Evidence based TA, anyone ?

[combotrader]

Sounds odd but true. We are vodoo magic people. We draw lines on charts and look at strange relations between price and volume. Even stranger , we use these to makemoney. Strange because according to an academic technical analyst, there is no proof it actually works. Welcome to the world of David Aronson.

I actually liked it. Chapter 4 teaches the non statistician the basics of sampling and probability. The real twists come lateron when the author tests single rules ( a huge number of them, including divergences), such as 91 day breakout, to SP500 over the last 25 years. None of these single rules have any statistically significant chances of producing profits. In another study hte head and shoulder pattern was also shown to have no benefit whatsoever. He does admit that most TA analysts use complex rules, ie more than one. He describes new ways of thinlking about TA - which I think is the book`s biggest achievement.

I dont want to make this into a review. The book goes onto describe new ways of doing monte carlo and AI means of inductive complex rule testing - which was fascinating.

The real question is - then why does it work ? Why are we making money out of divergences ? If you test the pin bar over the last 30yrs, sample it 10,000 times on different markets, do a white reality check ( read the book) and it does not work statistically - will you still use it ? Well...hell yeah !
But I will be more careful. And I will also do more testing of my rules using statistical inferencing. Is anyone else using neural networkrs, kernel regression.....(not me:cheesy:!)
CT.

 

[theroguetrader]

Sounds odd but true. We are vodoo magic people. We draw lines on charts and look at strange relations between price and volume. Even stranger , we use these to makemoney. Strange because according to an academic technical analyst, there is no proof it actually works. Welcome to the world of David Aronson.

I actually liked it. Chapter 4 teaches the non statistician the basics of sampling and probability. The real twists come lateron when the author tests single rules, such as 91 day breakout, to SP500 over the last 25 years. None of these single rules have any statistically significant chances of producing profits. In another study hte head and shoulder pattern was also shown to have no benefit whatsoever. He says that most TA analysts use complex rules, ie more than one.

I dont want to make this into a review. The book goes onto describe new ways of doing monte carlo and AI means of inductive complex rule testing - which was fascinating.

The real question is - then why does it work ? Why are we making money out of divergences ? If you test the pin bar over the last 30yrs, sample it 10,000 times on different markets, do a white reality check ( read the book) and it does not work statistically - will you still use it ? Well...hell yeah !
But I will be more careful. And I will also do more testing of my rules using statistical inferencing. Is anyone else using neural networkrs, kernel regression.....(not me:cheesy:!)
CT.

Does this use of " inductive complex rule testing " and use of " neural networks and " kernel regression " make him money ? or would it be easier (never liked that word complex)- just to buy rising prices and sell falling prices..?

 

[Mathemagician]

Outstanding book, IMO. It is one of the handful of books (5 out of the >100 of books that I own) that I actually recommend to people when asked.

jj

 

[DaxDestroyer]

Sounds odd but true. We are vodoo magic people. We draw lines on charts and look at strange relations between price and volume. Even stranger , we use these to makemoney. Strange because according to an academic technical analyst, there is no proof it actually works. Welcome to the world of David Aronson.

I actually liked it. Chapter 4 teaches the non statistician the basics of sampling and probability. The real twists come lateron when the author tests single rules ( a huge number of them, including divergences), such as 91 day breakout, to SP500 over the last 25 years. None of these single rules have any statistically significant chances of producing profits. In another study hte head and shoulder pattern was also shown to have no benefit whatsoever. He does admit that most TA analysts use complex rules, ie more than one. He describes new ways of thinlking about TA - which I think is the book`s biggest achievement.

I dont want to make this into a review. The book goes onto describe new ways of doing monte carlo and AI means of inductive complex rule testing - which was fascinating.

The real question is - then why does it work ? Why are we making money out of divergences ? If you test the pin bar over the last 30yrs, sample it 10,000 times on different markets, do a white reality check ( read the book) and it does not work statistically - will you still use it ? Well...hell yeah !
But I will be more careful. And I will also do more testing of my rules using statistical inferencing. Is anyone else using neural networkrs, kernel regression.....(not me:cheesy:!)
CT.

Outstanding book, IMO. It is one of the handful of books (5 out of the >100 of books that I own) that I actually recommend to people when asked.

jj

Yes, interesting book, BUT I have to admit, that after reading half of it and scanning through the rest, I was not convinced that his Proofs were "real world examples" for me.
Understandably he had to find simple rules as the computer needs that input, but as I use different rules, it did not help me.

And I am a bit "unhappy", that he teaches people that most common sense rules are wrong

Makes trading more difficult again.

