Backtested Systems, Lies, Damned Lies, & Statistics

[tech/a]

I personally think no, it wasn't, it had many similarities to the TURTLE system Now this is not a criticism (not yet anyway). Was the methodology (possibly undisclosed) examined in anyway for variables that were similar or the same within TechTrader?

Initially there was discussion about the Turtle Trading method and there was a method designed around that method--its far from the final T/T being used today.The only similarity is that the method is a longterm trend following method.

From this starting point we can ask the following questions;

Is this study larger, continued for longer, or otherwise more substantial than previous studies (studies equating to trading methodologies)

No.Its the longest I have seen walking forward though and public,there maybe more somewhere that are public and I like most would love to follow them.I was originally going to pull the public trading of T/T after year 1 but was convinced to leave it running.

Is the methodology of this study any more rigorous (in particular, does it cover any specific methodological criticisms of previous studies)

No. Infact it is involved in many of the critisisms I and many others have in system testing portfolio methods in particular.Single entity methods IE Futures can be studied more rigorously--but I'll get to that in a moment.

Will the numerical results of this study add significantly to a meta-analysis of previous studies?

No,but it has gone a long way in helping some find a methodology that suits them as there are a few hybrids I know of.

Is the population studied different in any way (has the study examined capitalization structures, duration of listing, economic environment, market environment, inflation, interest, alternate investment classes etc)

Yes and No.Yes by way of culling a universe of stocks simply based upon selection of allowable traded stock by the margin lender in the case in question that was/is BT (Bankers Trust). No in that the restrictions due to data are common to 99% of systems testers relying on retail data.


The single biggest problem a systems or method trade/developer has is the one of survivorship.Todays list is used for testing whether that be the BT list the ASX or ASX300 or any group of stocks you wish to include.That list is only valid for today.
Testing data 10 yrs ago is impossible we just dont have the list of stocks that were either included in the margin list or the ASX and your criteria could have selected/traded and been caught out by delisted stocks not in your data base---today.

So that brings the question which I have asked myself often as to how accurate then are the results of the method and could my "Blueprint" (The statistics which form the basis for comparison of performance as I trade the system forward) be out far enough to risk ruin?

I have perhaps answered my own question (Albeit convieniently) by reasoning the following.
Buy placing a set of conditions with a set of variables into a method and testing it over a great deal of timelines and stocks (Both in the AUST bourse and Overseas) results are consistantly within acceptable parameters.I have not tested one case where ruin would occur.
Before you get all tied up in accuracy I feel one aspect of this and any low risk parameter set method will distribute the risk broadly enough to avoid ruin if even 2 or 3 outliers of delisting hit it at once.No I dont have the stats on that but I do know that there are even fewer delistings of stocks in a sustained up trend than downtrend (From observations of charts).Thats what the method is designed to trade.
So by applying the method rules to all universes I have tested---as of yet I havent found a case for ruin.

Forward trading of 16 versions of this method from reports from those who keep in contact with me report that they are having similar results to the "Numbers" found in the original design traded on Reef.

This doesnt mean that there couldnt be a case of ruin waiting out there for anyone who got their numbers from trading at the worst time in history.
Yes, Ive also heard that even idiots can profit in a bullmarket--well I'm living proof,there are a great deal of us however that no matter how bullish a market is consistant profit avoids us and ruin is familiar.

While not statistically perfect due to the in ability to gain correct data.The entry,exit,stop and position sizing model has shown in every case I have found to be profitable and without a single case of ruin.

I'm also yet to see a test or a traded method which has numbers which fall below those in the "Blueprint" spat out in the original method.
If actual forward trading returned figures outside (To the negative) to those found in the "blueprint" then I would stop trading it.I'm also yet to see trading above the higher end of the Montecarlo results of the method,although I have achieved that end.

Now is it the best method around---I doubt it.
Could it be better---yes I'm sure it can as I have a better than "Original" I trade myself. I also trade the original.
Is it the most efficient---hardly.
Does it suit me ---absolutely.

Having said this I also believe that data has exactly the same effect on any method designed for portfolio trading and as such results will be similar to my own in respect of accuracy and acceptability to the designer.
Acceptable to some maybe not others.

Back to sleep.

 

[Milk Man]

Quick question: what is the Turtle Traders method? Ive heard of it before but Ive never seen exactly how it works.

