TA doesn't work?

[DesertCat]

Read some interesting posts by a reformed TA guy and thought I'd share them here to see what others thought. His basic point is the only traders making money over long periods of time aren't using any kind of TA, they are using statistical arbitrage or some kind of similar techniques.

[DesertCat]

[ QUOTE ]
> This isn't the place to debate fundamental and technical analysis.
> People much more erudite, and far richer, cannot agree. I won't
> presume to represent a viewpoint as I simply don't have the
background
> to do so.

Well, I think I do and I feel pretty confident telling you that TA
doesn't work. It's good for booksellers, software vendors and
brokers.

There are some things that might be useful to people (like stop
limits and buy stops can be used in some situations), but these are
not really TA.

I used to work for a very famous hedge fund. Most people outside of
the fund think they make money with TA. It's true, they use a lot
of charts, but I've been right there in his office when many trading
decisions were made, and let me tell you, there is a lot of
information and input that goes into the decisions. It's almost
never "oh man, look at that head and shoulders top! We've gotta
sell the neckline!".

Early on in my career, I also sent out a technical newsletter to
clients. I believed what I was doing and I thought I was right a
lot of the time, but my primary motivation was commission
generation. I wanted to find action points to motivate people to
trade. I see a stock break out to the upside to a new high, I would
send out an alert "You must buy this break out!" etc...

Of course I wanted to be right too, but if I was wrong, I was not
too concerned (because it was not my p/l).

If you think too hard and too deeply, then you will not generate too
many ideas every day, and if you don't do that, you won't generate
commissions, and if you don't do that, you will be fired.

Anyway, I've met many, many technicians over the years and it's
absolutely true what they say, most of them are not rich. The ones
that are rich got rich by being technicians at major brokerage
houses or by selling books.

I can't name a single one of those guys that made any serious money
in the markets. Of course, they all have their 'big trades' that
they like to brag about, but long term performance? None. Zip.

Look at funds, hedge funds and CTAs. How many with good long term
track records were created using TA? Not many. Some were done with
quantitative trading methods, but it's questionable whether it can
be considered TA...

Of course, John Henry, Richard Dennis and the Turtles (no, this is
not a rock band from the 60s) etc... are (or were) pretty well-known
technically oriented houses, but their track records are very
volatile and not very compelling (in my view). The blow-up rate of
these kinds of funds are pretty high.

Go to any of the major Wall Street firm's proprietary trading desks
and see how their money is made. They make money doing everything
from index arb, risk arb, convert arb, stat arb, volatility trading,
derivatives, special situations, tax arb, or whatever, but you will
never see someone trading off charts. (sure, many traders will look
at charts, but I've never seen anyone use that as their primary
decision making tool)

The fact is that none of the money these guys make come from charts-
reading.

Also, if you look at newsletters, how many newsletters with good
long-term track records were achieved using TA? Look at Joe
Granville's or Prechter's records.

I think one of the best long-term letters is Al Frank's Prudent
Speculator which did 20%/year for 20+ years (I'm not sure if it was
20, 25 or 30 years, but they do have a long, long record) and they
are 'valued/fundamentals' based.

So as much as I used to be into it (and I still look at charts from
time to time), and wanted to believe it works, the evidence is just
not there to support it.

Like I think I said before, there just is no Graham and Doddsville
in chartland. There isn't.

The major problem with chart analysis is that the books that teach
this stuff never show the stuff that doesn't work. They will show
you all the head and shoulders tops that lead to big declines in
prices, but they won't show you the h&s tops that didn't work!

Same with any of those mystical patterns.

It's like me publishing a book about politics. I can try to sell
you a 'system' that will allow you to guess the political
orientation of anybody just by looking at their shirt.

My hypothesis would be that if they are wearing a blue shirt, I will
claim that that person is a democrat.

I will go to the mall and ask every blue shirted person if they are
a democrat. If they say yes, I will take down their name and take a
picture of them.

I will do the same for people wearing non-blue shirts. If they
answer 'republican', I will take down their name and take a photo of
them.

In my book, I will state my thesis (or system), and then prove it by
stuffing the book with all the cases I found at the mall that proved
that my system worked.

You will see a hundred pictures of blue shirted people with the
caption "as my system had accurately predicted, this person is in
fact a democrat..."

Is this sound? Of course not. Does it matter? Not if your primary
objective is to sell books. To a non-thinking observer, this book
would be very compelling and convincing. It looks like it works
perfectly!

This is exactly how most of those TA books are written. I've spent
a lot of time talking people out of stupid trading strategies (even
at Wall Street firms)...

Many otherwise intelligent people get caught up in this at some
point... (but again, it's true that many serious investors/traders
have been known to look at charts from time to time even though
that's not how they make most of their money).

VE



[/ QUOTE ]

[DesertCat]

[ QUOTE ]

> I'm trying to understand here. If, as you state, low RSI points
>will coincide with bottoms in price, then buying at low RSI points
>should work to the same extent as a strategy of buying at price
>bottoms (or at least after price declines since nothing can call an
>exact bottom).