 

[Mathemagician]

Yes, interesting book, BUT I have to admit, that after reading half of it and scanning through the rest, I was not convinced that his Proofs were "real world examples" for me.
Understandably he had to find simple rules as the computer needs that input, but as I use different rules, it did not help me.

And I am a bit "unhappy", that he teaches people that most common sense rules are wrong

Makes trading more difficult again.

I'd be interested in hearing about some specific instances of proofs that do not translate into real-world examples. This is the greatest opportunity for learning.

It was my impression that the book was not at all about any specific rules, but rather a set of statistical techniques that can be applied to return series regardless of the underlying rule that generated them. It is also my impression that the chapter in which he does the data mining exercise was meant more to provide a comprehensive application of the techniques discussed rather than a search for a profitable trading rule.

I could easily be wrong here, though, as these thoughts are obviously subjective. Of course, I did read the entire book... :smart:

Thanks!
jj

 

[combotrader]

The authors main thrust is to describe ways of data mining - and why data mining is intrinsically flawed. This is the fools gold way of backtesting to get a profitable rule.

His proofs nothwithstanding - ways of analysing divergences between 2 prices or between price and indicator described in the book is illuminating.

His point is valid. Would you live in a house whose materials had not been tested in various situations ? Would you accept a treatment from your doctor which had not been statistically tested ? Then why trade you hard earned money on rules that are going to be right 50% of the time - at the very best - depending on which sample of trades you choose.

Perhaps it is easier to accept that one can make money by being wrong 70% of the time. It all boils down to risk control.

 

[DaxDestroyer]

I'd be interested in hearing about some specific instances of proofs that do not translate into real-world examples. This is the greatest opportunity for learning.

It was my impression that the book was not at all about any specific rules, but rather a set of statistical techniques that can be applied to return series regardless of the underlying rule that generated them. It is also my impression that the chapter in which he does the data mining exercise was meant more to provide a comprehensive application of the techniques discussed rather than a search for a profitable trading rule.

I could easily be wrong here, though, as these thoughts are obviously subjective. Of course, I did read the entire book... :smart:

Thanks!
jj

This is it You are right

Therefore nothing for me , no new input. All the examples he used, I don't trade.

Maybe I should be happy as I was afraid, that he proves me wrong

 

[BSD]

I don't backtest or anything and find most studies that purport to prove that TA works one way or the other to be pretty lacking as something relevant is usually missing from the equation, but anyway found this the other day which might be of interest:

"THURSDAY, JANUARY 10, 2008

Do Reversal Bars Really Work?
The market finally decided it had endured enough selling and put in a strong afternoon reversal. The bar looks nice on a chart, but is it indicative of a longer-term reversal? To test it I ran the following quantitative study (as usual each trade is $100,000):

Over the first 1-7 days, it appears to be a toss-up, but as you look a little bit further out there appears to be a solid edge to the upside. One reason the edge appears to be lower in the first few days is that the market has just made a large move up. Frequently this initial thrust takes a few days to digest (more on that lower down).

What I find especially compelling about this scenario are the size of the average winners - especially when compared to the average loser. If the reversal bar works, the expectation is for somewhere around a 4-5% follow through in the next 2-4 weeks based on these results. Note the win/loss ratios and profit factors once you get out more than 10 days. They're quite good.

For a more detailed look I evaluated the results 14 days out. That would put us at the end of the month and the next Fed meeting. Obviously no one will be worried about this backtest when the Fed is about to announce.

Fourteen days out 16 of 24 trades were winners. Of those sixteen winners, all of them pulled back at least 0.5% from the reversal day close at some point. The average drawdown among those 16 winners was 2.6%. The largest drawdown among winning trades was 6.6% which occurred after the reversal bar last August 6th. Five of the sixteen winners actually posted a lower low before turning higher again. In other words, it's probably not neccessary to chase this trade. There will most likely be some backing and filling which should allow for a better entry point or some scaling in.

This one gets stamped "quantifiable edge"."

LINK:
Quantifiable Edges: Do Reversal Bars Really Work?

 

[BSD]

To be honest, all I look at in trading is price.

Specifically,

A: What's the trend, and on various time frames at that...

B: Where is the next relevant Support / Resistance...

One doesn't need anything more does one.

Here we are back at that old question again, why unnecessarily make things more complicated than is warranted ?