I do see your point Ducati. I have the same reservations myself about backtested systems. However, I still believe that a backtested, mechanical system is the best way to consitently return a profit. As tech said, if parameters fall outside those of your testing then its back to the drawing board.

Having said this, is it then a bad idea to start up your own business? The figures are similarly bad to trading and they only have, what you call, methodological testing procedures. As I see it, people who start their own business have a pretty firm grip on the methodological side of things as they generally would have worked in the industry first. It is then, my theory, that it is the lack of risk management (and backtesting) that cannot assure everyone a profit.

Is it then a bad idea to start your own business? No...
Everything in life involves an element of risk. The way risk is managed directly influences our success; using backtesting can only be a positive for this. However, the methodology is an important part of designing a successful system.

 

[Bobby]

Quick question: what is the Turtle Traders method? Ive heard of it before but Ive never seen exactly how it works.

I do see your point Ducati. I have the same reservations myself about backtested systems. However, I still believe that a backtested, mechanical system is the best way to consitently return a profit. As tech said, if parameters fall outside those of your testing then its back to the drawing board.

Having said this, is it then a bad idea to start up your own business? The figures are similarly bad to trading and they only have, what you call, methodological testing procedures. As I see it, people who start their own business have a pretty firm grip on the methodological side of things as they generally would have worked in the industry first. It is then, my theory, that it is the lack of risk management (and backtesting) that cannot assure everyone a profit.

Is it then a bad idea to start your own business? No...
Everything in life involves an element of risk. The way risk is managed directly influences our success; using backtesting can only be a positive for this. However, the methodology is an important part of designing a successful system.

Hullo Milkman,
The term turtle traders was what a great trader (Richard Dennis ) called his new apprentices to the world of trading back in 1983, as most young turtles don't make it to an adult .
The original method is on the net just google it.
Bob.

 

[wayneL]

Hullo Milkman,
The term turtle traders was what a great trader (Richard Dennis ) called his new apprentices to the world of trading back in 1983, as most young turtles don't make it to an adult .
The original method is on the net just google it.
Bob.

http://www.tradingblox.net/Rules/turtlerules.pdf

 

[Milk Man]

Thanks guys.

 

[ducati916]

Prof.

quick question- Is it technical analysis or pure mechanical trading that you don't like?

Mostly technical analysis.
I actually think that most people would be far better off with some form of mechanical or formula methodology.

tech/a

Initially there was discussion about the Turtle Trading method and there was a method designed around that method--its far from the final T/T being used today.The only similarity is that the method is a longterm trend following method.

The discussion that revolved around the Turtle system identified flaws that TT sought to rectify?
If that is the case, then this would be a positive.
It also answers the question

Is the methodology of this study any more rigorous (in particular, does it cover any specific methodological criticisms of previous studies)

As the methodology was modified, or completely revamped, is there is no point of comparison with previous methodologies? If there is no point of commonality this would thereby reduce the total gross number of the population, as we cannot complete a meta-analysis thereby increasing the total population, and complying with a power calculation that may push the study into the *statistically significant*
If there are however points of commonality, viz. the *good bits*, how did you utilize this data?

Is this study larger, continued for longer, or otherwise more substantial than previous studies (studies equating to trading methodologies)

Forward trading of 16 versions of this method from reports from those who keep in contact with me report that they are having similar results to the "Numbers" found in the original design traded on Reef.

Therefore have you actually looked at the number of Turtle Traders that there were, how long they traded the system etc?
This is definitely a positive as more data is gathered, as I understand none of the other 16 portfolios are the same. However, how many do actually replicate positions. By way of example UTB was a 190% return in your public TT. Have any of the other 16 replicated the same results?
If not why not? What accounts for the variability? Is it systematic, or individual? Did testing indicate this variability prior to taking it live? If not why not? And so on.

The problem, and this has nothing to do with your research, is that trading methodologies, and research of trading methodologies are extremely secretive
People, Institutions, do not make available proprietary trading information.
Publically listed Investment companys have to disclose results, and positions, but do not have to disclose in any meaningful way their methodology.

This leaves very serious questions as to benchmarks.
This is a problem that can seriously bias your statistical results, and potentially cause serious consequences.