Well, this is the biggest problem with technical analysis as it is
applied: People look at big, major bottoms in stock prices and
notice that the RSI was below 30 or 20 or whatever every time.

So they convince themselves that the stock is near a bottom when the
RSI gets below 30 or 20.

This is one of those things that I've spent a lot of time testing.
We thought that if a stock is oversold with an RSI below 30, it was
even more oversold under 10.

As you can imagine, if you look at ALL instances, then the results
were random.

In fact, many people even developed trading systems based on BUYING
stocks/commods when RSI goes above 80 and SELL them when they go
below 20. Of course, this worked really well in strongly trending
markets.

Like many other of these simple systems, all one needs to do is to
predict whether the market/stock in the future will trend strongly,
or be stuck in a trading range. Then we can pick the appropriate
system to trade!

In the short term, market price movements are actually very close to
being random. Most of the tests I've done (with the best database
and the best software (and full time programmers) in the business at
the time) showed that to be true.

No matter what condition you looked at, price would be up or down
50% of the time on a daily, weekly or even monthly basis for most
stocks and commodities. Even if there was a divergence in it, the
number would be 51%, 52% or maybe even as high as 55 or 60% (but the
samples would be much smaller in these cases). Hardly anything to
be able to trade off of).

Now, this is not a refutation of TA or the RSI as an indicator, but
it is illustrative of the sort of self-deception or illogic that is
prevalent in the business. It is amazing how many intelligent
people don't understand this simple concept (of course, quants
understood this well):

The argument was:

Technical trader: "Man, this stock has been down for the last 10
days in a row! It's gotta bounce tommorow! What are the odds that
this stock can go down 11 days in a row!! Very low!!! This is a
great high probability trade!"

Me: "No, no, no. The odds of the stock being up is still 50%."

Technical trader: "No way, it's gotta be higher than that! What are
the odds that the stock could be down 11 days in a row?".

Me: "That's a different question and has nothing to do with what the
stock will do tommorow. The odds of a stock going up or down 50% is
not effected by the fact that the stock was down for the past 10
days... "

Technical trader: "Why not? The odds of a stock going down 11 days
in a row is highly unlikely, and if the stock declined tommorow,
that would be make it a highly inprobable event..."

Me (exasperated): "OK. What about if it was a coin. We flipped the
coin 10 times and it came up heads. What are the odds that the coin
will come up heads on the next flip?"

Technical trader: "Same. 11 heads in a row is highly unlikely so it
can't be 50%"

Me: "OK. never mind."

--

I actually had a conversation like this with a trader. It was
unbelievable. And yes, we did test the odds of a stock moving up or
down after all sorts of 'runs' (what happened after 10 up days in a
row etc...), and the results confirmed my point. It was usually
50/50 no matter what kind of conditions you put on it.

(by the way, the only time you get extreme results is when you have
very few data points. If you put so many conditions on, then it
reduces the number of occurences and hence the figure may deviate
from the 50% level. At that point, you don't have enough 'events'
to call your results 'significant').

Anyway, the above thinking, that a stock must move up since it went
down so much is at the heart of the RSI / Stochastics type
indicators. It might be a little different, but the basics of the
argument is exactly the same.

Another one of these silly indicators is the Bollinger band type
things with standard deviations. You put bands around a moving
average or something so that you can say when a price overshoots
outside of an average to 2 standard deviations whatever, "the stock
price has been in this area only 5% of the time! So it must revert
back to the mean at this point. You have a 95% chance of making
money on this trade!".

This suffers the same fallacy, because there are two ways that a
stock price can revert back to the moving average. The stock price
can come back down, or the moving average can move up and catch up
to it.

In these books, they will always show you the charts where the
argument looks compelling.

But this type of analysis also suffers from the same thing as the
RSI, stochastics. In strongly trending markets, you want to get
LONG when a stock price hits the upper standard deviation bands, and
sell or short when they hit the lower ones.

But then again, you have to know beforehand if it's going to trend
strongly or be stuck in a range.

That's why these things are useless. If you knew what you needed to
know to make these indicators useful, then you wouldn't need the
indicators!

If you have a sense of humor and a lot of patience, it's kind of fun
to work with these guys. (even though I've lost patience for this
nonsense long ago...)

> Anyway, I like to look at RSI (as one of many other factors)
>because I think buying on price dips is a good strategy. It may be
>pseudoscience but it gives me a signal to look at it besides just
>the line falling on the chart. I'd be open to other strategies for
>buying on price declines.

You can use these technical indicators for whatever you want, but
just don't expect it to mean anything. That's all. It's true that
it can be psychologically comforting (OK, let's ask Munger what he
thinks about doing something for psychological comfort at the
expense of logic and rationality!)

A good strategy for buying on price declines is to keep a
spreadsheet of intrinsic value of the companies you follow and a
threshold where you might be interested in buying it. And then
buying it when it gets to that price.

I think that's a great way to go.