"Support for Resistance: Technical Analysis and Intraday Exchange Rates

“Support” and “resistance” levels—points at which an exchange rate trend may be interrupted and reversed—are widely used for short-term exchange rate forecasting. Nevertheless, the levels’ ability to predict intraday trend interruptions has never been rigorously evaluated. This article undertakes such an analysis, using support and resistance levels provided to customers by six firms active in the foreign exchange market. The author offers strong evidence that the levels help to predict intraday trend interruptions. "
LINK:
Support for Resistance: Technical Analysis and Intraday Exchange Rates - Federal Reserve Bank of New York

That said, one doesn't really need a study to prove what simple eyeballing of every single chart you'll ever look at provides just as well, does one ?

One doesn't need a study either to prove that the sun rises when a new day dawns, does one.

 

[DaxDestroyer]

To be honest, all I look at in trading is price.

Specifically,

A: What's the trend, and on various time frames at that...

B: Where is the next relevant Support / Resistance...

One doesn't need anything more does one.

Here we are back at that old question again, why unnecessarily make things more complicated than is warranted ?

"Support for Resistance: Technical Analysis and Intraday Exchange Rates

“Support” and “resistance” levels—points at which an exchange rate trend may be interrupted and reversed—are widely used for short-term exchange rate forecasting. Nevertheless, the levels’ ability to predict intraday trend interruptions has never been rigorously evaluated. This article undertakes such an analysis, using support and resistance levels provided to customers by six firms active in the foreign exchange market. The author offers strong evidence that the levels help to predict intraday trend interruptions. "
LINK:
Support for Resistance: Technical Analysis and Intraday Exchange Rates - Federal Reserve Bank of New York

That said, one doesn't really need a study to prove what simple eyeballing of every single chart you'll ever look at provides just as well, does one ?

One doesn't need a study either to prove that the sun rises when a new day dawns, does one.

Yes and NO

I don't like it toooooo sophisticated, but SOME things, I add to increase my edge.
If you allow, Sir

 

[DaxDestroyer]

I don't backtest or anything and find most studies that purport to prove that TA works one way or the other to be pretty lacking as something relevant is usually missing from the equation, but anyway found this the other day which might be of interest:

"THURSDAY, JANUARY 10, 2008

Do Reversal Bars Really Work?
The market finally decided it had endured enough selling and put in a strong afternoon reversal. The bar looks nice on a chart, but is it indicative of a longer-term reversal? To test it I ran the following quantitative study (as usual each trade is $100,000):

Over the first 1-7 days, it appears to be a toss-up, but as you look a little bit further out there appears to be a solid edge to the upside. One reason the edge appears to be lower in the first few days is that the market has just made a large move up. Frequently this initial thrust takes a few days to digest (more on that lower down).

What I find especially compelling about this scenario are the size of the average winners - especially when compared to the average loser. If the reversal bar works, the expectation is for somewhere around a 4-5% follow through in the next 2-4 weeks based on these results. Note the win/loss ratios and profit factors once you get out more than 10 days. They're quite good.

For a more detailed look I evaluated the results 14 days out. That would put us at the end of the month and the next Fed meeting. Obviously no one will be worried about this backtest when the Fed is about to announce.

Fourteen days out 16 of 24 trades were winners. Of those sixteen winners, all of them pulled back at least 0.5% from the reversal day close at some point. The average drawdown among those 16 winners was 2.6%. The largest drawdown among winning trades was 6.6% which occurred after the reversal bar last August 6th. Five of the sixteen winners actually posted a lower low before turning higher again. In other words, it's probably not neccessary to chase this trade. There will most likely be some backing and filling which should allow for a better entry point or some scaling in.

This one gets stamped "quantifiable edge"."

LINK:
Quantifiable Edges: Do Reversal Bars Really Work?

This is an example of things, I like to read, but in daily trading could not care less as I don't hold positions in 95% of time

 

[BSD]

This is an example of things, I like to read, but in daily trading could not care less as I don't hold positions in 95% of time

Yup. Just added that because looks like some people here go for stuff like that.

I personally pretty much disregard these things.

Yes and NO

I don't like it toooooo sophisticated, but SOME things, I add to increase my edge.
If you allow, Sir

Bien sûr M. Carlos

Must work on my Spanish before I could hope to get that right I'm afraid.

 

[DaxDestroyer]

Yup. Just added that because looks like some people here go for stuff like that.

I personally pretty much disregard these things.


Bien sûr M. Carlos

Must work on my Spanish before I could hope to get that right I'm afraid.