Is the population studied different in any way (has the study examined capitalization structures, duration of listing, economic environment, market environment, inflation, interest, alternate investment classes etc)

Here I would put it to you that there are some very different fundamental factors that differentiate the populations studied;

1...ASX stocks do not utilize Market Makers in the same manner as the US
2...ASX stocks have far fewer adjustments of Capitalization structures (stock splits would be an example)
3...ASX stocks do not have anywhere near the Short-selling capabilities (you can short sell, but, it is not an easy thing to do)
4...Differences in liquidity, and thus spreads.
5...Differences in traded Options on stocks
6...Differences in volatility
7...Differences in volume
8...Differences in foreign stocks listed

I have perhaps answered my own question (Albeit convieniently) by reasoning the following.
Buy placing a set of conditions with a set of variables into a method and testing it over a great deal of timelines and stocks (Both in the AUST bourse and Overseas)

There are others, but just in these examples, there is enough variability to invalidate from a statistical point of view, calling them a common population.
By invalidating the gross numbers in the population, it must also call into question the validity of the found results due to the law of small numbers.
On the software, is there any *POWER CALCULATION* calculation?
This would determine the required number to generate a statistically significant result. I suspect there will not be any, as this would highlight immediately the *population* is woefully low, and no statistical accuracy can be inferred from the calculated data.

If we examine your inclusion criteria against our medical benchmark by asking the following two questions;

1......Who was included in the study?
Many medical trials routinely exclude patients with co-existing illness, those who do not speak the native language, using other medications etc.
This approach while scientifically clean has no logic behind the reasoning as clinical trial results will be used as "best practice", within wide patient groups.
It makes no sense, as, the requirement in the real world, is very different from the research protocol.

2...Who was excluded from the study?
For example, a randomised controlled trial may be restricted to patients with moderate or severe forms of heart failure, this selection policy could lead directly to false conclusions regarding patient populations with *mild* heart failure

The two questions illustrate the seriousness as to the distortions that can be induced from *selection of the population* and forms the selection bias that invalidates studies from drawing statistically derived conclusions.
It would seem that the selection of stocks to be included has some serious issues to be overcome prior to drawing statistically derived conclusions.

The single biggest problem a systems or method trade/developer has is the one of survivorship.Todays list is used for testing whether that be the BT list the ASX or ASX300 or any group of stocks you wish to include.That list is only valid for today.

Yes, mechanical testing fails on survivorship bias.
That also invalidates the statistical data generated.
However, you are jumping ahead of the curve, I have not yet worked my way through all the potential biases.

While not statistically perfect due to the in ability to gain correct data.The entry,exit,stop and position sizing model has shown in every case I have found to be profitable and without a single case of ruin.

Jumping ahead again...................
Data, quality, and surrounding issues will all be looked at.

milk

I do see your point Ducati. I have the same reservations myself about backtested systems. However, I still believe that a backtested, mechanical system is the best way to consitently return a profit. As tech said, if parameters fall outside those of your testing then its back to the drawing board.

Possibly it is for the majority, but then again possibly not.
The problem being that many do not really understand what they are in actuality trading. If you understand all the inherent weaknesses, and have allowances, or tolerances for the failings, then that is quite different from trading a computer print-out without any understanding.

By the time it falls outside of your parametres, you could be taking up your palatial residence at #3 dustbin. Remember some people leverage these systems to the hilt, trade them in futures markets etc.

I'm only using TT as it is a publically disclosed system.
Most are very secretive, and thus you'll never see the blow-ups.
If you are committed to a mechanical system, thats fine, I actually am quite pro-mechanical, but informed mechanical.

Having said this, is it then a bad idea to start up your own business? The figures are similarly bad to trading and they only have, what you call, methodological testing procedures. As I see it, people who start their own business have a pretty firm grip on the methodological side of things as they generally would have worked in the industry first. It is then, my theory, that it is the lack of risk management (and backtesting) that cannot assure everyone a profit

Well I would beg to differ.
Most start-ups go belly up.
Most who start small businesses are under capitalized, clueless, and destined to failure, not lasting past the first year or two.

Trading/Investing is, or should be, a small business.
How many coming into this *industry* knew what they were doing prior to entering? How many have a clue even after 3 yrs?

tech/a
I'll come back to you with the next set of questions, after I punish the gym equipment, Friday, is my relaxation day.

jog on
d998

 

[Milk Man]

http://www.tradingblox.net/Rules/turtlerules.pdf

Wow! That is pure gold, no, platinum! That is even more insightful than Nick's book. Everyone should read that before considering trading, even if not the system itself, but just the methodology.

