of an issue. IIRC, though, the crack spread contraction was pretty much universally expected.

On the other hand, obvious short term trends may already be priced, or even overpriced in. For example, a few weeks ago oil storage facilities in Rotterdam were maxed out, and tankers were waiting outside. One analyst expected (and bet on) a drop in price, but it didn't happen.

Because commodities are practically impossible to value in absolute terms, all you have to go on are relative changes. And if the commodity has been overpriced for a long time, even a significant pricedrop may be insufficient to make it an attractive buy.

It may be possible to make some money predicting short term trends if you're well-informed enough, and if you limit yourself to the most obvious trends. But I also think it's much more difficult, time-consuming and less profitable than some other forms of investing.

How that plays out in stock prices, I don't really know or understand. For example, I thought Buffett was crazy for buying COP when he did. Sure enough, the stock prices went down after he bought as in fact the crack spread did contract.

I think Buffett didn't care about the crack spread, he looked at the P/E of about 4. I'm pretty sure Buffett pays little attention to short trends, he's in it for the long term. If the price makes sense, that's enough for him.

Well, partly I don't believe the market is oh-so-efficient.

Neither do I.

But, the data on refinary outages is very reliable. My gut feeling is that the whole system is too complex to predict - too many differently motivated parties. Oil funds that have to buy COP when new investors come in, people buying to hold for 20 years, people trying to buy for short term price swings, buyers on rumors, etc. Trying to predict all of that to produce short term price predictions doesn't strike me as a plausible act.

It may be possible - but even if it is, why bother? There's much lower hanging fruit available.

 

[roger]

I think Buffett didn't care about the crack spread, he looked at the P/E of about 4. I'm pretty sure Buffett pays little attention to short trends, he's in it for the long term. If the price makes sense, that's enough for him.

Well, he is on record as considering it a mistake to buy then.

I told you in an earlier part of this report that last year I made a major mistake of commission (and maybe more; this one sticks out). Without urging from Charlie or anyone else, I bought a large amount of ConocoPhillips stock when oil and gas prices were near their peak. I in no way anticipated the dramatic fall in energy prices that occurred in the last half of the year. I still believe the odds are good that oil sells far higher in the future than the current $40-$50 price. But so far I have been dead wrong. Even if prices should rise, moreover, the terrible timing of my purchase has cost Berkshire several billion dollars.

The PE was more in the 9-10 range then wasn't it, with oil at an all time high?

I guess my working hypothesis on all this is yes, it is pretty hard, if not impossible, to predict prices for refiners, but the data on capacity coming on line was very solid, and the odds were pretty much against you at that point. There were less risky places for him to put his money. Yes, I know, hindsight, but I argued the same thing on a financial board about refiners last summer. If you know there's going to be a huge surplus in fertilizer next year, is it wise to buy a fertilizer business today? Was it wise to buy a homebuilder last September? Etc. You can't say with certainty where the price is going, but it sure appears that things can turn against you pretty easily with any of those purchases, long term holding or not.

Or so goes my still forming thoughts on this topic.

 

[egslim]

Well, he is on record as considering it a mistake to buy then.
The PE was more in the 9-10 range then wasn't it, with oil at an all time high?

Could be, I got my information here: http://seekingalpha.com/article/106310-buffett-increases-stake-in-cop-smart-move

I guess my working hypothesis on all this is yes, it is pretty hard, if not impossible, to predict prices for refiners, but the data on capacity coming on line was very solid, and the odds were pretty much against you at that point.

True. So the lesson is that if you focus on the long term, it pays to consider likely short term effects as well. Especially if the company you consider buying isn't already in the bargain-basement. Nobody said intelligent investing was easy.

ETA: The more a company is underpriced, the less its shareprice is likely to be affected by further short-term bad news.

 

[corplinx]

Indeed. So let's consider those stocktraders.

For every given stock there are many traders, they're different people almost every day, you don't know any of them, and like all humans they're often unpredictable even to those who do know them casually.

This isn't statistical physics, where all particles behave the same and obey relatively simple rules, so you can make statistical predictions. Each stocktrader is a complex individual.

With all those unknowns about the people who determine a stockprice, there's no rational reason to assume short term price fluctuations can be predicted. The irrational reason to believe that is because the human brain is wired to recognise patterns, and to make them up if they don't exist.

You're aware of that fallacy when related to gambling, but the exact same principle is at work with TA.

No its not. Imagine the roulette table is missing 0 (and 00 for fun since those are common as well). The probability of red/black is 50/50 and is unrelated to all previous spins.

The stock market is not a die throw. It is a market. People buy and sell. There are many factors that make it too random to model. If you trend a roulette tables spins you cannot use the information to make a prediction on future spins (assuming a perfect table with perfect dealers).

Let's pretend you want a 1995 Nissan mini-truck. You research and find that over the past year they have sold for between 1500 and 2500 dollars. You decide that if you buy that truck, you would rather pay closer to the minimum price than the maximum. Now, if you put the price volatility of that truck's listing prices and sell prices on a chart, its somehow flawed data.

Again, as the studies show, TA is not a panacea for _picking_ stocks. However, ignoring technicals for information about stock pricing for picks you have already made is just silly. Shrewd buyers want to buy things that are bargains and they will pass on a stock if it doesn't meet their entry point.

 

[roger]

Let's pretend you want a 1995 Nissan mini-truck. You research and find that over the past year they have sold for between 1500 and 2500 dollars. You decide that if you buy that truck, you would rather pay closer to the minimum price than the maximum. Now, if you put the price volatility of that truck's listing prices and sell prices on a chart, its somehow flawed data.

Again, as the studies show, TA is not a panacea for _picking_ stocks. However, ignoring technicals for information about stock pricing for picks you have already made is just silly. Shrewd buyers want to buy things that are bargains and they will pass on a stock if it doesn't meet their entry point.

You have me on ignore, but for the other readers...

1. studies show that in fact, your assumption in the first quoted paragraph is wrong. Read the studies I linked.

2. You have made a wrong assumption about the studies. They are not about picking stocks, they are about price predictions.

Read. Learn. Please. Stop assuming things based on poor wording on my part, where I said "stock picking" when I meant price prediction.

 

[egslim]

Let's pretend you want a 1995 Nissan mini-truck. You research and find that over the past year they have sold for between 1500 and 2500 dollars. You decide that if you buy that truck, you would rather pay closer to the minimum price than the maximum. Now, if you put the price volatility of that truck's listing prices and sell prices on a chart, its somehow flawed data.

Your line of reasoning is valid for almost every market in existance, except the stockmarket. That's because the stockmarket is subject to a unique peculiarity, compared to other markets.

In normal markets, like the one for 1995 Nissan mini-trucks, an increase in price leads to a drop in demand and vice versa. This leads to the famous supply and demand curves, which set the market price. The important thing is that this inverse relation between price and demand stabilises the market price. That stability is why your example for mini-trucks works.

The stockmarket is fundamentally different, because an increase in price actually leads to higher demand, and vice versa. This is a positive feedback loop, it's why the marketprice for a stock is unstable. And without stability from a negative feedback loop future trends can't be predicted from previous data.