But I would never buy something on an RSI or stochastic or anything
like that. I know it's psychologically appealing. When you see
that thing stuck down there under 20 for a long time, it's natural
to think "man, this stock's GOTTA bounce. It just HAS to!".

[/ QUOTE ]

[eastbay]

In general, this is false.

However, there's a lot of crap TA out there, and most people who think they know how to use it, don't. So the statement has some truth to it. It's just not absolute. TA can work. TA does work.

For rigorous evidence that TA can work, here is a good place to start:

http://www.pupress.princeton.edu/sample_chapters/lo/part1.html

eastbay

[Sniper]

Cat, there are no sure things, its all about putting the odds in your favor /images/graemlins/wink.gif

[DesertCat]

[ QUOTE ]

For rigorous evidence that TA can work, here is a good place to start:

http://www.pupress.princeton.edu/sample_chapters/lo/part1.html

eastbay

[/ QUOTE ]

To be honest, I couldn't understand the description of their test (I'm not a statistics major), so I can't comment on their proof.

But "Moreover, upon further investigation, we learned that over the past decade several investment firms--most notably, Morgan Stanley and D.E. Shaw--have been engaged in high-frequency equity trading strategies specifically designed to take advantage of the kind of patterns we uncovered in 1988. Previously known as "pairs trading" and now called "statistical arbitrage," these strategies have fared reasonably well until recently, and are now regarded as a very competitive and thin-margin business because of the proliferation of hedge funds engaged in these activities. "

Pair trades aren't TA, so that quote doesn't support it.

In general I disagree with the premise of "A Random Walk Down Wall Street", in that markets are always efficient. I think even Markiel has conceded that markets are just generally efficient, but not always.

But I believe the way to beat the market is by taking advantage of arbitrage opportunities and value discrepencies. I'm very skeptical that historical price data (TA) has any value in predicting future prices.

[crazy canuck]

To be honest, I couldn't understand the description of their test (I'm not a statistics major), so I can't comment on their proof.

Just take their word for it, they co-authored the Econometrics bible:

The Econometrics of Financial Markets (http://www.amazon.com/exec/obidos/tg/detail/-/0691043019/102-7529209-2720150?v=glance)

Also, from my experience RSI, stochastics, MA-s etc. have no predictive power (tho maybe some traders can generate signals from them using some fundamental info). I even implemented genetic programming that hasn't found anything (basically it evolves logical conditional statements combining various indicators and it is supposed to mimic the way traders think).

However, I found other systems that work and are much more complicated than simple TA. I wouldn't be surprised tho if these will be eliminated or their profitability will be greatly reduced in the next few years.

[Officer Farva]

In my limited experience, TA probably won't work by itself. But when I want to improve any of my fundamental or arb based models, TA works well. Its the synergy between having a reasoning behind owning or selling the stock and finding the correct time to execute that can be profitable.

[hedgeyerbets]

If TA adds alpha, it adds alpha in any situation. Are you contending that it adds such a small amount that it can't overcome transaction costs on its own.... ??

[Officer Farva]

[ QUOTE ]
If TA adds alpha, it adds alpha in any situation. Are you contending that it adds such a small amount that it can't overcome transaction costs on its own.... ??

[/ QUOTE ]

This is an invalid statement. I contend that TA may improve returns on certain subsets of the market (e.g. stocks with accerating ROE) but not the market at large.

[hedgeyerbets]

But accelerating ROE is a technical indicator, not a fundamental indicator. So applying other forms of TA to stocks with accelerating ROE is a completely technical strategy.

[Officer Farva]

[ QUOTE ]
But accelerating ROE is a technical indicator, not a fundamental indicator. So applying other forms of TA to stocks with accelerating ROE is a completely technical strategy.

[/ QUOTE ]

1) perhaps roe was a bad example but
2) acceralting ROE is very much profitability analysis not TA

[hedgeyerbets]

I disagree that accelerating ROE is a purely fundamental form of analysis (probably somewhere in the middle b/t ta and fa), but even if it were completely fa, the trade is still purely ta IMO. You are making the trade because you think that x form of ta works on stocks with accelerating ROE... not because you think that the fundamentals are strong.

[eastbay]

[ QUOTE ]

But I believe the way to beat the market is by taking advantage of arbitrage opportunities and value discrepencies. I'm very skeptical that historical price data (TA) has any value in predicting future prices.

[/ QUOTE ]

Well, sorry, but that's demonstrably untrue. It's really not that hard to find strategies which are profitable which do nothing but that.

The edges may be small, but they are there.

eastbay

support/resistance!

[eastbay]

[ QUOTE ]
support/resistance!

[/ QUOTE ]

Nah. At least, I don't know how to make those ideas work.

eastbay

[Sniper]

Support/Resistance hold, until they break.. what could be simpler /images/graemlins/wink.gif

Calculating RSI by hand

[Sniper]

Not to date myself too much, but one of my first tasks out of college was to automate away the work of an entire room full of nice little old ladies with pencils and spreadsheets... turned their entire workload into a single button push on my computer (ala George Jetson).