Do markets change? Trend Following.

[Trend Following]

Quote from Rimping:

How should we describe your character? Mean? Offensive?

Quote from Rimping:

I read your 'Trendfollowing" with with pleasure.

Thanks.

Quote from Rimping:

"Turtle Trader" less so. Especially the nonsense in chapters 11 and forth about "entrepreneurial skills".

I am always open to hear others' opinions about why some Turtles failed. Do you have a view?

 

[Rimping]

Quote from Trend Following:





I am always open to hear others' opinions about why some Turtles failed. Do you have a view?

Because some were not able to stick to the rules. Some were second guessing if I remember well.

Or do you mean to say: some failed AFTER the turtle experiment and why was that?
Well THAT might have to do with entrepreneurial skills. When you want to attract money for example, a smooth talk, a convincing personality and so on are very usefull.

But with trading it has nothing to do. I know of some people with excellent entrepreneurial skills who failed miserably in the markets. When you have to do with people all those above mentioned characteristics are very uselfull to put things to your hands. But markets are not impressed by "entrepreneurial skills" they are not impressed, they sweep you away just as easily.

 

[Rimping]

I apologize for the distorted sentences in the above post.
But it is still understandable I think.

 

[salvador90]

.

 

[Trend Following]

Quote from Rimping:

I know of some people with excellent entrepreneurial skills who failed miserably in the markets. When you have to do with people all those above mentioned characteristics are very uselfull to put things to your hands. But markets are not impressed by "entrepreneurial skills" they are not impressed, they sweep you away just as easily.

I never said markets were impressed by entrepreneurial skills. I was pointing out why people with the same systems could see some win and some fail.

 

[jack hershey]

Quote from TSGannGalt:

Heh... no one seems to comment on AFG's equity curve so I'll add...

1. The fact that the equity is log-normal, it doesn't provide the true picture of the actual model due to the compounding and it's vulnerable to a variation of a "selection bias" caused by equity curve compression. (Take a look at any index on a log chart and it looks like a very smooth upwards curve until the early 2000s...)

2. Because it is... log normal, there is definitely, at the least, a position sizing scheme included. So we're not sure whether the curve is generated by the risk management implemented or the signal. If AFJ Garner wants to prove a point from a pure "trading system" perspective, he needs to re-generate a risk-adjusted curve.

I have to emphasize that... risk management is a key component common to any kind of trading whether it be trend-following, mean reversion... discretionary... automated... etc. etc. etc.... I mean all trading requires risk management. Though, there is a danger that the RM scheme can be fitted to the curve (just like using MAE/MFE).

What I mean is... I can have a set of bad signals (random entry) and still make the equity curve positive by manipulating the "historical" equity curve. ("Historically possible"... in live trading, it doesn't work that way... Why? *drum roll* Markets Change)

Anyways... I have every reason to be skeptical about the equity curve and I wouldn't look at the continuous line and conclude anything at this point.

--------------------------------------------

AFJ Garner, no offense intended. Your post does serve the point, but everyone else is being a bit too optimistic and conclusive so I had to pull them back and provide them with a sense of perspective...

Most people who do analysis on equity curves do understand the relationship of the "signal" and the "financial management". They are parts of the same equation that represents the equity curve.

Usually people use families of log curves as well. Check out some of those for more personal incite.

 

[thetrendfollowe]

Quote from JezLiberty:

I was wondering... there must be some edge to the entry methodology as well; otherwise you might as well have random entries with trade management. There might be some market structure that cause trend continuation(?)

Actually I have run an edge calculation on a simple N-day breakout entry and it seems to offer a positive edge:
http://bit.ly/8RpMAY
meaning that even if the "hit rate" is <50% the potential gain is greater than potential loss overall

Jez.
You would be how good random entry actually is.
Run a few tests.
I would post some here but on the work computer at the moment.

Hit rate has nothing to do with size of gains or losses.
Average win size in relation to average loss size is the payoff factor.
But you already knew that.
So I must have missed your point?

 

[thetrendfollowe]

Quote from ssb11:

i would comment that i don't believe there is a more misunderstood stat than W%. it can be inversely correlated to profit. in nearly every bit of my research the quickest way to make more money is to lower your W%. any method can increase W% by increasing the width of the initial exit and by increasing the rate at which the stop follows a profitable trade or by decreasing the width of a profit objective. increasing the size of losing positions is another way to increase the W%. in fact one could make a method that had an 80-90% W% very easily.

the one suggestion that i would make to any new trader is to take whatever they are doing and try to decrease the number of winners in favor of the size of those winners.

I agree wholly.

The reason I mentioned W% is because its the only way to quantify or see if there is any edge in the entry.

Of course there doesn't have to be an edge in the entry for a method to have a positive expectancy.

 

[Muskoka Joe]

Quote from thetrendfollowe:

I agree wholly.

The reason I mentioned W% is because its the only way to quantify or see if there is any edge in the entry.

Of course there doesn't have to be an edge in the entry for a method to have a positive expectancy.

=============

I have seen that view expounded before, but never with any evidence that it is true... is there any evidence? The thought that random or worse entries can be tortured to produce winning trades is a glaring invitation to over optimization.

 

[thetrendfollowe]

Quote from Muskoka Joe:
I have seen that view expounded before, but never with any evidence that it is true... is there any evidence?

I've tested random entry myself, though it was years ago, when i was first getting started in systems design.
The results suprised me.

Do you have Amibroker?
Its pretty easy to do it with Ami I think.

I can run a test for you but will probably have to wait until the weekend, very busy this week.

Quote from Muskoka Joe:
thought that random or worse entries can be tortured to produce winning trades is a glaring invitation to over optimization.

How so?
If you run a system over the S&P500 for example, and say
RandomEntry:=0.2

Or however it was (cant remember the code) which means enter every one in five stocks alphabetically starting from the test start date and then put in your money management and your trade management (stops, exits) and then see the results.

It will suprise you.

Though it shouldn't.

The only reason trend following systems make money is the fact that they cut their losses and let their winners run. And this is all to do with stops and exits. And nothing to do with the entry.

Obviously, the situation is different with short term mean reversion, as you are relying on that high win% as you don't let winners run as much.