Systematic traders! Hear ye! Hear ye!

[Volker Butzlaff]

Quote from intradaybill:

Not clear enough...

@IntradayBill,

hmm...that's part of the content of my 03-11-11 contribution:

"...or

example 2 (system report with average results):
-------------------------------------------------------------
wins: 4
losses: 4 => trade ratio of 0.50
average profit: 160/4 = 40
average loss: 80/4 = 20 => pay-off ratio of 2.00..."

What's so difficult to understand? I think it's clear...

bye,
zentrader

 

[intradaybill]

Quote from Volker Butzlaff:

@IntradayBill,

hmm...that's part of the content of my 03-11-11 contribution:

"...or

example 2 (system report with average results):
-------------------------------------------------------------
wins: 4
losses: 4 => trade ratio of 0.50
average profit: 160/4 = 40
average loss: 80/4 = 20 => pay-off ratio of 2.00..."

What's so difficult to understand? I think it's clear...

bye,
zentrader

It isn't clear because you are using non-standard terminology. If you have kept up to date, people nowadays call your "trade ratio" win ratio and your "pay-off-ratio" avg. win to avg. loss ratio.

Just Google "pay-off-ratio" and then "average win to average loss ratio" to see what you get. I guess you never bothered.

Nevertheless, I just wanted to make sure we mean the same thing. So, now, you can tell us how you generate the trades for each simulation run, would you? Up to now, you have talked only about some well-known performance ratios.

After you have those ratios, what do you do with them?

In you "seminal" 03-11-11 post you claim:

Quote from Volker Butzlaff:

Imagine you have not 8 trades, but 1,000 trades in your historical system test and simulate this test in 10,000 or more simulation runs (resulting in over 10 millions of trades).

You will get the same MCS results concerning possible profits or (more important) possible DDs - independent if you use example 1 or example 2.

These are unjustifiable statements. You have to justify them. How do you get 1,000 trades? Be specific please. from two ratios, how do you get 1,000 trades?

 

[Volker Butzlaff]

Quote from intradaybill:

...These are unjustifiable statements...

@intradayBill,

normally with a typical system test you ave all information (I hope, i've not to explain a system test result sheet, because some of your questions let me taken to a feeling of beeing in a joke?).

Ok, let's widen the model for our special guest "intradaybill":
1. trade ratio 0.5 (40 winners, 40 losers)
2. pay-off ratio 2.0 (average win 40, average loss 20)
3. e.g. we have 1000 trades in 10 years of EOD trading in this historical test

Assumed the historical max.DD was 75 (as in my little example shown before), what is the benefit of a MCS stress test?

e.g. that a maxDD of 75 is only your "unique" historical experience. A maxDD of 600 is also possible, without changing the quality of your system (profit factor etc.)...

See the screenshot attached.

Hope it helps.

bye,
zentrader

 

[intradaybill]

Quote from Volker Butzlaff:

e.g. that a maxDD of 75 is only your "unique" historical experience. A maxDD of 600 is also possible, without changing the quality of your system (profit factor etc.)...

You say it is possible. It may not be possible depending on the specific system or it may be extremely impossible, like 20 losers in a raw. This is not MCS. This is something you pulled out of who knows what..

Quote from Volker Butzlaff:

normally with a typical system test you ave all information (I hope, i've not to explain a system test result sheet, because some of your questions let me taken to a feeling of beeing in a joke?).

You are so immersed in your own misconceptions that you think people questions are jokes.

Win ratio and avg. win to avg.loss ratio have a unique correspondence only to profit factor. Parameters such as drawdown cannot be mapped uniquely to those ratios. As a result, some of the numbers you produce, and you resist telling how, may not correspond to attainable levels by a specific system.

For example, if you have a trend following system with profit factor 5 and avg. win to avg. loss ratio 2 and a swing trading system with the same parameters, and both have DD =-10% in backtesting, then drawdown -20% is NOTequally likely to occur for those systems. Thus, when you produce your curves, you are misrepresenting one of the systems.

Now, if you don't understand this, I urge you to try. I am telling you, you are no doing MCS. This type of simulation typically involves changing inputs and observing outputs. In your case, the inputs are market changes, not even market data. The output is how the system responds to those changes. It is impossible to perform a valid stress test using just the output.

I also urge you to read the comments in this forum, especially by Howard Bandy

http://tinyurl.com/5wcqepq

In part Two of his post he explains when and how MCS is used

http://tinyurl.com/689l2xu

I have never seen anywhere, in no book, in no forum, no place whatsoever, anyone claiming that changing output performance parameters in ad-hoc ways is a MCS.

I have nothing more to say. Watch your method and especially your attitude. With that attitude, you will never learn because no too many people will be as patience as I was with you. You have a bad attitude.

Bye.

 

[Volker Butzlaff]

Quote from intradaybill:

You say it is possible. It may not be possible depending on the specific system or it may be extremely impossible, like 20 losers in a raw. This is not MCS...

@Intradaybill,

i thought that I talk against a wall. You have confirmed that 100%.

That "extremely impossible" risk is one important value of a risk-oriented MCS stress test. It's the risk, that the common user "intradayBill" estimates as "extremly impossible". This under-estimation of risks is the main reason for imploding trading accounts.

My used methods in the Zen Monte Carlo Simulator software are 100% MCS - independent of your individual opinion ...
... but anyway good luck for your future trading and risk management efforts!

bye,
zentrader

 

[dave4532]

Quote from Volker Butzlaff:

@Intradaybill,

i thought that I talk against a wall. You have confirmed that 100%.