:bananasmi

 

[ghotib]

Please, where is the study and/or the discussion of it that Tech was talking about in the the quotes at the start of this thread?

Hopefully if I can read that I won't have to bother you all with a stack of other questions about what we're talking about.

Cheers,

Ghoti

 

[tech/a]

tech/a



The discussion that revolved around the Turtle system identified flaws that TT sought to rectify?
If that is the case, then this would be a positive.
It also answers the question

As the methodology was modified, or completely revamped, is there is no point of comparison with previous methodologies? If there is no point of commonality this would thereby reduce the total gross number of the population, as we cannot complete a meta-analysis thereby increasing the total population, and complying with a power calculation that may push the study into the *statistically significant*
If there are however points of commonality, viz. the *good bits*, how did you utilize this data?

The turtle system was never pulled apart or studied to any degree.
All I/We looked at was how it entered and exited as a basis from which to start. The Good bits were that it traded a trend over long time frames.Nothing more complex than that.

Therefore have you actually looked at the number of Turtle Traders that there were, how long they traded the system etc?
This is definitely a positive as more data is gathered, as I understand none of the other 16 portfolios are the same. However, how many do actually replicate positions. By way of example UTB was a 190% return in your public TT. Have any of the other 16 replicated the same results?
If not why not? What accounts for the variability? Is it systematic, or individual? Did testing indicate this variability prior to taking it live? If not why not? And so on.

Re Turtle system answered above.
The 16 portfolios were started at various times myself and Daryl both traded UTB to its end.As for the other 15 traders using the method and or a hybrid i dont have individual trade results.But i would say that most got a good part of UTB.As start dates will obviously have an impact on which entries are triggered I would expect there to be both same and different stocks in portfolio's over different start dates. Incedently trading EVERY stock selected by the method (had you the funds) returns massive profit.

The problem, and this has nothing to do with your research, is that trading methodologies, and research of trading methodologies are extremely secretive
People, Institutions, do not make available proprietary trading information.
Publically listed Investment companys have to disclose results, and positions, but do not have to disclose in any meaningful way their methodology.

True and thats why I am happy to have this public,I want people like you to pull it to pieces and many have added constructive critisism over the years.Its as much a learning process to me as anyone.With something to atleast benchmark that is working there is a starting point for discussion.To this end it has/is serving its purpose---and making a buck while I/we learn more.

This leaves very serious questions as to benchmarks.
This is a problem that can seriously bias your statistical results, and potentially cause serious consequences.

There is one benchmark I need and that is Profit.I dont need to have a comparitive benchmark Im happy to improve on the singular benchmark printed by the method itself.

Here I would put it to you that there are some very different fundamental factors that differentiate the populations studied;

1...ASX stocks do not utilize Market Makers in the same manner as the US
2...ASX stocks have far fewer adjustments of Capitalization structures (stock splits would be an example)
3...ASX stocks do not have anywhere near the Short-selling capabilities (you can short sell, but, it is not an easy thing to do)
4...Differences in liquidity, and thus spreads.
5...Differences in traded Options on stocks
6...Differences in volatility
7...Differences in volume
8...Differences in foreign stocks listed

Absolutely correct but you are presuming that a method should work within its "blueprint" in every situation you place it in.I disagree the system and any system can be designed around any set condition and conditions.The 8 condition "variables" you mention above have little impact on the specifics of the method.

There are others, but just in these examples, there is enough variability to invalidate from a statistical point of view, calling them a common population.
By invalidating the gross numbers in the population, it must also call into question the validity of the found results due to the law of small numbers.
On the software, is there any *POWER CALCULATION* calculation?
This would determine the required number to generate a statistically significant result. I suspect there will not be any, as this would highlight immediately the *population* is woefully low, and no statistical accuracy can be inferred from the calculated data.

Well the population or Universe of stocks used in the test from the BT margin list is 327 stocks and the number of trades tested is over 800 (Possible but due to capital constraints only around 120 trades taken ---from memory I'm at the office). This has been replicated by countless other both using metastock/amibroker and wealthlab. Many select universes well away from the parameters set when the system was initially designed. Other than using far to small timeframes I am yet to recieve a test result showing ruin despite 100s if not 1000s of tests and various universes both in size and criteria.