My french is VERY limited, even after working for BNP some time ago Only some nasty words you never forget

 

[jm99]

yeah top book, would appeal more to quant traders.

hey jj from tradestation boards right. for the record all, mathemagician really knows his stuff

 

[jm99]

I don't backtest or anything and find most studies that purport to prove that TA works one way or the other to be pretty lacking as something relevant is usually missing from the equation, but anyway found this the other day which might be of interest:

"THURSDAY, JANUARY 10, 2008

Do Reversal Bars Really Work?
The market finally decided it had endured enough selling and put in a strong afternoon reversal. The bar looks nice on a chart, but is it indicative of a longer-term reversal? To test it I ran the following quantitative study (as usual each trade is $100,000):

Over the first 1-7 days, it appears to be a toss-up, but as you look a little bit further out there appears to be a solid edge to the upside. One reason the edge appears to be lower in the first few days is that the market has just made a large move up. Frequently this initial thrust takes a few days to digest (more on that lower down).

What I find especially compelling about this scenario are the size of the average winners - especially when compared to the average loser. If the reversal bar works, the expectation is for somewhere around a 4-5% follow through in the next 2-4 weeks based on these results. Note the win/loss ratios and profit factors once you get out more than 10 days. They're quite good.

For a more detailed look I evaluated the results 14 days out. That would put us at the end of the month and the next Fed meeting. Obviously no one will be worried about this backtest when the Fed is about to announce.

Fourteen days out 16 of 24 trades were winners. Of those sixteen winners, all of them pulled back at least 0.5% from the reversal day close at some point. The average drawdown among those 16 winners was 2.6%. The largest drawdown among winning trades was 6.6% which occurred after the reversal bar last August 6th. Five of the sixteen winners actually posted a lower low before turning higher again. In other words, it's probably not neccessary to chase this trade. There will most likely be some backing and filling which should allow for a better entry point or some scaling in.

This one gets stamped "quantifiable edge"."

LINK:
Quantifiable Edges: Do Reversal Bars Really Work?

interesting, but little statistical significance with only 24 trades imo

this is a similar time based mean reversion system.

http://www.trade2win.com/boards/us-indices/22794-dow-2007-a-733.html#post364942

off topic, its interesting the the last trade was a loser......If you take away the monthly filter the ec dipped during 2000-03 where these sort of mean reversion systems failed badly....another indication imo that we are entering a similar market condition

 

[BSD]

Then why trade you hard earned money on rules that are going to be right 50% of the time - at the very best - depending on which sample of trades you choose.

Perhaps it is easier to accept that one can make money by being wrong 70% of the time. It all boils down to risk control.

Does he really say that in the book ?

That's where I say again that most of these studies are inherently flawed because the authors fail to comprehend what is really success relevant and what isn't.

Hit rate most definitely is not success relevant.

Show me a trader who is "right" anything way over 50% of the time, and who does not achieve that via stop losses that are larger than your take profits, meaning it's more of a feel-well strategy than anything else, but it most decidely is not about the bottom line.

Systems that are right more often than they are wrong are decidedly less net profitable than their lower hit rate brethren:

Brett Steenbarger:

"...As a rule, maximizing batting average/minimizing drawdown comes at the cost of lowering overall system profitability...."

"William Eckhardt:

The Win/Loss Ratio
“One common adage on this subject that is completely wrongheaded is: You can’t go broke taking profits. That’s precisely how many traders do go broke. While amateurs go broke by taking large losses, professionals go broke by taking small profits. The problem in a nutshell is that human nature does not operate to maximize gain but rather to maximize the chance of a gain. The desire to maximize the number of winning trades (or minimize the number of losing trades) works against the trader. The success rate of trades is the least important performance statistic and may even be inversely related to performance. …

What really matters is the long-run distributions of outcomes from your trading techniques, systems, and procedures. But, psychologically, what seems of paramount importance is whether the positions that you have right now are going to work. Current positions seem to be crucial beyond any statistical justification. It’s quite tempting to bend your rules to make your current trades work, assuming that the favorability of your long-term statistics will take care of future profitability. Two of the cardinal sins of trading - giving losses too much rope and taking profits prematurely - are both attempts to make current positions more likely to succeed, to the severe detriment of long-term performance.”
Market Wizards

Having said that, it is obviously VERY important that people trade in a way they feel comfortable with, so if they want a high hit rate through larger TP's than SL's, than that is just fine.

Becoming net profitable at all is very much more important than total gains.

Finally, not to forget what are probably the two single most important factors one needs to fully understand in trading if one wants to make it, let alone make it big:

A: Mark Douglas excellent and very common-"sensical" observation that anything can happen anytime at all in markets overthrowing your clever analysis, all it takes is one big order going against you.