That "extremely impossible" risk is one important value of a risk-oriented MCS stress test. It's the risk, that the common user "intradayBill" estimates as "extremly impossible". This under-estimation of risks is the main reason for imploding trading accounts.

My used methods in the Zen Monte Carlo Simulator software are 100% MCS - independent of your individual opinion ...
... but anyway good luck for your future trading and risk management efforts!

bye,
zentrader

This thread is very interesting.

Having said that, I think I will agree with Bill. It doesn't make sense to consider too extreme possibilities.

I think Bill is saying that some valyes of DD may be too extreme for a particular system. For example, if I have a system that adjusts position size based on some realized DD levels, how could a stress test of the kind you are doing take that into account?

I don't even know if these stress-tests have any meaning at all. Forward tests are more useful IMO.

 

[Volker Butzlaff]

Quote from dave4532:

This thread is very interesting.

Having said that, I think I will agree with Bill. It doesn't make sense to consider too extreme possibilities.

I think Bill is saying that some valyes of DD may be too extreme for a particular system. For example, if I have a system that adjusts position size based on some realized DD levels, how could a stress test of the kind you are doing take that into account?

I don't even know if these stress-tests have any meaning at all. Forward tests are more useful IMO.

Dave,

it's clear that you don't have to trade "real" in this possible MCS simulated DDs. The only thing that the MCS stress test shows is, that these situations are a possible scenario for your trading system concept. You have to accept it and react (position sizing if possible or stop trading), if the DD is your absolute control size.

It's also possible that you as lucky trader never arrive at such an extreme DD point. The MCS shows only the possibility, but it's not the magic "crystal ball".

But your last sentence i don't understand. The MCS stress test is an implicit "forward test" of possible trade sequences based on the system test data.

Or you want to say forward testing with real data (partition back test/out-of-sample test). For this you have the same problem, that you have only one set of historical data. To solve this limit I've implemented the second part of MCS data simulation my software. So test data sets for infinite (forward) tests can be created based on some properties of the historical data and on the settings of a possible future market behaviour, that can be parameterized (scenarios).

bye,
zentrader

 

[intradaybill]

Quote from Volker Butzlaff:

@Intradaybill,

i thought that I talk against a wall. You have confirmed that 100%.

Either that or you cannot comprehend what I am talking about. You should also consider this possibility as viable.

Specifically, in your simulator "flyer" in your website you have an example where you show a min drawdown of 300 and a max drawdown of -5,775.

Question 1: how can drawdown have both minus and plus signs

Question 2: Given the system parameters used, you never explained, although asked repeatedly, numerous times, how you sample the possible values of drawdown.

Question 3: in your "flyer", what do the axes represent? One is profit, the vertical. There are no units there. What is the other axis, the horizontal. Why there are no labels?

Question 4: why does the drawdown become positive as shown in the example? (related to Q1). What is the meaning of positive drawdown?

It seems I have hit a wall asking you reasonable questions you resist answering.

 

[Volker Butzlaff]

Quote from intradaybill:

Either that or you cannot comprehend what I am talking about. You should also consider this possibility as viable.

Specifically, in your simulator "flyer" in your website you have an example where you show a min drawdown of 300 and a max drawdown of -5,775.

Question 1: how can drawdown have both minus and plus signs

Question 2: Given the system parameters used, you never explained, although asked repeatedly, numerous times, how you sample the possible values of drawdown.

Question 3: in your "flyer", what do the axes represent? One is profit, the vertical. There are no units there. What is the other axis, the horizontal. Why there are no labels?

Question 4: why does the drawdown become positive as shown in the example? (related to Q1). What is the meaning of positive drawdown?

It seems I have hit a wall asking you reasonable questions you resist answering.

@intradayBill,

a man - a word.

Quote from intradaybill:
...I have nothing more to say... [/B]

Already forgotten...


Ok - I'll answer your questions. Perhaps it helps...

Q1: the software uses as default a "zero start account" and calculates so-called "account draw downs". So it's possible, that in one simulation run there's never a negative account state. This is a method used in many other sciences or areas, but i know that in trading most people prefer to talk about only a temporate DD, so I updated the software concerning this maxDD behaviour optionally (see additional checkbox in my sample.gif)

Q2: i've always explained that I sample the average trades (trade ratio, pay-off ratio) and get so DDs in all simulation runs. Some DDs may be lower than in your historical "unique" test, some may be higher. And the last are the interesting...

Q3: you speak about this:
http://www.zentrader.de/mcsflyer_e.pdf

Ok, it's easy to understand, or? This is the distribution of MCS simulation runs. You see the distribution of profits (green) and possible DDs (red) in all simulation runs (compressed, because you have far more simulation runs than pixels on your screen). The horizontal axis is the zero axis.

Q4: see Q1


Any other questions?


bye,
zentrader

 

[intradaybill]

Quote from Volker Butzlaff:

Q1: the software uses as default a "zero start account" and calculates so-called "account draw downs". So it's possible, that in one simulation run there's never a negative account state. This is a method used in many other sciences or areas, but i know that in trading most people prefer to talk about only a temporate DD, so I updated the software concerning this maxDD behaviour optionally (see additional checkbox in my sample.gif)

Drawdown cannot have two extreme values with one being negative and the other being positive. Positive drawdown doesn't make sense in any science. Also, there are no equity curves with zero drawdown, except if we are talking straight line equity wth zero curvature everywhere. Even if you have a zero start account, drawdown cannot be positive.

Thus, you are doing something fundamentally wrong, as I suspected from the start. Ask anyone. They will tell you the same thing.

Do not take this as an assault but as an opportunity to improve your analysis. Before that though you must change your attitude. People may be trying to help you while you are behaving in an arrogant way.