If we examine your inclusion criteria against our medical benchmark by asking the following two questions;

1......Who was included in the study?
Many medical trials routinely exclude patients with co-existing illness, those who do not speak the native language, using other medications etc.
This approach while scientifically clean has no logic behind the reasoning as clinical trial results will be used as "best practice", within wide patient groups.
It makes no sense, as, the requirement in the real world, is very different from the research protocol.

2...Who was excluded from the study?
For example, a randomised controlled trial may be restricted to patients with moderate or severe forms of heart failure, this selection policy could lead directly to false conclusions regarding patient populations with *mild* heart failure

The two questions illustrate the seriousness as to the distortions that can be induced from *selection of the population* and forms the selection bias that invalidates studies from drawing statistically derived conclusions.
It would seem that the selection of stocks to be included has some serious issues to be overcome prior to drawing statistically derived conclusions.

Duc your again setting your own standard of one that must have a system satisfy every situation and every universe and under every condition.
I dont agree that that is necessary and think "so Far" I've been able to prove that point. I have made it clear the selection process for which the method was/is designed attempting to adapt it to other universes and in a bear market would have the same consequences as using asprin to treat diabetes.

The broader you design a method the more complex will be your criteria.

To the other extreme trading a singualr entity such as a future, is just as valid as trading every stock in every condition in every market in the world!
The simplistic and profitable method used to trade the Future would have little chance of being profitable in every concievable situation and this is the same for every method designed.They will/should be specific to their task.

Yes, mechanical testing fails on survivorship bias.
That also invalidates the statistical data generated.
However, you are jumping ahead of the curve, I have not yet worked my way through all the potential biases.

Here we disagree.Incomplete in a specific set of data yes invalidates??

 

[tech/a]

Ghotib.

If you mean TechTrader then there are 100s of pages written on the topic and it is traded live with each weeks results published by Darryl,who first asked me the question on wether a mechanical system could be designed and traded profitably---then the link and weeks of reading are here.

http://lightning.he.net/cgi-bin/suid/~reefcap/ultimatebb.cgi?ubb=forum;f=74

 

[ducati916]

tech/a

Duc your again setting your own standard of one that must have a system satisfy every situation and every universe and under every condition.
I dont agree that that is necessary and think "so Far" I've been able to prove that point. I have made it clear the selection process for which the method was/is designed attempting to adapt it to other universes and in a bear market would have the same consequences as using asprin to treat diabetes.

No not really.
The question that bothers me is this..........can mechanical systems backtested by the software that has been developed fairly claim statistically significant results

The answer already is a resounding no. However I wish to continue through the entire process to see if there are areas where the results are strong. We are at very early days.

The whole point of defining statistically significant results is rather important
There is a world of difference between a true statistically significant result, and one that is not.

Returning to the medical model for illustrative purposes; if a new drug, say Vioxx, was shown to have statistically significant correlation and causation to cardiac pathology, then, Merck could be in serious trouble, if not, then they get off.

If your tested trading methodology has a statistically significant result, that should mean that you are mathematically assured of the result, but, if it is not statistically significant, then you might get the result.

For all the above to be correct, the study design must be of a high methodological quality, otherwise it is mathematically impossible to claim a statistically significant result

The big advantage that mechanical traders enjoy over discretionary traders is a psychological buffer zone engendered by a belief that they are trading a mathematically proven methodology

What if this is a false premise?
Currently it certainly looks incredibly fragile.

Well the population or Universe of stocks used in the test from the BT margin list is 327 stocks and the number of trades tested is over 800 (Possible but due to capital constraints only around 120 trades taken ---from memory I'm at the office). This has been replicated by countless other both using metastock/amibroker and wealthlab. Many select universes well away from the parameters set when the system was initially designed. Other than using far to small timeframes I am yet to recieve a test result showing ruin despite 100s if not 1000s of tests and various universes both in size and criteria.

Total ruin is pretty unlikely, given adequate money management, which I know that you employ. Rather it is a failure to replicate the *success* that the system implies, this could be a major disappointment.

Certainly the timeframe is wholly inadequate, 1996 was the earliest data that was available I believe. The earliest data that I have heard of is 1982, again just nowhere near enough. I have seen examples of just 3/4 yrs being used, this is just delusional and could well engender a blow up if the drawdown was particularly aggressive.