B: Kenneth Grant, in "Trading Risk: Enhanced Profitability through Risk Control", depicts his experience as risk manager for some of the best and most successful hedge funds, amongst others Paul Tudor Jones funds and Steve Cohens SAC Capital, that:

ACROSS ALL TRADING STYLES, TIME FRAMES AND TRADERS, ONE RULE HOLDS TRUE:

10% OF ALL TRADES INEVITABLY ACCOUNT FOR 90% OF PROFITS !


Thing is, some wise guys might now assume that there is sthg they can do to improve that division of labour.

Problem is, doesn't work that way in real life.

In real life the law seems to be that it's pretty much always 20% of input that generates 80% of output, which is really all you need to concentrate on to achieve outstanding success, while in trading it's just a bit more extreme, 10% of input generates 90% of output.

 

[DaxDestroyer]

Does he really say that in the book ?

That's where I say again that most of these studies are inherently flawed because the authors fail to comprehend what is really success relevant and what isn't.

Hit rate most definitely is not success relevant.

Show me a trader who is "right" anything way over 50% of the time, and who does not achieve that via stop losses that are larger than your take profits, meaning it's more of a feel-well strategy than anything else, but it most decidely is not about the bottom line.

Systems that are right more often than they are wrong are decidedly less net profitable than their lower hit rate brethren:

Brett Steenbarger:

"...As a rule, maximizing batting average/minimizing drawdown comes at the cost of lowering overall system profitability...."

"William Eckhardt:

The Win/Loss Ratio
“One common adage on this subject that is completely wrongheaded is: You can’t go broke taking profits. That’s precisely how many traders do go broke. While amateurs go broke by taking large losses, professionals go broke by taking small profits. The problem in a nutshell is that human nature does not operate to maximize gain but rather to maximize the chance of a gain. The desire to maximize the number of winning trades (or minimize the number of losing trades) works against the trader. The success rate of trades is the least important performance statistic and may even be inversely related to performance. …

What really matters is the long-run distributions of outcomes from your trading techniques, systems, and procedures. But, psychologically, what seems of paramount importance is whether the positions that you have right now are going to work. Current positions seem to be crucial beyond any statistical justification. It’s quite tempting to bend your rules to make your current trades work, assuming that the favorability of your long-term statistics will take care of future profitability. Two of the cardinal sins of trading - giving losses too much rope and taking profits prematurely - are both attempts to make current positions more likely to succeed, to the severe detriment of long-term performance.”
Market Wizards

Having said that, it is obviously VERY important that people trade in a way they feel comfortable with, so if they want a high hit rate through larger TP's than SL's, than that is just fine.

Becoming net profitable at all is very much more important than total gains.

Finally, not to forget what are probably the two single most important factors one needs to fully understand in trading if one wants to make it, let alone make it big:

A: Mark Douglas excellent and very common-"sensical" observation that anything can happen anytime at all in markets overthrowing your clever analysis, all it takes is one big order going against you.

B: Kenneth Grant, in "Trading Risk: Enhanced Profitability through Risk Control", depicts his experience as risk manager for some of the best and most successful hedge funds, amongst others Paul Tudor Jones funds and Steve Cohens SAC Capital, that:

ACROSS ALL TRADING STYLES, TIME FRAMES AND TRADERS, ONE RULE HOLDS TRUE:

10% OF ALL TRADES INEVITABLY ACCOUNT FOR 90% OF PROFITS !


Thing is, some wise guys might now assume that there is sthg they can do to improve that division of labour.

Problem is, doesn't work that way in real life.

In real life the law seems to be that it's pretty much always 20% of input that generates 80% of output, which is really all you need to concentrate on to achieve outstanding success, while in trading it's just a bit more extreme, 10% of input generates 90% of output.

:clap::clap::clap: Markus has a field day today :clap::clap::clap:

 

[BSD]

Muchas gracias Carlos, but far too kind, I'm afraid I'm just a lazy bugger synthesizing the work that others have very kindly seen fit to provide the public realm - which then translates into me - with so that I may then put their sweat and tears to good use, and go about achieving my objectives.

 

[DaxDestroyer]

Muchas gracias Carlos, but far too kind, I'm afraid I'm just a lazy bugger synthesizing the work that others have very kindly seen fit to provide the public realm - which then translates into me - with so that I may then put their sweat and tears to good use, and go about achieving my objectives.

I'd rather call this "prudent behaviour"



You don't grow your food yourself, nor build your house with your own hands

So there is no reason to reinvent the wheel.

If everybody started at ZERO, nobody could ever progress

 

[Mathemagician]

Does he really say that in the book ?

No.

jj