1...ASX stocks do not utilize Market Makers in the same manner as the US
2...ASX stocks have far fewer adjustments of Capitalization structures (stock splits would be an example)
3...ASX stocks do not have anywhere near the Short-selling capabilities (you can short sell, but, it is not an easy thing to do)
4...Differences in liquidity, and thus spreads.
5...Differences in traded Options on stocks
6...Differences in volatility
7...Differences in volume
8...Differences in foreign stocks listed

Absolutely correct but you are presuming that a method should work within its "blueprint" in every situation you place it in.I disagree the system and any system can be designed around any set condition and conditions.The 8 condition "variables" you mention above have little impact on the specifics of the method.

But how do you know that?
The fact is you don't.
The design of the methodology is flawed and invalidates any conclusions that you draw on a statistical basis.

It may be that you are in point of fact correct, but without testing it with a robust statistical model, you and I are both exhibiting a bias, and that is what statistical modeling seeks to eliminate, or marginalize.

The 16 portfolios were started at various times myself and Daryl both traded UTB to its end.As for the other 15 traders using the method and or a hybrid i dont have individual trade results.But i would say that most got a good part of UTB.As start dates will obviously have an impact on which entries are triggered I would expect there to be both same and different stocks in portfolio's over different start dates.

This would imply that it was individual rather than systemic.

The turtle system was never pulled apart or studied to any degree.
All I/We looked at was how it entered and exited as a basis from which to start. The Good bits were that it traded a trend over long time frames.Nothing more complex than that.

I was always under the impression that TT (my mistake) was very similar to the Turtle methodology. As you know they blew it up. It would have been interesting to know if it was an individual or systemic blow-up.

jog on
d998

 

[tech/a]

Duc

I'll answer in more detail when I can.

But from the direction your going your going to be able to prove (and most anything can be proven mathematically) that due to umpteen reasons statistics from backtesting cannot be seen as "Statistically Significant"

I'm not going to argue that.

I'm also not going to stop trading the method either.
If I cant hit a golf ball perfectly everytime I'm not going to give up golf either.If I can hit it consistantly well often enough,completely miss it evry now and again,go around in a respectable score often enough for people to enjoy playing with me and every now and again hit an eagle,I'll keep playing.

Same goes with trading a potfolio.

 

[ducati916]

Moving swiftly on.

Sample Size
One crucial prerequisite prior to embarking upon the *testing* or study is the requirement to perform a sample size, or POWER CALCULATION In the words of D. Altman "a trial should be big enough to have a high chance of detecting, as a statistically significant, a worthwhile effect if it exists, and thus to be reasonably sure that no benefit exists if it is not found in the trial."

In our world, this will apply to profitability.
Intrestingly, and this is currently possibly being quite strongly exhibited by TT currently, what level of difference between *profitability* & *statistically significant profitability*.

Note, that this may not be the same measurement. You can have profitability, even high profitability, but it need not, and quite possibly will be statistically insignificant........viz. unrelated to the study. The obvious variable would be the presence of a strongly trending bull market.

By using a statistical nomogram the individual can prior to running the study calculate how large a sample is required to have in order to detect a moderate, high, or very high true difference between the profitability generated via the design or criteria, and variable market conditions.

The likelihood of detecting a true difference is known as the power of the study. It is common for medical studies to require a power of between 80%-90%. I have not seen any references to this calculation in any of the computer printouts from the software generated results.

jog on
d998

 

[ducati916]

tech/a

But from the direction your going your going to be able to prove (and most anything can be proven mathematically) that due to umpteen reasons statistics from backtesting cannot be seen as "Statistically Significant"

Quite possibly.

I'm also not going to stop trading the method either.

And that was not my expectation, nor wish.
It was simply that when a novice trader asks, how do you make a profit in the market? That the seemingly simple answer "its in the numbers" is trotted out with some caveats, and not promulgated as some mathematic guarantee.

The number of times I have been subjected to the (1+($profit/$loss)*(#wins/#losses)-1) nonsense, as if this was some form of money back guarantee. If an experienced, profitable trader, feels the requirement to offer advice, it should be offered on a true understanding, not some half-baked dimly understood regurgitation of a concept that they have swallowed whole without really doing the work.

Of course you could argue that, caveat emptor, but, hell I'm a nice guy.

jog on
d998

 

[tech/a]

Yes well we could argue statistical significance V data infinitum.
While in a longterm method such as that which we are discussing a larger cleaner data base would be preferable,my logic tells me that the statistical relevance if you like is---

If I set the conditions and the variables and over constant testing both from backtests and forward trading the results are consistantly within "blueprint results" regardless of how accurate they are---as even as un accurate as they could be a profit is found as "expectancy" in the tests and "actual" in the forward trades.If it were to fall outside the blueprint then trading would and should cease.
Until then the expectancy is intact and manifested in profit.

 

[wayneL]

Duc,

The expectancy equation is not nonsense. It is a tool... a measuring stick.

It can be used in an innappropriate way, and I suspect this is what you are refering to. I think "expectancy" is a misnomer in the best use of the equation is not described by that term.

But it is like saying a hammer is nonsense because you don't happen to be a carpenter, even if the hammer is given the wrong name. It is clearly a silly thing to say, nonsense even, and exposes a narrowness of thought common amongst the "learned".

Cheers

 

[tech/a]

If I take on one hand the entry.exit,and stop of T/T and apply it to a say 50 trades it will have an expectancy.
If I apply it to 100 trades same again.
1000 and so on.

As it was and is positive and the Numbers thrown out from 100s of tests give me a number of "blueprints" which fit within acceptable or Not acceptable limits--

My choice due to the confidence in the amount of testing I have done and the design of the method and its moneymanagement techniques applied to it coupled with a set of parameters that I must stay within to remain as "sample" allows me to trade the method until and or if it fails to meet the criteria for positive expectancy/profit.


Its really that simple.
Dont care how accurate it is compared to the standard set by Duc or X.


As for novice traders their confidence and ability to build a plan and trade a method with positive expectancy consistantly and over a long period of time may well lie in their ability and thoroughness of testing.
However to cover every variable and situation for Joe Average is not possible nor NECESSARY in my view.

The future may prove me wrong but in the meantime I'll take whatever it gives me--.

I'm trading the Numbers Duc and when they tell me not to----I'll stop.
Theyve been sufficient for 3.5 yrs so far. Infact all 3 methods have been.

 

[ghotib]

Ghotib.

If you mean TechTrader then there are 100s of pages written on the topic and it is traded live with each weeks results published by Darryl,who first asked me the question on wether a mechanical system could be designed and traded profitably---then the link and weeks of reading are here.

http://lightning.he.net/cgi-bin/suid/~reefcap/ultimatebb.cgi?ubb=forum;f=74

Thanks Tech/a

Don't know if I can just shut up and read for weeks, but I'll try

ghotib

 

[tech/a]

tech/a



The number of times I have been subjected to the (1+($profit/$loss)*(#wins/#losses)-1) nonsense, as if this was some form of money back guarantee. If an experienced, profitable trader, feels the requirement to offer advice, it should be offered on a true understanding, not some half-baked dimly understood regurgitation of a concept that they have swallowed whole without really doing the work.

jog on
d998

Actually Duc, If you/anyone has a positive expectancy model and they trade it within the numbers generated by that model then yes I would guarantee that you would profit. Provided of course that sufficient testing had taken place.I think that the 6 mths testing in the case of T/T has proven to be sufficient. Of course what is sufficient can be argued ad infinitum. I'm not the first or last to come to that conclusion.Stevo has a weekly method that is doing as well as T/T. As has Andrew.

If you can offer evidence to the contrary I'm all eyes.

 

[tech/a]

Trading Plan

This is the "half-baked dimly understood regurgitation of a concept that they have swallowed whole without really doing the work." That I know duc is NOT refering to.
But I am.
We are told constantly that to be successful HAVE A TRADING PLAN and even more dangerously ---DONT deviate from your plan--cause if you do whats the point of having a plan in the first place---Makes sence.

There is a MISSING LINK.
Without knowledge of how your plan performs you could be on the road to ruin with a poorly devised plan.

This is what this discussion is about.

The RESULTS of your plan.
Firstly it MUST have a proven positive expectancy over a period and number of tests that YOU have confidence in.

Where Duc and I differ (In part at least) is the accuracy and sample testing of any method.
I dont believe every concievable condition or variable need be met. I think and so far have proven (The future may prove me wrong),That provided you trade the Numbers of the tested positive expectancy method that have been generated during testing,and cease to trade it if actual trading figures fall outside those generated in testing---you'll make a profit within the parameters returned in testing.